WO2021007845A1 - 一种氧化铝生产指标的云-边协同预报系统及方法 - Google Patents
一种氧化铝生产指标的云-边协同预报系统及方法 Download PDFInfo
- Publication number
- WO2021007845A1 WO2021007845A1 PCT/CN2019/096512 CN2019096512W WO2021007845A1 WO 2021007845 A1 WO2021007845 A1 WO 2021007845A1 CN 2019096512 W CN2019096512 W CN 2019096512W WO 2021007845 A1 WO2021007845 A1 WO 2021007845A1
- Authority
- WO
- WIPO (PCT)
- Prior art keywords
- model
- forecast
- data
- alumina production
- cloud
- Prior art date
Links
- 238000004519 manufacturing process Methods 0.000 title claims abstract description 178
- 238000000034 method Methods 0.000 title claims abstract description 53
- TWNQGVIAIRXVLR-UHFFFAOYSA-N oxo(oxoalumanyloxy)alumane Chemical compound O=[Al]O[Al]=O TWNQGVIAIRXVLR-UHFFFAOYSA-N 0.000 title abstract 9
- 238000012549 training Methods 0.000 claims abstract description 138
- 238000012937 correction Methods 0.000 claims abstract description 57
- 238000011156 evaluation Methods 0.000 claims abstract description 21
- 238000013277 forecasting method Methods 0.000 claims abstract description 6
- PNEYBMLMFCGWSK-UHFFFAOYSA-N aluminium oxide Inorganic materials [O-2].[O-2].[O-2].[Al+3].[Al+3] PNEYBMLMFCGWSK-UHFFFAOYSA-N 0.000 claims description 148
- 238000007726 management method Methods 0.000 claims description 25
- 238000007405 data analysis Methods 0.000 claims description 21
- 238000013480 data collection Methods 0.000 claims description 12
- 238000007781 pre-processing Methods 0.000 claims description 12
- 238000004458 analytical method Methods 0.000 claims description 10
- 230000008676 import Effects 0.000 claims description 9
- 238000012545 processing Methods 0.000 claims description 5
- 238000012360 testing method Methods 0.000 claims description 5
- 238000009795 derivation Methods 0.000 claims description 3
- 238000010835 comparative analysis Methods 0.000 claims description 2
- 238000004090 dissolution Methods 0.000 description 6
- 230000009286 beneficial effect Effects 0.000 description 3
- 238000010586 diagram Methods 0.000 description 1
- 230000000694 effects Effects 0.000 description 1
- 238000005265 energy consumption Methods 0.000 description 1
- 238000012986 modification Methods 0.000 description 1
- 230000004048 modification Effects 0.000 description 1
- 238000012706 support-vector machine Methods 0.000 description 1
- 238000012546 transfer Methods 0.000 description 1
Images
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N20/00—Machine learning
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06Q—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES; SYSTEMS OR METHODS SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES, NOT OTHERWISE PROVIDED FOR
- G06Q10/00—Administration; Management
- G06Q10/04—Forecasting or optimisation specially adapted for administrative or management purposes, e.g. linear programming or "cutting stock problem"
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/21—Design or setup of recognition systems or techniques; Extraction of features in feature space; Blind source separation
- G06F18/214—Generating training patterns; Bootstrap methods, e.g. bagging or boosting
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06Q—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES; SYSTEMS OR METHODS SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES, NOT OTHERWISE PROVIDED FOR
- G06Q10/00—Administration; Management
- G06Q10/06—Resources, workflows, human or project management; Enterprise or organisation planning; Enterprise or organisation modelling
- G06Q10/063—Operations research, analysis or management
- G06Q10/0639—Performance analysis of employees; Performance analysis of enterprise or organisation operations
- G06Q10/06393—Score-carding, benchmarking or key performance indicator [KPI] analysis
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06Q—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES; SYSTEMS OR METHODS SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES, NOT OTHERWISE PROVIDED FOR
- G06Q10/00—Administration; Management
- G06Q10/06—Resources, workflows, human or project management; Enterprise or organisation planning; Enterprise or organisation modelling
- G06Q10/067—Enterprise or organisation modelling
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06Q—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES; SYSTEMS OR METHODS SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES, NOT OTHERWISE PROVIDED FOR
- G06Q50/00—Information and communication technology [ICT] specially adapted for implementation of business processes of specific business sectors, e.g. utilities or tourism
- G06Q50/04—Manufacturing
-
- C—CHEMISTRY; METALLURGY
- C01—INORGANIC CHEMISTRY
- C01F—COMPOUNDS OF THE METALS BERYLLIUM, MAGNESIUM, ALUMINIUM, CALCIUM, STRONTIUM, BARIUM, RADIUM, THORIUM, OR OF THE RARE-EARTH METALS
- C01F7/00—Compounds of aluminium
- C01F7/02—Aluminium oxide; Aluminium hydroxide; Aluminates
-
- Y—GENERAL TAGGING OF NEW TECHNOLOGICAL DEVELOPMENTS; GENERAL TAGGING OF CROSS-SECTIONAL TECHNOLOGIES SPANNING OVER SEVERAL SECTIONS OF THE IPC; TECHNICAL SUBJECTS COVERED BY FORMER USPC CROSS-REFERENCE ART COLLECTIONS [XRACs] AND DIGESTS
- Y02—TECHNOLOGIES OR APPLICATIONS FOR MITIGATION OR ADAPTATION AGAINST CLIMATE CHANGE
- Y02P—CLIMATE CHANGE MITIGATION TECHNOLOGIES IN THE PRODUCTION OR PROCESSING OF GOODS
- Y02P90/00—Enabling technologies with a potential contribution to greenhouse gas [GHG] emissions mitigation
- Y02P90/30—Computing systems specially adapted for manufacturing
Definitions
- the invention relates to the technical field of industrial cloud computing and edge-end collaborative forecasting, in particular to a cloud-edge collaborative forecasting system and method for alumina production indicators.
- the company does not allow staff to randomly acquire, modify, verify and test data such as variables and indicators at the industrial site.
- the requirements for quality and reduction of energy consumption are relatively difficult, and the existing alumina production index forecasting process is mainly through the separate processing of industrial process data collected by sensors and laboratory analysis data, and artificially re-process the processed results. Processing to ensure the consistency of data dimensions, etc., and use the data for model training, but this process is difficult to meet the real-time requirements of the forecasting process, so it is impossible to provide real-time feedback to the staff, and the staff cannot make real-time adjustments to product variables. Therefore, it is very necessary to establish an efficient and convenient forecasting system to better forecast the production indicators of the alumina production process.
- the technical problem to be solved by the present invention is to provide a cloud-side collaborative forecasting system and method for alumina production indicators in view of the above-mentioned shortcomings of the prior art, so as to realize the forecasting of the production indicators in the alumina production process.
- the present invention provides a cloud-side collaborative forecasting system for alumina production indicators, including a data acquisition device, a hardware platform, and software set on the hardware platform System;
- the data acquisition device is used to collect the actual data of the entire process of alumina production, and transmit the collected data to the cloud server and the alumina production indicator prediction computer at the edge at fixed time intervals;
- the hardware platform includes The cloud model training server and the alumina production index prediction computer at the edge;
- the software system includes software set on the cloud model training server and the alumina production index forecast computer set at the edge;
- the software set on the cloud model training server The software realizes the selection and management of the alumina production index forecast algorithm, and trains and evaluates the alumina production index forecast model through the actual data of the entire alumina production process collected by the data acquisition device, and uses the collected data In the production index forecast model, perform big data analysis on the model forecast results and actual production index data, and then correct the forecast model parameters;
- the software set on the cloud model training server includes a forecast model training program management unit, an algorithm and model library management unit, a big data analysis unit, a model training unit, a forecast model parameter correction unit, a forecast model analysis unit, and model correction Parameter download unit;
- the forecast model training program management unit is used to manage the training sample data used when training the forecast model and store the model parameters, so as to compare when analyzing the forecast model performance;
- the algorithm and model library management unit is used to manage the obtained alumina production index forecast model and the forecast algorithm used in the model training process and the common data processing algorithm built in the system, and provide the new, update and delete function of the forecast algorithm;
- the model training unit is used to perform model training based on the data generated by the actual operation of the alumina collected by the data collection device within a certain period of time and the selected prediction algorithm, and download the model training results to the model training evaluation unit; the model training The process includes four steps: data selection, data division, data preprocessing and algorithm selection;
- the big data analysis unit uses the trained model to calculate the correlation coefficient between the alumina production index forecast result and the actual value according to the collected alumina production operation data, and judges whether the model parameter correction is required according to the correlation coefficient ;
- the forecast model parameter correction unit continues to train the forecast model according to the latest production working condition data and according to the model training parameter configuration, so as to obtain corrected forecast model parameters adapted to the new working condition;
- the forecast model analysis unit uses the corrected forecast model parameters and the current forecast model parameters to predict and analyze the alumina production index data stored in the forecast model training program management module; if the corrected model parameters are used to predict the alumina production index If the result is better than the accuracy of the forecast model with current parameters, select the corrected forecast model parameters, otherwise keep the current forecast model parameters unchanged;
- the model correction parameter download unit provides the derivation function of the corrected forecast model parameters for downloading to the forecast model management unit of the alumina production index forecast computer at the edge; meanwhile, it also provides software in the alumina production index forecast computer at the edge.
- An online download interface that can automatically update the corrected model parameters remotely.
- the cloud model training server further includes a model training evaluation unit for evaluating the performance of the trained forecast model; the performance evaluation uses mean square error, average absolute value error, average absolute percentage error, and forecast value These indicators are compared with the actual value correlation coefficient.
- the model training unit also has a built-in parameter configuration module for configuring the parameters of the prediction algorithm for the specific training algorithm used in the model training process.
- the software of the alumina production index forecast computer set at the edge includes a forecast model management unit and an intelligent forecast unit;
- the forecast model management unit realizes the management of the local edge model training program, including a model selection module and a model parameter correction module;
- the model selection module selects the alumina production index forecast model corresponding to the production process according to the judgment result of the alumina production process For forecasting, the edge end and the cloud have the same multiple alumina production index forecast model training schemes;
- the model parameter correction module imports the forecast model parameters corrected by the cloud forecast model training server, or directly performs remote update of the correction parameters for the The accuracy of the model prediction results can still be guaranteed under the new working conditions;
- the intelligent forecasting unit uses a trained forecast model at the edge end to use the actual data of the entire alumina production process passed to the edge end to predict alumina production indicators for different production processes.
- the present invention also provides a cloud-side collaborative forecasting method for alumina production indicators, including three parts: the collection of alumina production indicator data, the operation of the cloud training server and the operation of the edge-end alumina production indicator prediction computer;
- the alumina production index data collection collects actual operation data generated by alumina production through a data collection device, and transmits the collected data to the model training unit of the cloud training server and the alumina production index prediction computer intelligent prediction unit at the edge. ;
- the operation of the cloud training server includes a model training process and a model parameter correction process
- the model training process is:
- Step 1 The model training unit of the cloud training server conducts model training based on the data generated by the actual operation of the alumina collected by the data collection device within a certain period of time and the selected prediction algorithm, and downloads the model training results to the model training evaluation unit;
- the model training process includes the following processes:
- Data division select the data division method to divide the selected data into training set and test set;
- Data preprocessing choose different data preprocessing algorithms according to different goals and data forms
- Algorithm selection select the algorithm for model training from the algorithm and model library management unit;
- Step 2 The model training evaluation unit evaluates the performance of the forecast model trained by the model training unit; the performance evaluation indicators include the mean square error, the average absolute value error, the average absolute percentage error, and the correlation coefficient between the forecast value and the actual value;
- the model parameter correction process is:
- Step C1 The big data analysis unit uses the trained model to calculate the correlation coefficient between the alumina production index forecast result and the actual value according to the collected alumina production operation data, and judges whether the model parameters are required according to the correlation coefficient Correction, if the correlation coefficient between the alumina production index forecast result and the actual value is less than the correlation coefficient threshold set according to the specific production index requirements, then perform step 2 for model parameter correction, otherwise no model parameter correction;
- Step C2 According to the big data analysis result obtained by the big data analysis unit, the forecast model parameter correction unit continues to train the forecast model according to the forecast model parameter configuration set by the model training unit, thereby obtaining corrected forecast model parameters adapted to the big data analysis result;
- Step C3 The forecast model analysis unit uses the corrected forecast model parameters and the current forecast model parameters to predict and analyze the alumina production index data stored in the forecast model training program management module; if the corrected model parameters are used to predict the alumina production index If the result of is better than the forecast model accuracy of the current parameters, select the corrected forecast model parameters, otherwise keep the current forecast model parameters unchanged;
- the operation process of the alumina product index prediction computer at the edge end is:
- Step S1 The model selection module selects the forecast model of alumina production indicators to be predicted according to the alumina production process
- Step S2 The model parameter correction module imports the forecast model parameters corrected on the cloud forecast model training server through the model correction parameter download unit, or directly performs remote update of the correction parameters in the cloud;
- Step S3 The intelligent forecasting unit uses the trained forecasting model at the edge to forecast alumina production indicators in different production processes.
- the cloud-side collaborative forecasting system and method for alumina production indicators provided by the present invention, based on the existing alumina production process requirements, perform alumina production indicators on the cloud server
- the selection of forecasting algorithm, the parameter configuration of training model and the process of model training, evaluation and correction provide powerful computing resources for alumina production index forecast model training; use cloud-trained models at the edge to perform real-time forecasting of alumina production indexes.
- the marginal resources are saved, and it is convenient to view the forecast results of alumina production indicators in time.
- Fig. 1 is a structural block diagram of a cloud-side collaborative forecasting system for alumina production indicators provided by an embodiment of the present invention
- Fig. 2 is a flowchart of a cloud-side collaborative forecasting method for alumina production indicators provided by an embodiment of the present invention.
- a cloud-side collaborative forecasting system for alumina production indicators includes a data acquisition device, a hardware platform, and a software system set on the hardware platform; the data acquisition device is used to collect The actual data of the whole process of the alumina production process, and the collected data are transmitted to the cloud server and the alumina production indicator prediction computer at the edge at fixed time intervals; the hardware platform includes the cloud model training server and the alumina production at the edge Index prediction computer; the software system includes software set on the cloud model training server and the alumina production indicator prediction computer software set on the edge; the software set on the cloud model training server realizes the evaluation of the alumina production indicator prediction algorithm Select management, and train and evaluate the alumina production index forecast model through the actual data of the entire alumina production process collected by the data acquisition device, and apply the collected data to the production index forecast model to predict the results of the model Perform big data analysis with actual actual production index data, and then correct the forecast model parameters; the software of the alumina production index forecast computer set at the
- the software set on the cloud model training server includes a forecast model training program management unit, an algorithm and model library management unit, a big data analysis unit, a model training unit, a forecast model parameter correction unit, a forecast model analysis unit, and a model correction parameter download unit ;
- the forecast model training program management unit is used to manage the training sample data used when training the forecast model and store the model parameters, so as to compare when analyzing the forecast model performance;
- the algorithm and model library management unit is used to manage the obtained alumina production index forecast model and the forecast algorithm used in the model training process and the common data processing algorithm built in the system, and provide the new, update and delete function of the forecast algorithm;
- the model training unit is used to perform model training based on the data generated by the actual operation of the alumina collected by the data collection device within a certain period of time and the selected prediction algorithm, and download the model training results to the model training evaluation unit;
- the model training The process includes four steps: data selection, data division, data preprocessing, and algorithm selection; at the same time, the model training unit also includes a parameter configuration module, which is used to configure the parameters of the forecast algorithm for the specific training algorithm used in the model training process;
- the big data analysis unit uses the trained model to calculate the correlation coefficient between the alumina production index forecast result and the actual value according to the collected alumina production operation data, and judges whether the model parameter correction is required according to the correlation coefficient ;
- the forecast model parameter correction unit continues to train the forecast model according to the latest production working condition data and according to the model training parameter configuration, so as to obtain corrected forecast model parameters adapted to the new working condition;
- the forecast model analysis unit uses the corrected forecast model parameters and the current forecast model parameters to predict and analyze the alumina production index data stored in the forecast model training program management module; if the corrected model parameters are used to predict the alumina production index If the result is better than the accuracy of the forecast model with current parameters, select the corrected forecast model parameters, otherwise keep the current forecast model parameters unchanged;
- the model correction parameter download unit provides the derivation function of the corrected forecast model parameters for downloading to the forecast model management unit of the alumina production index forecast computer at the edge; meanwhile, it also provides software in the alumina production index forecast computer at the edge.
- the online download interface for automatic remote update of the corrected model parameters.
- the cloud model training server also includes a model training evaluation unit for evaluating the performance of the trained forecast model; the performance evaluation adopts mean square error, average absolute value error, average absolute percentage error, and forecast value and actual value. Correlation coefficient of these indicators for comparative analysis.
- the software of the alumina production index forecast computer set at the edge includes a forecast model management unit and an intelligent forecast unit;
- the forecast model management unit realizes the management of the local edge model training program, including a model selection module and a model parameter correction module;
- the model selection module selects the alumina production index forecast model corresponding to the production process according to the judgment result of the alumina production process For forecasting, the edge end and the cloud have the same multiple alumina production index forecast model training schemes;
- the model parameter correction module imports the forecast model parameters corrected by the cloud forecast model training server, or directly performs remote update of the correction parameters for the The accuracy of the model prediction results can still be guaranteed under the new working conditions;
- the intelligent forecasting unit uses a trained forecast model at the edge end to use the actual data of the entire alumina production process passed to the edge end to predict alumina production indicators for different production processes.
- a cloud-side collaborative forecasting method for alumina production indicators includes three parts: the collection of alumina production indicator data, the operation of the cloud training server and the operation of the edge alumina production indicator prediction computer;
- the alumina production index data collection collects actual operation data generated by alumina production through a data collection device, and transmits the collected data to the forecast model training program management unit of the cloud training server and the alumina production index forecast computer at the edge Data preprocessing unit;
- the operation of the cloud training server includes a model training process and a model parameter correction process
- the model training process is:
- Step 1 The model training unit of the cloud training server performs model training according to the data generated by the actual operation of alumina collected by the data collection device within a certain period of time and the selected prediction algorithm, and downloads the model training results to the model training evaluation unit;
- the model training process includes the following processes:
- 500 alumina dissolution process sample data are selected for a certain period of time for model training, and the predicted production index is selected as the dissolution rate, and the selected data are shown in Table 1;
- Data division select the data division method to divide the selected data into training set and test set;
- 80% of the data is selected as the training set, 20% of the data is selected as the test set, and a random division method is selected;
- Data preprocessing choose different data preprocessing algorithms according to different goals and data forms
- standardized algorithms are used for data preprocessing, and the data preprocessing algorithms can perform operations such as new creation and update in the algorithm management module by themselves;
- Algorithm selection select the algorithm for model training from the algorithm and model library management unit;
- This embodiment selects the ⁇ support vector machine algorithm for model training.
- the parameters that the algorithm needs to configure are the penalty factor C, the error accuracy requirement ⁇ , and the kernel function.
- Step 2 The model training evaluation unit evaluates the performance of the forecast model trained by the model training unit; the performance evaluation indicators include the mean square error, the average absolute value error, the average absolute percentage error, and the correlation coefficient between the forecast value and the actual value;
- the average absolute error is selected to evaluate the performance of the model.
- the average absolute error value is 0.2, which is less than the set threshold 0.5, indicating that the model has a good training effect and can be used as a dissolution rate prediction model and stored in the prediction model training program management unit;
- the model parameter correction process is:
- Step C1 The big data analysis unit uses the trained model to calculate the correlation coefficient between the alumina production index forecast result and the actual value according to the collected new alumina production operation data, and judges whether it needs to be performed according to the correlation coefficient Model parameter correction. If the correlation coefficient between the forecast result of alumina production index and the actual value is less than the correlation coefficient threshold set according to the specific production index requirements, perform step 2 for model parameter correction, otherwise no model parameter correction;
- Step C2 According to the big data analysis result obtained by the big data analysis unit, the forecast model parameter correction unit continues to train the forecast model according to the forecast model parameter configuration set by the model training unit, thereby obtaining corrected forecast model parameters adapted to the big data analysis result;
- Step C3 The forecast model analysis unit uses the corrected forecast model parameters and the current forecast model parameters to predict and analyze the alumina production index data stored in the forecast model training program management module; if the corrected model parameters are used to predict the alumina production index If the result of is better than the forecast model accuracy of the current parameters, select the corrected forecast model parameters, otherwise keep the current forecast model parameters unchanged;
- the operation process of the alumina product index prediction computer at the edge end is:
- Step S1 The model selection module selects the alumina production index forecast model to be predicted according to the alumina production process
- Step S2 The model parameter correction module imports the forecast model parameters corrected on the cloud forecast model training server through the model correction parameter download unit, or directly performs remote update of the correction parameters in the cloud;
- Step S3 The intelligent forecasting unit uses the trained forecasting model at the edge to forecast alumina production indicators in different production processes.
Landscapes
- Engineering & Computer Science (AREA)
- Business, Economics & Management (AREA)
- Human Resources & Organizations (AREA)
- Theoretical Computer Science (AREA)
- Strategic Management (AREA)
- Economics (AREA)
- General Physics & Mathematics (AREA)
- Physics & Mathematics (AREA)
- Entrepreneurship & Innovation (AREA)
- Tourism & Hospitality (AREA)
- Marketing (AREA)
- Development Economics (AREA)
- General Business, Economics & Management (AREA)
- Operations Research (AREA)
- Quality & Reliability (AREA)
- Game Theory and Decision Science (AREA)
- Data Mining & Analysis (AREA)
- Life Sciences & Earth Sciences (AREA)
- Chemical & Material Sciences (AREA)
- Organic Chemistry (AREA)
- Computer Vision & Pattern Recognition (AREA)
- General Engineering & Computer Science (AREA)
- Evolutionary Computation (AREA)
- Artificial Intelligence (AREA)
- Educational Administration (AREA)
- Software Systems (AREA)
- Bioinformatics & Cheminformatics (AREA)
- Evolutionary Biology (AREA)
- Inorganic Chemistry (AREA)
- Geology (AREA)
- Bioinformatics & Computational Biology (AREA)
- Mathematical Physics (AREA)
- Computing Systems (AREA)
- Medical Informatics (AREA)
- Manufacturing & Machinery (AREA)
- Health & Medical Sciences (AREA)
- General Health & Medical Sciences (AREA)
- Primary Health Care (AREA)
- Management, Administration, Business Operations System, And Electronic Commerce (AREA)
- Feedback Control In General (AREA)
Abstract
一种氧化铝生产指标的云‑边协同预报系统,预报系统通过在云端模型训练服务器对氧化铝生产过程指标与变量进行预报算法选择、参数配置及模型训练,并对训练好的模型进行评估及参数校正,得到最优的训练模型;同时在边缘端的氧化铝生产指标预报计算机上对氧化铝生产过程数据进行预处理,并通过从云端导入训练好的模型参数,进而使用训练好的预报模型针对不同生产过程对氧化铝生产指标进行预报。及一种氧化铝生产指标的云‑边协同预报方法。上述系统及方法通过云端服务器为氧化铝生产指标预报模型训练提供强大的计算资源,通过边缘端的计算机提供实时便捷的氧化铝生产指标预报。
Description
本发明涉及工业云计算与边缘端协同预报技术领域,尤其涉及一种氧化铝生产指标的云-边协同预报系统及方法。
由于氧化铝生产全流程工艺复杂,涉及各个工序与大量生产数据,由于安全等原因,企业不允许工作人员在工业现场随意进行变量与指标等数据的获取、修改、验证与测试,因此对于提高产品质量以及降低能耗等要求相对困难,并且现有的氧化铝生产指标预报过程主要是通过将传感器等采集到的工业过程数据与实验室化验分析数据进行单独处理,人为将处理后的结果进行再次处理以保证数据维度等的一致,并以此数据进行模型训练,但这个过程难以满足预报过程的实时性要求,因此无法向工作人员进行实时反馈,工作人员无法对产品变量等进行实时性调整。所以非常需要建立一个高效、便捷的预报系统,以便更好的对氧化铝生产过程的生产指标进行预报。
发明概述
问题的解决方案
本发明要解决的技术问题是针对上述现有技术的不足,提供一种氧化铝生产指标的云-边协同预报系统及方法,实现对氧化铝生产过程生产指标进行预报。
为解决上述技术问题,本发明所采取的技术方案是:一方面,本发明提供一种氧化铝生产指标的云-边协同预报系统,包括数据采集装置、硬件平台和设置在硬件平台上的软件系统;所述数据采集装置用于采集氧化铝生产过程的全流程实际数据,并将采集到的数据以固定时间间隔传输给云端服务器和边缘端的氧化铝生产指标预报计算机上;所述硬件平台包括云端模型训练服务器与边缘端的氧化铝生产指标预报计算机;所述软件系统包括设置在云端模型训练服务器 的软件和设置在边缘端的氧化铝生产指标预报计算机的软件;所述设置在云端模型训练服务器的软件实现对氧化铝生产指标预报算法的选择管理,并将通过数据采集装置采集到的氧化铝生产过程的全流程实际数据对氧化铝生产指标预报模型进行训练、评估,并将采集到的数据运用到生产指标预报模型中,将模型预报结果与实际生产指标数据进行大数据分析,进而对预报模型参数进行校正;所述设置在边缘端的氧化铝生产指标预报计算机的软件能够实现从云端模型训练服务器导入云端预报模型训练服务器校正后的预报模型参数,或者直接进行校正参数的远程更新,进而将数据采集装置采集到的氧化铝生产过程的全流程实际数据传入边缘端,在边缘端使用训练好的预报模型针对不同生产过程对氧化铝生产指标进行预报。
优选地,所述设置在云端模型训练服务器的软件包括预报模型训练方案管理单元、算法与模型库管理单元、大数据分析单元、模型训练单元、预报模型参数校正单元、预报模型分析单元和模型校正参数下载单元;
所述预报模型训练方案管理单元用于管理训练预报模型时使用的训练样本数据和存储模型参数,以便在分析预报模型性能时进行对比;
所述算法与模型库管理单元用于管理得到的氧化铝生产指标预报模型及模型训练过程所采用的预报算法和系统内置的常用数据处理算法,并提供预报算法的新建、更新与删除功能;
所述模型训练单元用于根据一定时间内数据采集装置采集的氧化铝实际运行所产生的数据及选择的预报算法进行模型训练,并将模型训练结果下传给模型训练评估单元;所述模型训练过程包括数据选择、数据划分、数据预处理和算法选择四步;
所述大数据分析单元根据采集到的氧化铝生产运行数据运用训练好的模型进行氧化铝生产指标预报结果与实际值之间的相关性系数计算,并根据相关性系数判断是否需要进行模型参数校正;
所述预报模型参数校正单元根据最近的生产工况数据,按照模型训练参数配置继续训练预报模型,从而得到适应新工况的校正后的预报模型参数;
所述预报模型分析单元采用校正后的预报模型参数与当前预报模型参数对预报 模型训练方案管理模块中存储的氧化铝生产指标数据进行预报分析;如果采用校正后的模型参数预报氧化铝生产指标的结果比当前参数的预报模型准确性好,则选择校正后的预报模型参数,否则维持当前预报模型参数不变;
所述模型校正参数下载单元提供校正后的预报模型参数的导出功能,以便下载到边缘端的氧化铝生产指标预报计算机的预报模型管理单元中;同时还提供了边缘端的氧化铝生产指标预报计算机中软件可以自动远程更新校正后的模型参数的在线下载接口。
优选地,所述云端模型训练服务器上还包括模型训练评估单元用于对训练得到的预报模型进行性能评估;所述性能评估采用均方误差、平均绝对值误差、平均绝对百分误差及预报值与实际值相关性系数这些指标进行对比分析。
优选地,所述模型训练单元还内置参数配置模块,用于对模型训练过程采用的具体训练算法进行预报算法的参数配置。
优选地,所述设置在边缘端的氧化铝生产指标预报计算机的软件包括预报模型管理单元和智能预报单元;
所述预报模型管理单元实现对本地边缘端模型训练方案的管理,包括模型选择模块与模型参数校正模块;模型选择模块根据氧化铝生产过程的判断结果,选择对应生产过程的氧化铝生产指标预报模型进行预报,边缘端与云端均有相同的多种氧化铝生产指标预报模型训练方案;模型参数校正模块导入云端预报模型训练服务器校正后的预报模型参数,或者直接进行校正参数的远程更新,以便在新的工况下仍然能够保证模型预报结果的准确性;
所述智能预报单元在边缘端使用训练好的预报模型运用传入边缘端的氧化铝生产全流程实际数据针对不同生产过程对氧化铝生产指标进行预报。
另一方面,本发明还提供一种氧化铝生产指标的云-边协同预报方法,包括氧化铝生产指标数据的采集、云端训练服务器的运行和边缘端氧化铝生产指标预报计算机的运行三部分;
所述氧化铝生产指标数据采集通过数据采集装置采集氧化铝生产所产生的实际运行数据,并将采集到的数据传输给云端训练服务器的模型训练单元和边缘端的氧化铝生产指标预报计算机智能预报单元;
所述云端训练服务器的运行包括模型训练过程与模型参数校正过程;
所述模型训练过程为:
步骤1:云端训练服务器的模型训练单元根据一定时间内数据采集装置采集的氧化铝实际运行所产生的数据及选择的预报算法进行模型训练,并将模型训练结果下传给模型训练评估单元;所述模型训练过程包括以下过程:
数据选择:根据训练目的对数据采集装置采集到的氧化铝生产全流程数据将进行区间划分;
数据划分:选用数据划分方法将选择的数据划分为训练集与测试集;
数据预处理:根据不同目标及数据形式选用不同数据预处理算法;
算法选择:从算法与模型库管理单元选择进行模型训练的算法;
步骤2:模型训练评估单元对模型训练单元训练得到的预报模型进行性能评估;性能评估指标包括均方误差、平均绝对值误差、平均绝对百分误差及预报值与实际值的相关性系数;
所述模型参数校正过程为:
步骤C1:大数据分析单元根据采集到的氧化铝生产运行数据运用训练好的模型进行氧化铝生产指标预报结果与实际值之间的相关性系数计算,并根据相关性系数判断是否需要进行模型参数校正,如果氧化铝生产指标预报结果与实际值之间的相关性系数小于根据具体生产指标要求设定的相关性系数阈值,则执行步骤2进行模型参数校正,否则不进行模型参数校正;
步骤C2:预报模型参数校正单元根据大数据分析单元得到的大数据分析结果,按照模型训练单元设置的预报模型参数配置继续训练预报模型,从而得到适应大数据分析结果的校正后的预报模型参数;
步骤C3:预报模型分析单元采用校正后的预报模型参数与当前预报模型参数对预报模型训练方案管理模块中存储的氧化铝生产指标数据进行预报分析;如果采用校正后的模型参数预报氧化铝生产指标的结果比当前参数的预报模型准确性好,则选择校正后的预报模型参数,否则维持当前预报模型参数不变;
所述边缘端的氧化铝产品指标预报计算机的运行过程为:
步骤S1:模型选择模块根据氧化铝生产过程选择所要预测的氧化铝生产指标预 报模型;
步骤S2:模型参数校正模块通过模型校正参数下载单元导入在云端预报模型训练服务器校正后的预报模型参数,或者在云端直接进行校正参数的远程更新;
步骤S3:智能预报单元在边缘端使用训练好的预报模型针对不同生产过程的氧化铝生产指标进行预报。
发明的有益效果
采用上述技术方案所产生的有益效果在于:本发明提供的一种氧化铝生产指标的云-边协同预报系统及方法,基于现有的氧化铝全流程生产要求,在云端服务器进行氧化铝生产指标预报算法的选择,训练模型的参数配置及模型训练、评估校正过程,为氧化铝生产指标预报模型训练提供强大的计算资源;在边缘端使用云端训练好的模型进行氧化铝生产指标的实时预测,节省了边缘端资源,便于及时查看氧化铝生产指标的预报结果。
对附图的简要说明
图1为本发明实施例提供的一种氧化铝生产指标的云-边协同预报系统的结构框图;
图2为本发明实施例提供的一种氧化铝生产指标的云-边协同预报方法的流程图。
发明实施例
下面结合附图和实施例,对本发明的具体实施方式作进一步详细描述。以下实施例用于说明本发明,但不用来限制本发明的范围。
本实施例中,一种氧化铝生产指标的云-边协同预报系统,如图1所示,包括数据采集装置、硬件平台和设置在硬件平台上的软件系统;所述数据采集装置用于采集氧化铝生产过程的全流程实际数据,并将采集到的数据以固定时间间隔传输给云端服务器和边缘端的氧化铝生产指标预报计算机上;所述硬件平台包 括云端模型训练服务器与边缘端的氧化铝生产指标预报计算机;所述软件系统包括设置在云端模型训练服务器的软件和设置在边缘端的氧化铝生产指标预报计算机的软件;所述设置在云端模型训练服务器的软件实现对氧化铝生产指标预报算法的选择管理,并通过数据采集装置采集到的氧化铝生产过程的全流程实际数据对氧化铝生产指标预报模型进行训练、评估,并将采集到的数据运用到生产指标预报模型中,将模型预报结果与实际实际生产指标数据进行大数据分析,进而对预报模型参数进行校正;所述设置在边缘端的氧化铝生产指标预报计算机的软件能够实现从云端模型训练服务器导入云端预报模型训练服务器校正后的预报模型参数,或者直接进行校正参数的远程更新,进而将数据采集装置采集到的氧化铝生产过程的全流程实际数据传入边缘端,在边缘端使用训练好的预报模型针对不同生产过程对氧化铝生产指标进行预报。
所述设置在云端模型训练服务器的软件包括预报模型训练方案管理单元、算法与模型库管理单元、大数据分析单元、模型训练单元、预报模型参数校正单元、预报模型分析单元和模型校正参数下载单元;
所述预报模型训练方案管理单元用于管理训练预报模型时使用的训练样本数据和存储模型参数,以便在分析预报模型性能时进行对比;
所述算法与模型库管理单元用于管理得到的氧化铝生产指标预报模型及模型训练过程所采用的预报算法和系统内置的常用数据处理算法,并提供预报算法的新建、更新与删除功能;
所述模型训练单元用于根据一定时间内数据采集装置采集的氧化铝实际运行所产生的数据及选择的预报算法进行模型训练,并将模型训练结果下传给模型训练评估单元;所述模型训练过程包括数据选择、数据划分、数据预处理和算法选择四步;同时,模型训练单元还包括参数配置模块,用于对模型训练过程采用的具体训练算法进行预报算法的参数配置;
所述大数据分析单元根据采集到的氧化铝生产运行数据运用训练好的模型进行氧化铝生产指标预报结果与实际值之间的相关性系数计算,并根据相关性系数判断是否需要进行模型参数校正;
所述预报模型参数校正单元根据最近的生产工况数据,按照模型训练参数配置 继续训练预报模型,从而得到适应新工况的校正后的预报模型参数;
所述预报模型分析单元采用校正后的预报模型参数与当前预报模型参数对预报模型训练方案管理模块中存储的氧化铝生产指标数据进行预报分析;如果采用校正后的模型参数预报氧化铝生产指标的结果比当前参数的预报模型准确性好,则选择校正后的预报模型参数,否则维持当前预报模型参数不变;
所述模型校正参数下载单元提供校正后的预报模型参数的导出功能,以便下载到边缘端的氧化铝生产指标预报计算机的预报模型管理单元中;同时还提供了边缘端的氧化铝生产指标预报计算机中软件的自动远程更新校正后的模型参数的在线下载接口。
所述云端模型训练服务器上还包括模型训练评估单元用于对训练得到的预报模型进行性能评估;所述性能评估采用均方误差、平均绝对值误差、平均绝对百分误差及预报值与实际值相关性系数这些指标进行对比分析。
所述设置在边缘端的氧化铝生产指标预报计算机的软件包括预报模型管理单元和智能预报单元;
所述预报模型管理单元实现对本地边缘端模型训练方案的管理,包括模型选择模块与模型参数校正模块;模型选择模块根据氧化铝生产过程的判断结果,选择对应生产过程的氧化铝生产指标预报模型进行预报,边缘端与云端均有相同的多种氧化铝生产指标预报模型训练方案;模型参数校正模块导入云端预报模型训练服务器校正后的预报模型参数,或者直接进行校正参数的远程更新,以便在新的工况下仍然能够保证模型预报结果的准确性;
所述智能预报单元在边缘端使用训练好的预报模型运用传入边缘端的氧化铝生产全流程实际数据针对不同生产过程对氧化铝生产指标进行预报。
一种氧化铝生产指标的云-边协同预报方法,如图2所示,包括氧化铝生产指标数据的采集、云端训练服务器的运行和边缘端氧化铝生产指标预报计算机的运行三部分;
所述氧化铝生产指标数据采集通过数据采集装置采集氧化铝生产所产生的实际运行数据,并将采集到的数据传输给云端训练服务器的预报模型训练方案管理单元和边缘端的氧化铝生产指标预报计算机的数据预处理单元;
所述云端训练服务器的运行包括模型训练过程与模型参数校正过程;
所述模型训练过程为:
步骤1云端训练服务器的模型训练单元根据一定时间内数据采集装置采集的氧化铝实际运行所产生的数据及选择的预报算法进行模型训练,并将模型训练结果下传给模型训练评估单元;所述模型训练过程包括以下过程:
数据选择:根据训练目的对数据采集装置采集到的氧化铝生产全流程数据将进行区间划分;
本实施例中,选取某段时间内500个氧化铝溶出过程样本数据进行模型训练,选取预报的生产指标为溶出率,选取的数据如表1所示;
表1某段时间内500个氧化铝溶出过程运行样本数据
数据划分:选用数据划分方法将选择的数据划分为训练集与测试集;
本实施例中,选用80%数据为训练集,20%数据为测试集,并选用随机划分方式;
数据预处理:根据不同目标及数据形式选用不同数据预处理算法;
本实施例选用标准化算法进行数据预处理,数据预处理算法可自行在算法管理模块进行新建、更新等操作;
算法选择:从算法与模型库管理单元选择进行模型训练的算法;
本实施例选用ε支持向量机算法进行模型训练,该算法需要配置的参数分别为 惩罚因子C,误差精度要求ε和核函数,本实施例中的核函数选择RBF函数,参数C和ε分别为C=52,ε=1.9。
步骤2模型训练评估单元对模型训练单元训练得到的预报模型进行性能评估;性能评估指标包括均方误差、平均绝对值误差、平均绝对百分误差及预报值与实际值的相关性系数;
本实施例中,选用平均绝对误差对模型性能进行评估,平均绝对误差值为0.2,小于设定的阈值0.5,表明该模型训练效果很好,可以作为溶出率预报模型存入预报模型训练方案管理单元;
所述模型参数校正过程为:
步骤C1:大数据分析单元根据采集到的新的氧化铝生产运行数据运用训练好的模型进行氧化铝生产指标预报结果与实际值之间的相关性系数计算,并根据相关性系数判断是否需要进行模型参数校正,如果氧化铝生产指标预报结果与实际值之间的相关性系数小于根据具体生产指标要求设定的相关性系数阈值,则执行步骤2进行模型参数校正,否则不进行模型参数校正;
步骤C2:预报模型参数校正单元根据大数据分析单元得到的大数据分析结果,按照模型训练单元设置的预报模型参数配置继续训练预报模型,从而得到适应大数据分析结果的校正后的预报模型参数;
步骤C3:预报模型分析单元采用校正后的预报模型参数与当前预报模型参数对预报模型训练方案管理模块中存储的氧化铝生产指标数据进行预报分析;如果采用校正后的模型参数预报氧化铝生产指标的结果比当前参数的预报模型准确性好,则选择校正后的预报模型参数,否则维持当前预报模型参数不变;
所述边缘端的氧化铝产品指标预报计算机的运行过程为:
步骤S1:模型选择模块根据氧化铝生产过程选择所要预测的氧化铝生产指标预报模型;
步骤S2:模型参数校正模块通过模型校正参数下载单元导入在云端预报模型训练服务器校正后的预报模型参数,或者在云端直接进行校正参数的远程更新;
步骤S3:智能预报单元在边缘端使用训练好的预报模型针对不同生产过程的氧化铝生产指标进行预报。
本实例中,给出4组运用云端训练好的预报模型在边缘端对生产指标溶出率进行智能预报,得到的预报结果如表2所示;
表2生产指标溶出率的智能预报结果
[Table 1]
最后应说明的是:以上实施例仅用以说明本发明的技术方案,而非对其限制;尽管参照前述实施例对本发明进行了详细的说明,本领域的普通技术人员应当理解:其依然可以对前述实施例所记载的技术方案进行修改,或者对其中部分或者全部技术特征进行等同替换;而这些修改或者替换,并不使相应技术方案的本质脱离本发明权利要求所限定的范围。
Claims (7)
- 一种氧化铝生产指标的云-边协同预报系统,其特征在于:包括数据采集装置、硬件平台和设置在硬件平台上的软件系统;所述数据采集装置用于采集氧化铝生产过程的全流程实际数据,并将采集到的数据以固定时间间隔传输给云端服务器和边缘端的氧化铝生产指标预报计算机上;所述硬件平台包括云端模型训练服务器与边缘端的氧化铝生产指标预报计算机;所述软件系统包括设置在云端模型训练服务器的软件和设置在边缘端的氧化铝生产指标预报计算机的软件;所述设置在云端模型训练服务器的软件实现对氧化铝生产指标预报算法的选择管理,并将通过数据采集装置采集到的氧化铝生产过程的全流程实际数据对氧化铝生产指标预报模型进行训练、评估,并将采集到的数据运用到生产指标预报模型中,将模型预报结果与实际生产指标数据进行大数据分析,进而对预报模型参数进行校正;所述设置在边缘端的氧化铝生产指标预报计算机的软件能够实现从云端模型训练服务器导入云端预报模型训练服务器校正后的预报模型参数,或者直接进行校正参数的远程更新,进而将数据采集装置采集到的氧化铝生产过程的全流程实际数据传入边缘端,在边缘端使用训练好的预报模型针对不同生产过程对氧化铝生产指标进行预报。
- 根据权利要求1所述的一种氧化铝生产指标的云-边协同预报系统,其特征在于:所述设置在云端模型训练服务器的软件包括预报模型训练方案管理单元、算法与模型库管理单元、大数据分析单元、模型训练单元、预报模型参数校正单元、预报模型分析单元和模型校正参数下载单元;所述预报模型训练方案管理单元用于管理训练预报模型时使用的训练样本数据和存储模型参数,以便在分析预报模型性能时进行对比;所述算法与模型库管理单元用于管理得到的氧化铝生产指标预报 模型及模型训练过程所采用的预报算法和系统内置的常用数据处理算法,并提供预报算法的新建、更新与删除功能;所述模型训练单元用于根据一定时间内数据采集装置采集的氧化铝实际运行所产生的数据及选择的预报算法进行模型训练,并将模型训练结果下传给模型训练评估单元;所述模型训练过程包括数据选择、数据划分、数据预处理和算法选择四步;所述大数据分析单元根据采集到的氧化铝生产运行数据运用训练好的模型进行氧化铝生产指标预报结果与实际值之间的相关性系数计算,并根据相关性系数判断是否需要进行模型参数校正;所述预报模型参数校正单元根据最近的生产工况数据,按照模型训练参数配置继续训练预报模型,从而得到适应新工况的校正后的预报模型参数;所述预报模型分析单元采用校正后的预报模型参数与当前预报模型参数对预报模型训练方案管理模块中存储的氧化铝生产指标数据进行预报分析;如果采用校正后的模型参数预报氧化铝生产指标的结果比当前参数的预报模型准确性好,则选择校正后的预报模型参数,否则维持当前预报模型参数不变;所述模型校正参数下载单元提供校正后的预报模型参数的导出功能,以便下载到边缘端的氧化铝生产指标预报计算机的预报模型管理单元中;同时还提供了边缘端的氧化铝生产指标预报计算机中软件自动远程更新校正后的模型参数的在线下载接口。
- 根据权利要求2所述的一种氧化铝生产指标的云-边协同预报系统,其特征在于:所述云端模型训练服务器上还包括模型训练评估单元用于对训练得到的预报模型进行性能评估;所述性能评估采用均方误差、平均绝对值误差、平均绝对百分误差及预报值与实际值相关性系数这些指标进行对比分析。
- 根据权利要求3所述的一种氧化铝生产指标的云-边协同预报系统,其特征在于:所述模型训练单元还内置参数配置模块,用于对 模型训练过程采用的具体训练算法进行预报算法的参数配置。
- 根据权利要求4所述的一种氧化铝生产指标的云-边协同预报系统,其特征在于:所述设置在边缘端的氧化铝生产指标预报计算机的软件包括预报模型管理单元和智能预报单元;所述预报模型管理单元实现对本地边缘端模型训练方案的管理,包括模型选择模块与模型参数校正模块;模型选择模块根据氧化铝生产过程的判断结果,选择对应生产过程的氧化铝生产指标预报模型进行预报,边缘端与云端均有相同的多种氧化铝生产指标预报模型训练方案;模型参数校正模块导入云端预报模型训练服务器校正后的预报模型参数,或者直接进行校正参数的远程更新,以便在新的工况下仍然能够保证模型预报结果的准确性;所述智能预报单元在边缘端使用训练好的预报模型运用传入边缘端的氧化铝生产全流程实际数据针对不同生产过程对氧化铝生产指标进行预报。
- 一种氧化铝生产指标的云-边协同预报方法,基于权利要求5所述的一种氧化铝生产指标的云-边协同预报系统进行预报,其特征在于:包括氧化铝生产指标数据的采集、云端训练服务器的运行和边缘端氧化铝生产指标预报计算机的运行三部分;所述氧化铝生产指标数据采集通过数据采集装置采集氧化铝生产所产生的实际运行数据,并将采集到的数据传输给云端训练服务器的模型训练单元和边缘端的氧化铝生产指标预报计算机智能预报单元;所述云端训练服务器的运行包括模型训练过程与模型参数校正过程;所述模型训练过程为:步骤1:云端训练服务器的模型训练单元根据一定时间内数据采集装置采集的氧化铝实际运行所产生的数据及选择的预报算法进行模型训练,并将模型训练结果下传给模型训练评估单元;步骤2:模型训练评估单元对模型训练单元训练得到的预报模型进行性能评估;性能评估指标包括均方误差、平均绝对值误差、平均绝对百分误差及预报值与实际值的相关性系数;所述模型参数校正过程为:步骤C1:大数据分析单元根据采集到的氧化铝生产运行数据运用训练好的模型进行氧化铝生产指标预报结果与实际值之间的相关性系数计算,并根据相关性系数判断是否需要进行模型参数校正,如果氧化铝生产指标预报结果与实际值之间的相关性系数小于根据具体生产指标要求设定的相关性系数阈值,则执行步骤2进行模型参数校正,否则不进行模型参数校正;步骤C2:预报模型参数校正单元根据大数据分析单元得到的大数据分析结果,按照模型训练单元设置的预报模型参数配置继续训练预报模型,从而得到适应大数据分析结果的校正后的预报模型参数;步骤C3:预报模型分析单元采用校正后的预报模型参数与当前预报模型参数对预报模型训练方案管理模块中存储的氧化铝生产指标数据进行预报分析;如果采用校正后的模型参数预报氧化铝生产指标的结果比当前参数的预报模型准确性好,则选择校正后的预报模型参数,否则维持当前预报模型参数不变;所述边缘端的氧化铝产品指标预报计算机的运行过程为:步骤S1:模型选择模块根据氧化铝生产过程选择所要预测的氧化铝生产指标预报模型;步骤S2:模型参数校正模块通过模型校正参数下载单元导入在云端预报模型训练服务器校正后的预报模型参数,或者在云端直接进行校正参数的远程更新;步骤S3:智能预报单元在边缘端使用训练好的预报模型针对不同生产过程的氧化铝生产指标进行预报。
- 根据权利要求6所述的一种氧化铝生产指标的云-边协同预报方法 ,其特征在于:所述模型训练过程具体包括以下过程:数据选择:根据训练目的对数据采集装置采集到的氧化铝生产全流程数据将进行区间划分;数据划分:选用数据划分方法将选择的数据划分为训练集与测试集;数据预处理:根据不同目标及数据形式选用不同数据预处理算法;算法选择:从算法与模型库管理单元选择进行模型训练的算法。
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
US17/597,189 US20220309393A1 (en) | 2019-07-16 | 2019-07-18 | Cloud-edge collaboration forecasting system and method for aluminum oxide production indexes |
Applications Claiming Priority (2)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201910640654.3A CN110363355B (zh) | 2019-07-16 | 2019-07-16 | 一种氧化铝生产指标的云-边协同预报系统及方法 |
CN201910640654.3 | 2019-07-16 |
Publications (1)
Publication Number | Publication Date |
---|---|
WO2021007845A1 true WO2021007845A1 (zh) | 2021-01-21 |
Family
ID=68219577
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
PCT/CN2019/096512 WO2021007845A1 (zh) | 2019-07-16 | 2019-07-18 | 一种氧化铝生产指标的云-边协同预报系统及方法 |
Country Status (3)
Country | Link |
---|---|
US (1) | US20220309393A1 (zh) |
CN (1) | CN110363355B (zh) |
WO (1) | WO2021007845A1 (zh) |
Cited By (2)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN113220455A (zh) * | 2021-05-21 | 2021-08-06 | 润联软件系统(深圳)有限公司 | 一种基于云边协同的数据预处理方法和系统 |
CN113487156A (zh) * | 2021-06-30 | 2021-10-08 | 同济大学 | 一种基于云-边架构的工人行为监控识别方法及装置 |
Families Citing this family (4)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN110719210B (zh) * | 2019-12-05 | 2020-08-28 | 赣江新区智慧物联研究院有限公司 | 一种基于云边协同的工业设备预测性维护方法 |
CN111258984B (zh) * | 2020-01-17 | 2021-06-22 | 东北大学 | 工业大数据环境下的产品质量端-边-云协同预报方法 |
CN117149361B (zh) * | 2023-10-30 | 2023-12-29 | 北京万界数据科技有限责任公司 | 一种训练模型多端协同优化系统 |
CN118485451A (zh) * | 2024-05-24 | 2024-08-13 | 海南天地神农科技有限公司 | 基于源头数据验证机制的农产品溯源方法及系统 |
Citations (7)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN101187803A (zh) * | 2007-12-06 | 2008-05-28 | 宁波思华数据技术有限公司 | 基于数据挖掘技术的氨合成装置生产优化方法 |
CN103955760A (zh) * | 2014-04-22 | 2014-07-30 | 江苏祥兆文具有限公司 | 铝杆氧化染色工艺参数优化专家系统 |
CN104076686A (zh) * | 2013-03-29 | 2014-10-01 | 沈阳铝镁设计研究院有限公司 | 一种氧化铝生产过程动态成本控制方法 |
US8965625B2 (en) * | 2012-02-01 | 2015-02-24 | Afton Chemical Corporation | System and method for extending a lubricant discard interval |
CN104570739A (zh) * | 2015-01-07 | 2015-04-29 | 东北大学 | 基于云和移动终端的选矿多生产指标优化决策系统及方法 |
CN108092319A (zh) * | 2017-12-13 | 2018-05-29 | 国家电网公司 | 一种短期风电功率预测的不确定性分析方法及装置 |
CN109032094A (zh) * | 2018-08-15 | 2018-12-18 | 东北大学 | 一种基于核磁共振分析仪的原油快速评价建模云平台 |
Family Cites Families (3)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN104299045B (zh) * | 2014-09-24 | 2017-05-10 | 东北大学 | 一种选矿生产全流程精矿产量预报系统及方法 |
US10949807B2 (en) * | 2017-05-04 | 2021-03-16 | Servicenow, Inc. | Model building architecture and smart routing of work items |
CN108375808A (zh) * | 2018-03-12 | 2018-08-07 | 南京恩瑞特实业有限公司 | Nriet基于机器学习的大雾预报方法 |
-
2019
- 2019-07-16 CN CN201910640654.3A patent/CN110363355B/zh active Active
- 2019-07-18 US US17/597,189 patent/US20220309393A1/en active Pending
- 2019-07-18 WO PCT/CN2019/096512 patent/WO2021007845A1/zh active Application Filing
Patent Citations (7)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN101187803A (zh) * | 2007-12-06 | 2008-05-28 | 宁波思华数据技术有限公司 | 基于数据挖掘技术的氨合成装置生产优化方法 |
US8965625B2 (en) * | 2012-02-01 | 2015-02-24 | Afton Chemical Corporation | System and method for extending a lubricant discard interval |
CN104076686A (zh) * | 2013-03-29 | 2014-10-01 | 沈阳铝镁设计研究院有限公司 | 一种氧化铝生产过程动态成本控制方法 |
CN103955760A (zh) * | 2014-04-22 | 2014-07-30 | 江苏祥兆文具有限公司 | 铝杆氧化染色工艺参数优化专家系统 |
CN104570739A (zh) * | 2015-01-07 | 2015-04-29 | 东北大学 | 基于云和移动终端的选矿多生产指标优化决策系统及方法 |
CN108092319A (zh) * | 2017-12-13 | 2018-05-29 | 国家电网公司 | 一种短期风电功率预测的不确定性分析方法及装置 |
CN109032094A (zh) * | 2018-08-15 | 2018-12-18 | 东北大学 | 一种基于核磁共振分析仪的原油快速评价建模云平台 |
Cited By (2)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN113220455A (zh) * | 2021-05-21 | 2021-08-06 | 润联软件系统(深圳)有限公司 | 一种基于云边协同的数据预处理方法和系统 |
CN113487156A (zh) * | 2021-06-30 | 2021-10-08 | 同济大学 | 一种基于云-边架构的工人行为监控识别方法及装置 |
Also Published As
Publication number | Publication date |
---|---|
CN110363355A (zh) | 2019-10-22 |
US20220309393A1 (en) | 2022-09-29 |
CN110363355B (zh) | 2022-11-29 |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
WO2021007845A1 (zh) | 一种氧化铝生产指标的云-边协同预报系统及方法 | |
WO2021179572A1 (zh) | 运维系统异常指标检测模型优化方法、装置及存储介质 | |
CN104392390B (zh) | 一种基于topsis模型的智能变电站二次设备评估方法 | |
CN110352660B (zh) | 一种脱绒棉种活力快速无损检测信息处理方法和装置 | |
WO2021007871A1 (zh) | 一种基于云边协同的氧化铝生产运行优化系统及方法 | |
CN105388876B (zh) | 获得基于班组的间歇型化工生产工艺符合度的方法及装置 | |
CN117420967B (zh) | 一种软件采集数据存储性能提升方法和系统 | |
CN104407688A (zh) | 基于树回归的虚拟化云平台能耗测量方法及系统 | |
CN106610616B (zh) | 一种基于大数据分析的性能自动调优方法及装置 | |
CN114155072A (zh) | 基于大数据分析的财务预测模型构建方法及系统 | |
JP2022510667A (ja) | エネルギー消費量を予測するための方法および装置 | |
CN117556366B (zh) | 基于数据筛选的数据异常检测系统及方法 | |
CN110413703A (zh) | 基于人工智能的监控指标数据的分类方法及相关设备 | |
CN115396981A (zh) | 一种基于大数据技术的智能监测系统 | |
CN118485311B (zh) | 基于云的数控设备管理系统及方法 | |
CN110378586B (zh) | 基于动态闭环知识管理的变电设备缺陷预警方法及系统 | |
CN116681426A (zh) | 一种工业互联网设备预测性维护系统及方法 | |
CN116938676A (zh) | 一种基于数据源分析的通信风险联合预警方法 | |
WO2020034248A1 (zh) | 一种基于核磁共振分析仪的原油快速评价建模云平台 | |
CN110147403A (zh) | 农业大数据融合方法、装置、设备及存储介质 | |
CN117196257A (zh) | 配电网工程辅助结算系统 | |
CN108826595A (zh) | 一种空气净化器的控制方法及系统 | |
CN101706659A (zh) | 一种车载设备生产的过程控制系统及其实现方法 | |
CN116933628A (zh) | 一种企业用电负荷监测方法和系统 | |
CN111080037A (zh) | 一种基于深度神经网络的短期电力负荷预测方法及装置 |
Legal Events
Date | Code | Title | Description |
---|---|---|---|
121 | Ep: the epo has been informed by wipo that ep was designated in this application |
Ref document number: 19937451 Country of ref document: EP Kind code of ref document: A1 |
|
NENP | Non-entry into the national phase |
Ref country code: DE |
|
122 | Ep: pct application non-entry in european phase |
Ref document number: 19937451 Country of ref document: EP Kind code of ref document: A1 |