An intelligent control method for industrial production systems based on time series prediction

Abstract

The present invention relates to the technical field of time series prediction, and in particular to an intelligent control method for an industrial production system based on time series prediction, including collecting historical time series data to improve data quality; using preprocessed data as a training set to construct a time series prediction model, and using an algorithm to adjust model parameters to accurately capture time series characteristics; predicting controlled variables at future time points through the time series prediction model, and formulating an optimal control strategy; performing weighted operations based on prediction results of different controlled variables, introducing process variables to correct and adjust reference values ​​of manipulated variables, and realizing multivariable coordinated control; obtaining actual values ​​of controlled variables in real time, combining models to perform error analysis on predicted values ​​of controlled variables at future time points, and dynamically adjusting. The present invention is used to solve the technical problems of inaccurate data prediction and unoptimized control strategies in industrial production systems in existing solutions.

Description

Technical Field

[0001] The present invention relates to the technical field of time series prediction, and in particular to an intelligent control method for an industrial production system based on time series prediction. Background Art

[0002] Time series forecasting, whose full name is time series forecasting, is a technology for forecasting based on time series data. It refers to using historical time series data for statistics and analysis based on the development trends and changing laws of objective things, digging out the changing laws of things, and predicting the values ​​or trends of specific objects in the future based on these laws.

[0003] With the continuous development of science and technology, time series prediction technology has been widely used in many fields, including finance, transportation, energy, and medical care. The current control of industrial production systems still faces some challenges, including the following aspects: data quality and integrity. In the industrial production process, due to equipment failure, sensor damage, or data transmission problems, historical time series data may be missing or outliers, which increases the difficulty of data preprocessing and affects the accuracy and reliability of the subsequent time series prediction model; model selection and training problems. Different industrial production systems have different characteristics and requirements. Selecting a suitable time series prediction model is still a problem. Model training requires a lot of computing resources and time, and there is a risk of overfitting, which affects the generalization ability of the model; the challenge of multivariable coordinated control. Industrial production systems often contain multiple controlled variables and manipulated variables, which influence and restrict each other. How to formulate an effective control strategy and achieve coordinated control among multiple variables is a complex problem. In response to these challenges, corresponding solutions and technical means need to be adopted to improve the control level and operation efficiency of industrial production systems. Summary of the invention

[0004] The purpose of the present invention is to solve the problems in the background technology and to propose an intelligent control method for an industrial production system based on time series prediction.

[0005] In order to achieve the above object, the present invention adopts the following technical solutions:

[0006] An intelligent control method for an industrial production system based on time series prediction, comprising:

[0007] Step 1: Collect the historical time series data of the industrial production system and pre-process the historical time series data; the historical time series data of the industrial production system comes from the sensors, controllers, databases, etc. of the industrial production system, and records the historical values ​​of controlled variables (such as temperature, pressure, flow, etc.), manipulated variables (such as valve opening, motor speed, etc.) and disturbance variables (such as raw material quality changes, ambient temperature fluctuations, etc.);

[0008] Step 2: Build a time series prediction model, and use the preprocessed historical time series data as a training data set to train the time series prediction model;

[0009] Step 3: Use the trained time series prediction model to predict the value of the controlled variable at a future time point, and continuously adjust the value of the manipulated variable through an iterative optimization process so that the predicted value of the controlled variable at a future time point is close to the predetermined target value, thereby finding the optimal control strategy;

[0010] Step 4: For the manipulated variables that affect multiple controlled variables, weighted operations are performed based on the prediction results of different controlled variables; process variables are introduced to correct and adjust the reference values ​​of the manipulated variables;

[0011] Step 5: Obtain the actual value of the controlled variable in real time through the sensor, perform error analysis on the predicted value of the controlled variable at future time points in combination with the time series prediction model, and make dynamic adjustments based on the error analysis results.

[0012] It should be noted that the application object of the intelligent control method of an industrial production system based on time series prediction proposed in the present invention can be data analysis of various time series data generated in the industrial production process. These data include but are not limited to historical records of controlled variables and manipulated variables such as temperature, pressure, flow, speed, motor current, valve opening, and data on disturbance factors that may affect these variables (such as raw material quality, ambient temperature, etc.). Specifically, it can be to comprehensively monitor and accurately analyze the controlled variables, manipulated variables, and disturbance variables through time series prediction technology, and use the constructed time series prediction model to accurately predict the controlled variable values ​​at future time points, providing a scientific basis for the formulation of control strategies, and through iterative optimization. ization process, constantly adjust the value of the manipulated variable, so that the predicted value of the controlled variable is close to the predetermined target value, thereby realizing the formulation of the optimal control strategy. For the manipulated variables that affect multiple controlled variables, the coordinated control between multiple variables is realized through weighted operations and the introduction of process variables. At the same time, combined with the real-time data obtained by the sensor and the error analysis results, the various manipulated variables and control strategies in the industrial production system are dynamically adjusted, realizing real-time monitoring and intelligent control of the industrial production process, improving the response speed and flexibility of the system, enhancing the stability and reliability of the industrial production system, and helping to formulate more reasonable and effective production plans and control strategies, providing strong support for the sustainable development of enterprises.

[0013] Furthermore, the historical time series data of the industrial production system is collected, and the process of preprocessing the historical time series data includes:

[0014] Acquire historical time series data through an industrial production system, wherein the historical time series data includes controlled variables, manipulated variables, and disturbance variables;

[0015] Perform data cleaning on the collected historical time series data, where data cleaning includes outlier detection and missing value filling;

[0016] It can be understood that outlier detection refers to identifying points that deviate significantly from the overall data distribution by calculating the statistics of each data point (such as mean, standard deviation). These points are caused by measurement errors, equipment failures, or abnormal events and need to be removed or replaced.

[0017] Rearrange all preprocessed data in time series , as the final result of preprocessing.

[0018] Furthermore, the process of outlier detection for the collected historical time series data includes:

[0019] The Z-Score method is used to detect outliers: a single observation of any variable is marked as , then there is an outlier detection formula:

[0020] In the formula, represents the outlier detection value; represents the mean of any variable, represents the standard deviation of any variable; if , then it is believed that is an outlier, among which, Represents the preset outlier detection threshold.

[0021] Furthermore, the process of filling missing values ​​for the collected historical time series data includes:

[0022] Use linear interpolation to fill missing values: For any variable In time If the value of Indicates the missing values ​​to be filled, then use and Interpolate the values ​​at the moment:

[0023] In the formula, Indicates time Number.

[0024] Furthermore, a time series prediction model is constructed, and the preprocessed historical time series data is used as a training data set. The process of training the time series prediction model includes:

[0025] Choosing a time series forecasting model ,definition Network Layers , the number of hidden units in each layer ;in, is a structural parameter that defines the depth of the model;

[0026] Extracting manipulated variables from preprocessed historical time series data , the controlled variable , and the disturbance variable ;

[0027] The preprocessed historical time series data Divide into training set and validation set ;

[0028] Assume that the preprocessed historical time series data Include time points, of which the first 80% are used as training sets and the remaining 20% ​​are used as validation sets;

[0029] make For time point The input vector contains the information of the manipulated variables and the disturbance variables, that is, ;at the same time, For time point The controlled variable target value;

[0030] The training process of the time series prediction model is as follows:

[0031] B1. Setting model parameters ; Among them, the model parameters include The weight matrix and bias vector of ;

[0032] B2. At each time point On, through Layer calculation prediction value :

[0033] In the formula, Respectively represent the hidden state and memory unit state at the previous time point;

[0034] B3. Calculate the loss function using the mean square error formula :

[0035] In the formula, Indicates the total number of time points;

[0036] B4. Update model parameters using Adam optimizer :

[0037] In the formula, denote the unbias-corrected first- and second-order moment estimates, respectively; represents the hyperparameters used to control the decay rates of the first-order moment and the second-order moment respectively; Represents the loss function About Model Parameters The gradient of Respectively First-order and second-order moment estimates after correction for moment bias; Iterate to The model parameter values ​​at time instant; Represents the model parameter value for the next iteration; Represents the learning rate, which is used to control the step size of model parameter update; represents a small constant to avoid division by zero, ensuring that the denominator is not zero;

[0038] B5. Evaluate model performance on the validation set and adjust model parameters as needed , hyperparameters , learning rate , or model structure parameters and ;

[0039] Repeat steps B2-B5 until the model's performance on the validation set reaches a predetermined standard or the maximum number of iterations is reached;

[0040] It is understandable that through multiple iterative updates, the model parameters gradually move towards the direction of minimizing the loss function and finally achieve the optimization goal; the above steps construct and train a The time series prediction model is used to predict the controlled variables in industrial production systems and provide a basis for intelligent control.

[0041] Furthermore, the trained time series prediction model is used to predict the controlled variable value at a future time point, and the value of the manipulated variable is continuously adjusted through an iterative optimization process so that the predicted value of the controlled variable at a future time point is close to the predetermined target value, thereby finding the optimal control strategy. The process includes:

[0042] Use the trained The time series prediction model is marked as ;

[0043] Obtaining manipulated variables and disturbance variable sequences to predict controlled variables at future time points;

[0044] Setting the controlled variable Target value ;

[0045] pass Get the predicted value of the controlled variable at a future time point ;

[0046] definition is the manipulated variable vector; represents the index of the manipulated variable vector;

[0047] Setting manipulated variables The initial value and optimization range of

[0048] The predicted value is close to the target value as the optimization goal, and the objective function is defined :

[0049] In the formula, Represents the total number of predicted future time points; The number representing the predicted future time point;

[0050] Use the gradient descent algorithm to optimize the manipulated variables and update the manipulated variables , to minimize :

[0051] In the formula, Indicates The value of the manipulated variable at the iteration; Indicates the operation variable The index of the iteration; Represents the learning rate or step size; Represents the objective function For manipulated variables The gradient of

[0052] In the optimization process, by manipulating the variables Minimum and maximum limits are imposed to ensure that the actual control constraints are met; the actual constraint control of the manipulated variable is expressed as:

[0053] In the formula, Represents the manipulated variables Minimum and maximum limits for ;

[0054] For the optimization termination condition, when the manipulated variable When the update amount is less than the preset update amount threshold or reaches the maximum number of iterations, the optimization process is stopped; the termination condition of the optimization process is expressed as:

[0055] In the formula, Indicates the preset update amount threshold; Indicates the maximum number of iterations, that is, the optimization process terminates when this number is reached;

[0056] It can be understood that as long as any of the conditions that the update amount is less than the threshold or the maximum number of iterations is reached, the optimization process will stop immediately;

[0057] Use the optimized manipulated variables Update control strategies and apply them in industrial production systems.

[0058] Furthermore, for the manipulated variables that affect multiple controlled variables, a weighted operation is performed based on the prediction results of different controlled variables; the process of introducing the process variable to correct and adjust the reference value of the manipulated variable includes:

[0059] Assume that controlled variable, denoted as , each controlled variable at a future time point The predicted values ​​are ;in, represents the number of controlled variables; Indicates the current time point, that is, the specific time when the prediction is made; Represents the time interval between the future time point and the current time point. is any positive real number or real number;

[0060] exist manipulated variables, denoted as ;in, Indicates the number of manipulated variables; some manipulated variables may affect multiple controlled variables simultaneously;

[0061] For each controlled variable , define a corresponding weight , used to reflect The relative importance of the control objectives; among them, Represents an index, used for traversal controlled variable , for each , each corresponds to a controlled variable and its predicted value ;

[0062] Calculate weighted prediction error , as the weighted sum of the prediction errors of all controlled variables:

[0063] In the formula, express At a future time Expected value;

[0064] Introducing process variables ; These variables may affect the optimal setting of the manipulated variables;)

[0065] According to the current value of the process variable, the reference value of the manipulated variable is adjusted using the correction function:

[0066] In the formula, Represents the manipulated variable The initial test reference value; Indicates the corrected value; is the correction function; Represents the index of the manipulated variable, used for traversal Manipulated variables , for each Each corresponds to a manipulated variable , which may be adjusted by an optimization algorithm to minimize the weighted prediction error .

[0067] Furthermore, the actual value of the controlled variable is obtained in real time through the sensor, and the error analysis of the predicted value of the controlled variable at a future time point is performed in combination with the time series prediction model, and the process of dynamic adjustment according to the error analysis result includes:

[0068] Obtain the actual value of the controlled variable monitored in real time , and calculate the predicted value with the time series forecasting model Error :

[0069] ; In the formula, Indicates the current time point of real-time monitoring; Indicates the next time point of real-time monitoring;

[0070] When performing error analysis between the actual value and the predicted value of the controlled variable, set the error threshold , determine whether it exceeds the set error threshold, so as to obtain the error analysis result; among which, if the calculated error value Exceeding the set error threshold , then dynamically adjust each manipulated variable in the industrial production system:

[0071] In the formula, Indicates that according to the error value The new value of the manipulated variable after dynamic adjustment; represents the manipulated variable value of the time series forecasting model based on the information at the previous time point; Represents the gain coefficient of the controller, which determines the speed and intensity of the system's response to the error. A controller is a device or algorithm that receives the error signal between the system output (i.e., the actual value of the controlled variable) and the model's predicted value, and then outputs a control signal (the change in the manipulated variable) to adjust the system's behavior so that the output is as close to the set target value as possible;

[0072] Each dynamic adjustment will automatically trigger an early warning prompt and feedback to technical personnel for review;

[0073] It is understandable that whenever a parameter, status or configuration within the system changes, the system will immediately recognize the change and automatically trigger an early warning prompt; after the early warning prompt is triggered, the system will feedback this information to the responsible technician, who will immediately review the change to determine whether it is in line with expectations, whether it is safe, or whether further intervention is required.

[0074] Compared with the existing technology, the advantages of the intelligent control method of an industrial production system based on time series prediction provided by the present invention are:

[0075] The present invention collects historical time series data of industrial production systems, pre-processes the historical time series data, ensures that the training data of the model is representative and accurate, avoids model deviations caused by data quality problems, improves the efficiency and automation of data processing, and reduces errors caused by manual intervention; by constructing a time series prediction model, and inputting the pre-processed historical time series data as a training data set into the model, the time series prediction model is trained, thereby constructing a model that can accurately predict the values ​​of controlled variables at future time points, providing a scientific basis for the formulation of control strategies, improving the generalization ability and robustness of the model, and enabling it to cope with prediction tasks under different working conditions and disturbance conditions;

[0076] The present invention predicts the controlled variable value at a future time point by using a trained time series prediction model, and continuously adjusts the value of the manipulated variable through an iterative optimization process, so that the predicted value of the controlled variable at a future time point is close to the predetermined target value, thereby finding the optimal control strategy, realizing accurate control of the controlled variable, and improving the stability and efficiency of the production process; by weighting the manipulated variables that affect multiple controlled variables according to the prediction results of each controlled variable, introducing process variables to correct and adjust the target value of the manipulated variable, realizing coordinated control among multiple variables, improving the stability and coordination of the entire production system, and reducing production fluctuations and losses caused by improper control of a single variable;

[0077] The present invention obtains the actual value of the controlled variable in real time through the sensor, performs error analysis on the predicted value of the controlled variable at a future time point in combination with the time series prediction model, and makes dynamic adjustments based on the error analysis results to ensure that the intelligent adjustment control always remains in the optimal state, thereby improving the flexibility and response speed of the production process.

[0078] In summary, the present invention collects and preprocesses historical time series data, constructs and trains a time series prediction model, uses the model to make future predictions, and combines iterative optimization and real-time adjustment strategies to achieve precise control of controlled variables in the industrial production process, significantly improves the accuracy, depth and breadth of time series data analysis, can comprehensively capture data changes, and ensure that the industrial production system is always in the optimal operating state, thereby enhancing the stability and reliability of the industrial production system and ensuring the normal implementation of a subsequent intelligent control method for industrial production systems based on time series prediction. BRIEF DESCRIPTION OF THE DRAWINGS

[0079] Figure 1 The present invention is a flowchart of an intelligent control method for an industrial production system based on time series prediction. DETAILED DESCRIPTION

[0080] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the implementation regulations described 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 creative work are within the scope of protection of the present invention.

[0081] Reference Figure 1 , an intelligent control method for an industrial production system based on time series prediction, comprising:

[0082] Step 1: Collect the historical time series data of the industrial production system and pre-process the historical time series data; the historical time series data of the industrial production system comes from the sensors, controllers, databases, etc. of the industrial production system, and records the historical values ​​of controlled variables (such as temperature, pressure, flow, etc.), manipulated variables (such as valve opening, motor speed, etc.) and disturbance variables (such as raw material quality changes, ambient temperature fluctuations, etc.);

[0083] Step 2: Build a time series prediction model, and use the preprocessed historical time series data as a training data set to train the time series prediction model;

[0084] Step 3: Use the trained time series prediction model to predict the value of the controlled variable at a future time point, and continuously adjust the value of the manipulated variable through an iterative optimization process so that the predicted value of the controlled variable at a future time point is close to the predetermined target value, thereby finding the optimal control strategy;

[0085] Step 4: For the manipulated variables that affect multiple controlled variables, weighted operations are performed based on the prediction results of different controlled variables; process variables are introduced to correct and adjust the reference values ​​of the manipulated variables;

[0086] Step 5: Obtain the actual value of the controlled variable in real time through the sensor, perform error analysis on the predicted value of the controlled variable at future time points in combination with the time series prediction model, and make dynamic adjustments based on the error analysis results.

[0087] It should be noted that the application object of the intelligent control method of an industrial production system based on time series prediction proposed in the present invention can be data analysis of various time series data generated in the industrial production process. These data include but are not limited to historical records of controlled variables and manipulated variables such as temperature, pressure, flow, speed, motor current, valve opening, and data on disturbance factors that may affect these variables (such as raw material quality, ambient temperature, etc.). Specifically, it can be to comprehensively monitor and accurately analyze the controlled variables, manipulated variables, and disturbance variables through time series prediction technology, and use the constructed time series prediction model to accurately predict the controlled variable values ​​at future time points, providing a scientific basis for the formulation of control strategies, and through iterative optimization. ization process, constantly adjust the value of the manipulated variable, so that the predicted value of the controlled variable is close to the predetermined target value, thereby realizing the formulation of the optimal control strategy. For the manipulated variables that affect multiple controlled variables, the coordinated control between multiple variables is realized through weighted operations and the introduction of process variables. At the same time, combined with the real-time data obtained by the sensor and the error analysis results, the various manipulated variables and control strategies in the industrial production system are dynamically adjusted, realizing real-time monitoring and intelligent control of the industrial production process, improving the response speed and flexibility of the system, enhancing the stability and reliability of the industrial production system, and helping to formulate more reasonable and effective production plans and control strategies, providing strong support for the sustainable development of enterprises.

[0088] See also Figure 1 The present invention provides an intelligent control method for an industrial production system based on time series prediction. In the step 1, the historical time series data of the industrial production system is collected, and the step of preprocessing the historical time series data includes:

[0089] Step 101: Acquire historical time series data through an industrial production system, wherein the historical time series data includes controlled variables, manipulated variables, and disturbance variables;

[0090] Step 102: Perform data cleaning on the collected historical time series data, wherein data cleaning includes outlier detection and missing value filling;

[0091] In step 102, outlier detection refers to identifying points that significantly deviate from the overall data distribution by calculating the statistics of each data point (such as mean, standard deviation). These points are caused by measurement errors, equipment failures or abnormal events and need to be removed or replaced. The specific steps of outlier detection and missing value filling are as follows:

[0092] The Z-Score method is used to detect outliers: a single observation of any variable is marked as , then there is an outlier detection formula:

[0093] In the formula, represents the outlier detection value; represents the mean of any variable, represents the standard deviation of any variable; if , then it is believed that is an outlier, among which, Represents the preset outlier detection threshold;

[0094] For example, the temperature data in the controlled variable For example, there is a sequence of temperature observations {15, 16, 17, 15, 14, 200, 16, 17, 15, 16}, and these data are recorded in a stable state; use the outlier detection formula to calculate (taking the value 200 as an example):

[0095] In the formula, The number representing this set of temperature observations;

[0096] Preset outlier detection threshold If is 2, , therefore, the value 200 is considered an outlier;

[0097] Use linear interpolation to fill missing values: For any variable In time If the value of Indicates the missing values ​​to be filled, then use and Interpolate the values ​​at the moment:

[0098] In the formula, Indicates time Serial number;

[0099] For example, valve opening data sequence ,in, Indicates missing values ​​that need to be filled;

[0100] If the time differences are equal, and , The values ​​of are 52 and 54 respectively;

[0101] According to the linear interpolation formula, the missing values ​​after calculation ;

[0102] Therefore, the missing values ​​are filled with 53;

[0103] Step 103: re-arrange all pre-processed data in time series to obtain , as the final result of preprocessing.

[0104] See also Figure 1 The present invention provides an intelligent control method for an industrial production system based on time series prediction. In the step 2, a time series prediction model is constructed, and the preprocessed historical time series data is used as a training data set. The steps of constructing the time series prediction model include:

[0105] Step 201: Select a time series prediction model ,definition Network Layers , the number of hidden units in each layer ;in, is a structural parameter that defines the depth of the model;

[0106] Step 202: Extracting manipulated variables from preprocessed historical time series data , the controlled variable , and the disturbance variable ;

[0107] Step 203: preprocess the historical time series data Divide into training set and validation set ;

[0108] Step 204: Assume that the preprocessed historical time series data Include time points, of which the first 80% are used as training sets and the remaining 20% ​​are used as validation sets;

[0109] Step 205: For time point The input vector contains the information of the manipulated variables and the disturbance variables, that is, ;at the same time, For time point The controlled variable target value;

[0110] Step 206: The time series prediction model training process is as follows:

[0111] B1. Setting model parameters ; Among them, the model parameters include The weight matrix and bias vector of ;

[0112] B2. At each time point On, through Layer calculation prediction value :

[0113] In the formula, Respectively represent the hidden state and memory unit state at the previous time point;

[0114] B3. Calculate the loss function using the mean square error formula :

[0115] In the formula, Indicates the total number of time points;

[0116] B4. Update model parameters using Adam optimizer :

[0117] In the formula, denote the unbias-corrected first- and second-order moment estimates, respectively; represents the hyperparameters used to control the decay rates of the first-order moment and the second-order moment respectively; Represents the loss function About Model Parameters The gradient of Respectively First-order and second-order moment estimates after correction for moment bias; Iterate to The model parameter values ​​at time instant; Represents the model parameter value for the next iteration; Represents the learning rate, which is used to control the step size of model parameter update; represents a small constant to avoid division by zero, ensuring that the denominator is not zero;

[0118] B5. Evaluate model performance on the validation set and adjust model parameters as needed , hyperparameters , learning rate , or model structure parameters and ;

[0119] Repeat steps B2-B5 until the model's performance on the validation set reaches a predetermined standard or the maximum number of iterations is reached;

[0120] In step 206, through multiple iterations of updating, the model parameters gradually move towards the direction of minimizing the loss function, and finally reach the optimization goal; the above steps construct and train a The time series prediction model is used to predict the controlled variables in industrial production systems and provide a basis for intelligent control.

[0121] See also Figure 1 The present invention provides an intelligent control method for an industrial production system based on time series prediction. In step three, the value of a controlled variable at a future time point is predicted using a trained time series prediction model, and the value of the manipulated variable is continuously adjusted through an iterative optimization process so that the predicted value of the controlled variable at the future time point is close to a predetermined target value, thereby finding the optimal control strategy. The steps include:

[0122] Step 301: Use the trained The time series prediction model is marked as ;

[0123] Step 302: Obtain a sequence of manipulated variables and disturbance variables to predict controlled variables at future time points;

[0124] Step 303: Setting the controlled variable Target value ;

[0125] Step 304: Pass Get the predicted value of the controlled variable at a future time point ;

[0126] Step 305: Definition is the manipulated variable vector; represents the index of the manipulated variable vector;

[0127] Step 306: Set manipulated variables The initial value and optimization range of

[0128] Step 307: Taking the predicted value close to the target value as the optimization goal, define the objective function :

[0129] In the formula, Represents the total number of predicted future time points; The number representing the predicted future time point;

[0130] Step 308: Optimize the manipulated variables using the gradient descent algorithm and update the manipulated variables. , to minimize :

[0131] In the formula, Indicates The value of the manipulated variable at the iteration; Indicates the operation variable The index of the iteration; Represents the learning rate or step size; Represents the objective function For manipulated variables The gradient of

[0132] Step 309: During the optimization process, by manipulating the variables Minimum and maximum limits are imposed to ensure that the actual control constraints are met; the actual constraint control of the manipulated variable is expressed as:

[0133] In the formula, Represents the manipulated variables Minimum and maximum limits for ;

[0134] Step 310: For the optimization termination condition, when the manipulated variable When the update amount is less than the preset update amount threshold or reaches the maximum number of iterations, the optimization process is stopped; the termination condition of the optimization process is expressed as:

[0135] In the formula, Indicates the preset update amount threshold; Indicates the maximum number of iterations, that is, the optimization process terminates when this number is reached;

[0136] In step 310, as long as any one of the conditions of the update amount being less than the threshold or the maximum number of iterations being reached is met, the optimization process will stop immediately;

[0137] Step 311: Use the optimized manipulated variables Update control strategies and apply them in industrial production systems.

[0138] See also Figure 1 The present invention provides an intelligent control method for an industrial production system based on time series prediction. In step 4, for a manipulated variable affecting multiple controlled variables, a weighted operation is performed based on the prediction results of different controlled variables; and the step of introducing a process variable to correct and adjust a reference value of the manipulated variable comprises:

[0139] Step 401: Assume that controlled variable, denoted as , each controlled variable at a future time point The predicted values ​​are ;in, represents the number of controlled variables; Indicates the current time point, that is, the specific time when the prediction is made; Represents the time interval between the future time point and the current time point. is any positive real number or real number;

[0140] Step 402: Existence manipulated variables, denoted as ;in, represents the number of manipulated variables;

[0141] In step 402, some manipulated variables may affect multiple controlled variables simultaneously;

[0142] Step 403: For each controlled variable , define a corresponding weight , used to reflect The relative importance of the control objectives; among them, Represents an index, used for traversal controlled variable , for each , each corresponds to a controlled variable and its predicted value ;

[0143] Step 404: Calculate weighted prediction error , as the weighted sum of the prediction errors of all controlled variables:

[0144] In the formula, express At a future time Expected value;

[0145] Step 405: Introduce process variables ,

[0146] In step 405, these variables may affect the optimal setting of the manipulated variables;

[0147] Step 406: According to the current value of the process variable, the reference value of the manipulated variable is adjusted using the correction function:

[0148] In the formula, Represents the manipulated variable The initial test reference value; Indicates the corrected value; is the correction function; Represents the index of the manipulated variable, used for traversal Manipulated variables , for each Each corresponds to a manipulated variable , which may be adjusted by an optimization algorithm to minimize the weighted prediction error .

[0149] See also Figure 1 The present invention provides an intelligent control method for an industrial production system based on time series prediction. In step 5, the actual value of the controlled variable is obtained in real time by a sensor, and the predicted value of the controlled variable at a future time point is analyzed by combining a time series prediction model, and the steps of dynamically adjusting according to the error analysis result include:

[0150] Step 501: Obtain the actual value of the controlled variable monitored in real time , and calculate the predicted value with the time series forecasting model Error :

[0151] ; In the formula, Indicates the current time point of real-time monitoring; Indicates the next time point of real-time monitoring;

[0152] Step 502: When performing error analysis between the actual value and the predicted value of the controlled variable, set the error threshold , determine whether it exceeds the set error threshold, so as to obtain the error analysis result; among which, if the calculated error value Exceeding the set error threshold , then dynamically adjust the various manipulated variables in the industrial production system to ensure that the industrial production system operates stably in the best state:

[0153] In the formula, Indicates that according to the error value The new value of the manipulated variable after dynamic adjustment; represents the manipulated variable value of the time series forecasting model based on the information at the previous time point; Represents the gain coefficient of the controller, which determines the speed and intensity of the system's response to the error. A controller is a device or algorithm that receives the error signal between the system output (i.e., the actual value of the controlled variable) and the model's predicted value, and then outputs a control signal (the change in the manipulated variable) to adjust the system's behavior so that the output is as close to the set target value as possible;

[0154] Step 503: Each dynamic adjustment will automatically trigger an early warning prompt and provide feedback to the technical staff for review;

[0155] In step 503, whenever a parameter, state or configuration in the system changes, the system will immediately identify the change and automatically trigger an early warning prompt; after the early warning prompt is triggered, the system will feed back this information to the responsible technician, who will immediately review the change to determine whether it is in line with expectations, whether it is safe, or whether further intervention is required.

[0156] In the embodiment of the present invention, by comprehensively collecting historical time series data related to process control in the industrial production system, including controlled variables (such as temperature, pressure, flow, etc.), manipulated variables (such as valve opening, motor speed, etc.) and disturbance variables (such as raw material quality changes, ambient temperature fluctuations, etc.), the collected original time series data is cleaned to eliminate abnormal values ​​and missing values ​​in the data, improve data quality, and provide a reliable data basis for subsequent modeling. According to the characteristics and needs of the industrial production system, a suitable time series prediction model is selected, and the pre-processed historical time series data is used as a training data set and input into the model. The model parameters are adjusted through an iterative optimization algorithm (such as a gradient descent method) so that the model can accurately capture the time series characteristics and laws in the data. By using the trained time series prediction model to predict future time points The controlled variable value is predicted, and the value of the manipulated variable is continuously adjusted by the iterative optimization process, so that the predicted value of the controlled variable at the future time point is as close to the predetermined target value as possible, so as to realize the precise control of the controlled variable. For the manipulated variables that affect multiple controlled variables, a weighted operation is performed based on the prediction results of different controlled variables to comprehensively consider the importance of each controlled variable, and the reference value of the manipulated variable is corrected and adjusted by introducing the process variable to more accurately reflect the changes in the actual production process. The actual value of the controlled variable is obtained in real time by the sensor, and the error analysis of the predicted value of the controlled variable at the future time point is performed in combination with the time series prediction model to evaluate whether the actual industrial production plan meets the target requirements. According to the error analysis results, the control strategy of the actual industrial production process is dynamically adjusted to improve the prediction accuracy and control effect. In summary, the examples of the present invention involve data processing, comprehensive analysis and intelligent optimization decision-making, and solve the technical problems of inaccurate data prediction and unoptimized control strategy of the industrial production system in the existing scheme. In actual situations, more data and context information may be needed to make specific decisions and optimization schemes.

[0157] In addition, the formulas involved in the above are all calculated by removing dimensions and taking their numerical values. They are a formula that is closest to the actual situation obtained by collecting a large amount of data and performing software simulation. The proportional coefficient in the formula and the various preset thresholds in the analysis process are set by technical personnel in this field according to actual conditions or obtained by simulating a large amount of data; the size of the proportional coefficient is to quantify each parameter to obtain a specific value for subsequent comparison. The size of the proportional coefficient depends on the amount of sample data and the preliminary setting of the corresponding processing coefficient for each group of sample data by technical personnel in this field; as long as it does not affect the proportional relationship between the parameter and the quantized value.

[0158] Each embodiment in this specification is described in a progressive manner, and the same or similar parts between the embodiments can be referred to each other, and each embodiment focuses on the differences from other embodiments. In particular, for the device embodiment, since it is basically based on the method embodiment, the description is relatively simple, and the relevant parts can be referred to the partial description of the method embodiment.

[0159] For the convenience of description, the above device is described in various units according to their functions. Of course, when implementing the present application, the functions of each unit can be implemented in the same or multiple software and / or hardware.

[0160] It will be appreciated by those skilled in the art that embodiments of the present invention may be provided as methods, systems, or computer program products. Therefore, the present invention may take the form of a complete hardware embodiment, a complete software embodiment, or an embodiment combining software and hardware. Furthermore, the present invention may take the form of a computer program product implemented on one or more computer-usable storage media (including but not limited to disk storage, CD-ROM, optical storage, etc.) containing computer-usable program code.

[0161] The present invention is described with reference to flowcharts and / or block diagrams of methods, devices (systems), and computer program products according to embodiments of the present invention. It should be understood that each process and / or block in the flowchart and / or block diagram, as well as the combination of processes and / or blocks in the flowchart and / or block diagram, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, a special-purpose computer, an embedded processor, or other programmable data processing device to generate a machine, so that the instructions executed by the processor of the computer or other programmable data processing device generate instructions for implementing the processes in the flowchart and / or block diagram. Figure 1 A process or multiple processes and / or boxes Figure 1 A device that provides the functions specified in a block or multiple blocks.

[0162] These computer program instructions may also be stored in a computer-readable memory capable of directing a computer or other programmable data processing device to operate in a specific manner, so that the instructions stored in the computer-readable memory produce an article of manufacture comprising an instruction device, which implements the process Figure 1 A process or multiple processes and / or boxes Figure 1 A function specified in one or more boxes.

[0163] These computer program instructions can also be loaded onto a computer or other programmable data processing device so that a series of operating steps are executed on the computer or other programmable device to produce a computer-implemented process, thereby providing instructions for implementing the process. Figure 1 A process or multiple processes and / or boxes Figure 1 The steps for the functions specified in one or more boxes.

[0164] Secondly: In the drawings of the embodiments disclosed in the present invention, only the structures related to the embodiments disclosed in the present invention are involved, and other structures can refer to the general design. In the absence of conflict, the same embodiment and different embodiments of the present invention can be combined with each other;

[0165] Finally: The above description is only a preferred specific implementation manner of the present invention, but the protection scope of the present invention is not limited thereto. Any technician familiar with the technical field can make equivalent replacements or changes according to the technical scheme and inventive concept of the present invention within the technical scope disclosed by the present invention, which should be covered by the protection scope of the present invention.

Claims

1. An intelligent control method for an industrial production system based on time series prediction, characterized in that: Step 1: Collect historical time series data of the industrial production system and pre-process the historical time series data; wherein the historical time series data includes controlled variables, manipulated variables and disturbance variables; Step 2: Build a time series prediction model, and use the preprocessed historical time series data as a training data set to train the time series prediction model; Step 3: Use the trained time series prediction model to predict the value of the controlled variable at a future time point, and continuously adjust the value of the manipulated variable through an iterative optimization process so that the predicted value of the controlled variable at a future time point is close to the predetermined target value, thereby finding the optimal control strategy; The step three uses the trained time series prediction model to predict the controlled variable value at a future time point, and continuously adjusts the manipulated variable value through an iterative optimization process so that the predicted value of the controlled variable at a future time point is close to the predetermined target value, thereby finding the optimal control strategy. The process includes: Use the trained The time series prediction model is marked as ; Obtaining manipulated variables and disturbance variable sequences to predict controlled variables at future time points; Setting the controlled variable Target value ; pass Get the predicted value of the controlled variable at a future time point ; definition is the manipulated variable vector; where represents the index of the manipulated variable vector; Setting manipulated variables The initial value and optimization range of The predicted value is close to the target value as the optimization goal, and the objective function is defined : In the formula, Represents the total number of predicted future time points; The number representing the predicted future time point; Use the gradient descent algorithm to optimize the manipulated variables and update the manipulated variables , to minimize : In the formula, Indicates The value of the manipulated variable at the iteration; Indicates the operation variable The index of the iteration; Represents the learning rate or step size; Represents the objective function For manipulated variables The gradient of In the optimization process, by manipulating the variables Minimum and maximum limits are imposed to ensure that the actual control constraints are met; the actual constraint control of the manipulated variable is expressed as: In the formula, Represents the manipulated variables Minimum and maximum limits for ; For the optimization termination condition, when the manipulated variable When the update amount is less than the preset update amount threshold or reaches the maximum number of iterations, the optimization process is stopped; the termination condition of the optimization process is expressed as: In the formula, Indicates the preset update amount threshold; Indicates the maximum number of iterations, that is, the optimization process terminates when this number is reached; Use the optimized manipulated variables Update control strategies for application in industrial production systems; Step 4: For the manipulated variables that affect multiple controlled variables, weighted operations are performed based on the prediction results of different controlled variables; process variables are introduced to correct and adjust the reference values ​​of the manipulated variables; The fourth step, for the manipulated variables affecting multiple controlled variables, performs a weighted operation based on the prediction results of different controlled variables; and introduces process variables to correct and adjust the reference values ​​of the manipulated variables, includes: Assume that controlled variable, denoted as , each controlled variable at a future time point The predicted values ​​are ;in, represents the number of controlled variables; Indicates the current time point; Represents the time interval between the future time point and the current time point. is any positive real number or real number; exist manipulated variables, denoted as ;in, represents the number of manipulated variables; For each controlled variable , define a corresponding weight , used to reflect The relative importance of the control objectives; among them, Represents an index, used for traversal controlled variable ; Calculate weighted prediction error , as the weighted sum of the prediction errors of all controlled variables: In the formula, express At a future time Expected value; Introducing process variables ; According to the current value of the process variable, the reference value of the manipulated variable is adjusted using the correction function: In the formula, Represents the manipulated variable The initial test reference value; Indicates the corrected value; is the correction function; Represents the index of the manipulated variable, used for traversal Manipulated variables ; Step 5: Obtain the actual value of the controlled variable in real time through the sensor, perform error analysis on the predicted value of the controlled variable at the future time point in combination with the time series prediction model, and make dynamic adjustments based on the error analysis results; The process of obtaining the actual value of the controlled variable in real time through the sensor, performing error analysis on the predicted value of the controlled variable at a future time point in combination with the time series prediction model, and making dynamic adjustments based on the error analysis results includes: Obtain the actual value of the controlled variable monitored in real time , and calculate the predicted value with the time series forecasting model Error : ; In the formula, Indicates the current time point of real-time monitoring; Indicates the next time point of real-time monitoring; When performing error analysis between the actual value and the predicted value of the controlled variable, set the error threshold , determine whether it exceeds the set error threshold, so as to obtain the error analysis result; among which, if the calculated error value Exceeding the set error threshold , then dynamically adjust each manipulated variable in the industrial production system: In the formula, Indicates that according to the error value The new value of the manipulated variable after dynamic adjustment; represents the manipulated variable value of the time series forecasting model based on the information at the previous time point; Represents the gain coefficient of the controller, which determines the speed and intensity of the system's response to the error; Every dynamic adjustment will automatically trigger an early warning prompt and provide feedback to technical personnel for review.

2. The intelligent control method of an industrial production system based on time series prediction according to claim 1 is characterized in that: In the step 1, the historical time series data of the industrial production system is collected, and the process of preprocessing the historical time series data includes: Acquire historical time series data through an industrial production system, wherein the historical time series data includes controlled variables, manipulated variables, and disturbance variables; Perform data cleaning on the collected historical time series data, where data cleaning includes outlier detection and missing value filling; Rearrange all preprocessed data in time series , as the final result of preprocessing.

3. The intelligent control method of an industrial production system based on time series prediction according to claim 2 is characterized in that: The process of outlier detection for the collected historical time series data includes: The Z-Score method is used to detect outliers: a single observation of any variable is marked as , then there is an outlier detection formula: In the formula, represents the outlier detection value; represents the mean of any variable, represents the standard deviation of any variable; if , then it is believed that is an outlier, among which, Represents the preset outlier detection threshold.

4. The intelligent control method of an industrial production system based on time series prediction according to claim 2 is characterized in that: The process of filling missing values ​​for the collected historical time series data includes: Use linear interpolation to fill missing values: For any variable In time If the value of Indicates the missing values ​​to be filled, then use and Interpolate the values ​​at the moment: In the formula, Indicates time Number.

5. The intelligent control method of an industrial production system based on time series prediction according to claim 1 is characterized in that: In the step 2, a time series prediction model is constructed, and the preprocessed historical time series data is used as a training data set. The process of training the time series prediction model includes: Choosing a time series forecasting model ,definition Network Layers , the number of hidden units in each layer ;in, is a structural parameter that defines the depth of the model; Extracting manipulated variables from preprocessed historical time series data , the controlled variable , and the disturbance variable ; The preprocessed historical time series data Divide into training set and validation set ; Assume that the preprocessed historical time series data Include time points, of which the first 80% are used as training sets and the remaining 20% ​​are used as validation sets; make For time point The input vector contains the information of the manipulated variables and the disturbance variables, that is, ;at the same time, For time point The controlled variable target value; The training process of the time series prediction model is as follows: B1. Setting model parameters ; Among them, the model parameters include The weight matrix and bias vector of ; B2. At each time point On, through Layer calculation prediction value : In the formula, Respectively represent the hidden state and memory unit state at the previous time point; B3. Calculate the loss function using the mean square error formula : In the formula, Indicates the total number of time points; B4. Update model parameters using Adam optimizer : In the formula, denote the unbias-corrected first- and second-order moment estimates, respectively; represents the hyperparameters used to control the decay rates of the first-order moment and the second-order moment respectively; Represents the loss function About Model Parameters The gradient of Respectively First-order and second-order moment estimates after correction for moment bias; Iterate to The model parameter values ​​at time instant; Represents the model parameter value for the next iteration; Represents the learning rate, which is used to control the step size of model parameter update; represents a small constant to avoid division by zero, ensuring that the denominator is not zero; B5. Evaluate model performance on the validation set and adjust model parameters as needed , hyperparameters , learning rate , or model structure parameters and ; Repeat steps B2-B5 until the performance of the model on the validation set reaches a predetermined standard or the maximum number of iterations is reached.

Images