Method and system for predicting working procedure power consumption data of production system
By constructing a linear regression model and using the LSTM model, considering the power transmission line loss and output measurement error, the problem of large prediction error in process power consumption in cement production is solved, and higher prediction accuracy and energy optimization effect are achieved.
Patent Information
- Application Number
- CN202510133184.7
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2025-02-06
- Publication Date
- 2025-06-10
AI Technical Summary
The prior art has large errors in the prediction of process power consumption during cement production, and the problems of power transmission line loss and output metering error cannot be effectively considered.
By collecting historical production data, a linear regression model is constructed to consider the power transmission line loss and output metering error, and combined with the LSTM model to predict the process power consumption to improve the prediction accuracy.
Effectively process long-term dependencies in time series data, improve the accuracy of process power consumption prediction, can guide production scheduling and optimize energy use, and reduce energy consumption costs.
Smart Images
Figure CN120124787A_ABST
Abstract
Description
Technical Field
[0001] The present invention belongs to the field of production control. Specifically, the present invention relates to a method and system for predicting the power consumption data of production system processes. Background Art
[0002] As a typical resource-intensive and energy-consuming industry, the energy conservation and environmental protection work in the cement industry has become the focus of attention.
[0003] The electricity consumption in the cement production process is mainly reflected in aspects such as the clinker burning system, cement grinding system, material conveying system, and auxiliary production system. In order to reduce electricity consumption and improve energy utilization efficiency, cement enterprises have taken a series of energy-saving measures, such as optimizing the production process, improving equipment performance, and strengthening energy management. At present, in order to timely grasp the energy consumption of each process in cement production, in the prior art, the energy consumption of relevant processes is often statistically calculated through measurement devices such as smart meters, belt scales / rotor scales for electricity and output.
[0004] However, due to the fact that in the real-time monitoring process of electricity measurement devices such as smart meters, the problems of line loss and amortization are not considered, resulting in a difference between the power consumption data statistically calculated in production and the power consumption statistically calculated in finance; due to the monitoring errors of output measurement devices such as belt scales / rotor scales themselves, there is a difference between the output data collected through the production system and the actual output data. Therefore, there is a problem of large errors in the prediction of the existing process power consumption. Summary of the Invention
[0005] The present invention aims to overcome the deficiencies of the prior art and proposes a method and system for predicting the power consumption data of production system processes to achieve the following objectives: Based on production data such as electricity consumption and output of the production system, considering the line loss of electricity transmission and the measurement error of output, the process power consumption is predicted through a neural network, thereby improving the prediction accuracy.
[0006] To achieve the above objectives, the technical solution adopted by the present invention is: A method for predicting the power consumption data of production system processes, the method comprising the following steps:
[0007] Step S1, collect the electricity consumption data and product output data in historical production, and after preprocessing these historical data, construct an electricity consumption data set and a product output data set;
[0008] Step S2, considering the line loss of electricity transmission and the measurement error of output, respectively construct corresponding linear regression models based on the corresponding data sets to determine the relationship between the process power consumption and the output;
[0009] Step S3: After temporalizing the power consumption dataset and the product output dataset respectively, calculate the corresponding process power consumption based on the linear regression model and form a process power consumption dataset.
[0010] Step S4: Construct an LSTM model; divide the temporalized power consumption dataset, the product output dataset, and the process power consumption dataset into a training set and a test set, use the training set to train the LSTM model, and verify and optimize the model performance through the test set; finally obtain a trained LSTM model.
[0011] Step S5: Apply the trained LSTM model to actual power consumption prediction, and predict the power consumption of each process according to the input power consumption data and product output data.
[0012] Preferably, in step S1, the power consumption dataset includes the real-time power consumption x e collected in the production system and the actual power consumption data y e ; the product output dataset includes the real-time output data x f and the actual output data y f .
[0013] Preferably, step S2 includes:
[0014] Step S21: Construct a first linear regression model representing the relationship between the real-time power consumption x e and the actual power consumption data y e , expressed as:
[0015] y e =k 1 (x e -x e b 1 );
[0016] where k 1 is the regression coefficient of the first linear regression model, and b 1 is the fixed error term of the first linear regression model, that is, the power transmission line loss coefficient;
[0017] Step S22: Construct a second linear regression model representing the relationship between the real-time output data x f and the actual output data y f , expressed as:
[0018] y f =k 2 (x f -x f b 2 );
[0019] where k 2is the regression coefficient of the second linear regression model, b 2 is the fixed error term of the second linear regression model, i.e., the error of the production measurement equipment;
[0020] Step S23: Perform linear regression analysis on the first and second linear regression models respectively through the power consumption data set and the product output data set to determine the first and second linear regression models.
[0021] Preferably, after the first and second linear regression models are determined, the process power consumption y m is calculated by the formula: y m = y f / y e .
[0022] Preferably, in step S4, the time-series power consumption data set and the product output data set are used as the input samples for training and testing, and the calculated process power consumption data is used as the output sample to train and test the long short-term memory network model LSTM.
[0023] Preferably, in step S4, the construction of the LSTM model includes setting the time step and the number of features; setting the loss function as MSE and the activation function as RELU.
[0024] Meanwhile, based on the above method, the present application proposes a prediction system for the process power consumption data of a production system. The prediction system includes a power collection system, a production control system, and an energy management system. Among them, the energy management system is respectively connected to the power collection system and the production control system. The power collection system is used to collect the power consumption data of the production system; the production control system is used to collect the product output data of the production system; the energy management system is used to predict the process power consumption data according to the above prediction method.
[0025] Preferably, the power collection system includes a smart meter, a data collection gateway, and a time-series database memory. The power consumption data collected by the smart meter is forwarded through the data collection gateway and stored in the time-series database memory.
[0026] Preferably, the production control system includes a PLC, which is connected to the energy management system and is used to collect the product output data of the production equipment in the production system and adjust the operation of the production system according to the commands of the energy management system.
[0027] Preferably, the energy management system includes a data acquisition and control unit, an energy efficiency monitoring layer, and an energy efficiency data analysis layer that are connected in sequence. Among them, the data acquisition and control layer directly acts on the power consumption data acquisition system and the production control system to collect the power consumption data and product output data. The energy efficiency monitoring layer is used to display the power consumption data, product output data, and predicted process power consumption data in real time. The energy efficiency data analysis layer is used to predict the future process power consumption data based on the current power consumption data and product output data, and issue commands to adjust the production system according to the prediction results.
[0028] The technical effects of the present invention are as follows: (1) In the prediction of the process power consumption of cement enterprises, the process power consumption data may be affected by factors such as power consumption and output changes in the past long period. The present invention uses the long short-term memory network model LSTM to predict the process power consumption, which can effectively process the long-term dependence relationship in time series data. (2) In the prediction, the influence of power transmission line loss and output measurement error is fully considered, which improves the prediction accuracy of the process power consumption. (3) In the present invention, the trained LSTM model can be applied to actual power consumption prediction, and the prediction results can be used to guide production scheduling, optimize energy use, and reduce energy consumption costs. BRIEF DESCRIPTION OF THE DRAWINGS
[0029] Figure 1 It is a flowchart of a method for predicting the process power consumption data of a production system according to an embodiment of the present invention;
[0030] Figure 2 It is a structural diagram of a system for predicting the process power consumption data of a production system according to an embodiment of the present invention. DETAILED DESCRIPTION OF THE EMBODIMENTS
[0031] Next, with reference to the drawings, through the description of the embodiments, the specific embodiments of the present invention will be further described in detail. The purpose is to help those skilled in the art have a more complete, accurate, and in-depth understanding of the inventive concept and technical solution of the present invention, and to facilitate its implementation. It should be noted that the terms "first", "second", etc. used in this application are only for the convenience of describing the technical solution to distinguish different components, and do not limit this application. To make the technical solution of the present invention clearer, the present invention will be explained and illustrated through the following embodiments.
[0032] This embodiment provides a method for predicting the process power consumption data of a production system. Using the LSTM algorithm, for the massive historical production data of cement enterprises, in this embodiment, mainly power consumption data and product output data (in actual application, other data such as equipment status can also be added. Correspondingly, just increase the input of the LSTM algorithm), and considering the influence of power transmission line loss and output measurement error, a process power consumption prediction model is generated. By continuously modifying the model, the real-time generated process power consumption data will be infinitely close to the true consumption data finally calculated manually, improving the prediction accuracy.
[0033] As Figure 1 shown, the method includes the following steps:
[0034] Step S1: Collect the power consumption data and product output data in historical production, and preprocess these historical data to construct a power consumption data set and a product output data set;
[0035] Step S2: Considering the power transmission line loss and output measurement error, respectively construct corresponding linear regression models based on the corresponding data sets to determine the relationship between the process power consumption and the output;
[0036] Step S3: After respectively time-sequencing the power consumption data set and the product output data set, calculate the corresponding process power consumption based on the linear regression model and form a process power consumption data set;
[0037] Step S4: Construct an LSTM model; divide the time-sequenced power consumption data set, product output data set, and process power consumption data set into a training set and a test set, use the training set to train the LSTM model, and verify and optimize the model performance through the test set; finally obtain the trained LSTM model;
[0038] Step S5: Apply the trained LSTM model to actual power consumption prediction, and predict the power consumption of each process according to the input power consumption data and product output data.
[0039] Specifically, in step S1 of this embodiment, after collecting the power consumption data and product output data in historical production, these historical data need to be preprocessed. The preprocessing includes data cleaning processes such as removing outliers and filling in missing values to ensure the accuracy and integrity of the data. In this embodiment, the power consumption data set includes the real-time power consumption x e and the actual power consumption data y e , the real-time power consumption x e is directly collected by power measurement devices such as smart meters, and the actual power consumption data y eIt is often calculated based on the data collected from the enterprise's reporting system; the product output data set includes real-time output data x f and actual output data y f . The real-time output data x f is collected in real time by the production control system; the actual output data y f is usually obtained by manual statistics.
[0040] This embodiment takes into account the influence of power transmission line loss and output measurement error. Correspondingly, step S2 of this embodiment includes:
[0041] Step S21, construct a first linear regression model representing the relationship between real-time power consumption x e and actual power consumption data y e , expressed as:
[0042] y e =k 1 (x e -x e b 1 );
[0043] where k 1 is the regression coefficient of the first linear regression model, and b 1 is the fixed error term of the first linear regression model, that is, the power transmission line loss coefficient;
[0044] Step S22, construct a second linear regression model representing the relationship between real-time output data x f and actual output data y f , expressed as:
[0045] y f =k 2 (x f -x f b 2 );
[0046] where k 2 is the regression coefficient of the second linear regression model, and b 2 is the fixed error term of the second linear regression model, that is, the output measurement equipment error;
[0047] Step S23, respectively perform linear regression analysis on the first and second linear regression models through the power consumption data set and the product output data set to determine the first and second linear regression models, that is, analyze and obtain k 1 , b 1 , k 2 , b 2 .
[0048] In step S3 of this embodiment, after the first and second linear regression models are determined, the production record equipment error and the power transmission line loss coefficient are basically fixed, so that the relationship between the actual power consumption and the production remains unchanged; therefore, the process power consumption y m is calculated by the formula: y m = y f / y e .
[0049] Then, after the power consumption data set and the product output data set are respectively serialized (arranged and processed in chronological order), the corresponding process power consumption is calculated according to the above formula and a process power consumption data set is formed.
[0050] In step S4 of this embodiment, a long short-term memory neural network model LSTM is used to calculate the process power consumption. The LSTM model belongs to the prior art and is good at dealing with and predicting long-term dependencies in time series data. First, define the LSTM model structure, including an input layer, a hidden layer (LSTM layer), and an output layer. The input layer receives the processed feature data, the hidden layer learns the time series features of the data through LSTM units, and the output layer predicts the future process power consumption. Then, according to the above structure, the LSTM model is constructed, including setting the time step (the time step refers to the length of the time series considered by the model in one processing), the number of features (the number of features refers to the dimension of the feature vector corresponding to each time step input into the LSTM model. For example, in this embodiment, the inputs are power consumption and production, so the number of features is 2); setting the loss function as MSE (mean squared error) and the activation function as RELU (rectified linear unit).
[0051] After the LSTM model is constructed, the serialized power consumption data set and product output data set are used as input samples for training and testing, and the calculated process power consumption data is used as the output sample to train and test the long short-term memory network model LSTM. The training set and the test set are usually divided in the ratio of 80% training set and 20% test set. The training set is used to train the LSTM model, and the performance of the model is verified and optimized through the test set. During the training process, techniques such as early stopping can be used to prevent overfitting.
[0052] During the training process, the LSTM model can be optimized according to the prediction results of the test set, including adjusting the parameters of the LSTM layer, increasing or decreasing the number of layers, adjusting the learning rate, etc., so as to continuously improve the prediction accuracy of the model.
[0053] Finally, in step S5 of this embodiment, the trained LSTM model can be applied to actual power consumption prediction, and the prediction results can be used to guide production scheduling, optimize energy use, and reduce energy consumption costs.
[0054] Meanwhile, based on the above method, this application proposes a prediction system for the power consumption data of production system processes, as Figure 2 shown. The prediction system includes an electricity consumption data acquisition system, a production control system, and an energy management system. Among them, the energy management system is respectively connected to the electricity consumption data acquisition system and the production control system. The electricity consumption data acquisition system is used to collect the electricity consumption data of the production system; the production control system is used to collect the product output data of the production system; the energy management system is used to predict the process power consumption data according to the above prediction method.
[0055] The electricity consumption data acquisition system of this embodiment includes an intelligent electric meter, a data acquisition gateway, and a time series database memory. The electricity consumption data collected by the intelligent electric meter is forwarded through the data acquisition gateway and stored in the time series database memory.
[0056] The production control system of this embodiment includes a PLC (Programmable Logic Controller). The PLC is connected to the energy management system. The PLC is used to collect the product output data of the production equipment in the production system by controlling product output metering devices such as belt scales / rotor scales. At the same time, the PLC can also adjust the operation of the production system according to the commands of the energy management system. In an advanced control system, the PLC can also be combined with a human-machine interface, so that users can also adjust the production equipment as needed.
[0057] The energy management system of this embodiment includes a data acquisition and control unit, an energy efficiency monitoring layer, and an energy efficiency data analysis layer that are connected in sequence.
[0058] The data acquisition and control layer directly acts on the electricity consumption data acquisition system and the production control system, and is used to collect the electricity consumption data and product output data and send them to the energy efficiency monitoring layer.
[0059] The energy efficiency monitoring layer is used to display in real time the electricity consumption data, product output data, and predicted process power consumption data to achieve comprehensive monitoring and management of the energy use of the entire enterprise. It is also used to monitor the electricity consumption data and product output data, and further transfer them to the energy efficiency data analysis layer after preliminary processing for in-depth data analysis and optimal scheduling.
[0060] The energy efficiency data analysis layer is used to predict the future process power consumption data according to the current electricity consumption data and product output data according to the method of the present invention, and issue commands to adjust the production system according to the prediction results. The commands are fed back to the production control system through the data acquisition and control layer.
[0061] The present invention has been described by way of example in conjunction with the accompanying drawings. Obviously, the specific implementation of the present invention is not limited by the above methods. As long as various non-substantive improvements are made by adopting the method concept and technical solution of the present invention; or without improvement, the above concept and technical solution of the present invention are directly applied to other occasions, they are all within the protection scope of the present invention.
Claims
1. A method for predicting power consumption data of a production system process, characterized by: The method comprises the following steps: Step S1, collecting power consumption data and product output data in historical production, and constructing power consumption data set and product output data set after preprocessing these historical data; Step S2, considering the power transmission line loss and the output measurement error, construct corresponding linear regression models based on the corresponding data sets to determine the relationship between the process power consumption and the output; Step S3, after the power consumption data set and the product output data set are time-series respectively, the corresponding process power consumption is calculated based on the linear regression model to form a process power consumption data set; Step S4, constructing an LSTM model; dividing the time-series power consumption data set, product output data set, and process power consumption data set into a training set and a test set, using the training set to train the LSTM model, and using the test set to verify and optimize the model performance; finally obtaining a trained LSTM model; Step S5: Apply the trained LSTM model to the actual power consumption prediction, and predict the power consumption of each process according to the input power consumption data and product output data.
2. The method for predicting power consumption data of a production system process according to claim 1, characterized in that: In step S1, the power consumption data set includes real-time power consumption x collected in the production system. e and actual power consumption data e ; The product output data set includes real-time output data x f and actual production data y f .
3. The method for predicting power consumption data of a production system process according to claim 2, characterized in that: The step S2 comprises: Step S21: construct a representation of real-time power consumption x e and actual power consumption data e The first linear regression model of the relationship between is expressed as: y e =k1(x e -x e b1); Among them, k1 is the regression coefficient of the first linear regression model, b1 is the fixed error term of the first linear regression model, that is, the power transmission line loss coefficient; Step S22: construct a real-time production data x f and actual production data y f The second linear regression model of the relationship between is expressed as: y f =k2(x f -x f b2); Among them, k2 is the regression coefficient of the second linear regression model, b2 is the fixed error term of the second linear regression model, that is, the error of the output measurement equipment; Step S23: performing linear regression analysis on the first and second linear regression models respectively through the power consumption data set and the product output data set to determine the first and second linear regression models.
4. The method for predicting power consumption data of a production system process according to claim 3, characterized in that: After the first and second linear regression models are determined, the process power consumption y m The calculation formula is: m =y f / y e .
5. The method for predicting power consumption data of a production system process according to claim 4, characterized in that: In step S4, the time-series power consumption data set and product output data set are used as input samples for training and testing, and the calculated process power consumption data is used as output samples to train and test the long short-term memory network model LSTM.
6. A method for predicting power consumption data of a production system process according to claim 4 or 5, characterized in that: In step S4, the construction of the LSTM model includes setting the time step and the number of features; setting the loss function to MSE and the activation function to RELU.
7. A system for predicting power consumption data of a production system process according to the prediction method according to any one of claims 1 to 6, characterized in that: The prediction system includes a power collection system, a production control system, and an energy management system, wherein the energy management system is connected to the power collection system and the production control system, respectively. The power collection system is used to collect power consumption data of the production system; the production control system is used to collect product output data of the production system; and the energy management system is used to predict process power consumption data according to the prediction method according to any one of claims 1-6.
8. The system for predicting power consumption data of a production system process according to claim 7, characterized in that: The power collection system includes a smart meter, a data collection gateway, and a time series database memory. The power consumption data collected by the smart meter is forwarded through the data collection gateway and stored in the time series database memory.
9. The system for predicting power consumption data of a production system process according to claim 7, characterized in that: The production control system includes a PLC, which is connected to the energy management system and is used to collect product output data of production equipment in the production system and adjust the operation of the production system according to the command of the energy management system.
10. The system for predicting power consumption data of a production system process according to claim 7, characterized in that: The energy management system includes a data acquisition and control unit, an energy efficiency monitoring layer, and an energy efficiency data analysis layer which are connected in sequence, wherein the data acquisition and control layer directly acts on the power acquisition system and the production control system to collect the power consumption data and the product output data; the energy efficiency monitoring layer is used to display the power consumption data, the product output data, and the predicted process power consumption data in real time; the energy efficiency data analysis layer is used to predict the future process power consumption data based on the current power consumption data and the product output data, and to issue commands to adjust the production system based on the predicted results.