CN110456756B - Method suitable for online evaluation of global operation state in continuous production process - Google Patents

Method suitable for online evaluation of global operation state in continuous production process Download PDF

Info

Publication number
CN110456756B
CN110456756B CN201910880712.XA CN201910880712A CN110456756B CN 110456756 B CN110456756 B CN 110456756B CN 201910880712 A CN201910880712 A CN 201910880712A CN 110456756 B CN110456756 B CN 110456756B
Authority
CN
China
Prior art keywords
layer
real
production process
evaluation
continuous production
Prior art date
Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
Active
Application number
CN201910880712.XA
Other languages
Chinese (zh)
Other versions
CN110456756A (en
Inventor
王雅琳
李灵
袁小锋
孙备
吴东哲
王思哲
阳春华
Current Assignee (The listed assignees may be inaccurate. Google has not performed a legal analysis and makes no representation or warranty as to the accuracy of the list.)
Central South University
Original Assignee
Central South University
Priority date (The priority date is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the date listed.)
Filing date
Publication date
Application filed by Central South University filed Critical Central South University
Publication of CN110456756A publication Critical patent/CN110456756A/en
Application granted granted Critical
Publication of CN110456756B publication Critical patent/CN110456756B/en
Active legal-status Critical Current
Anticipated expiration legal-status Critical

Links

Images

Classifications

    • G—PHYSICS
    • G05—CONTROLLING; REGULATING
    • G05B—CONTROL OR REGULATING SYSTEMS IN GENERAL; FUNCTIONAL ELEMENTS OF SUCH SYSTEMS; MONITORING OR TESTING ARRANGEMENTS FOR SUCH SYSTEMS OR ELEMENTS
    • G05B19/00—Programme-control systems
    • G05B19/02—Programme-control systems electric
    • G05B19/418—Total factory control, i.e. centrally controlling a plurality of machines, e.g. direct or distributed numerical control [DNC], flexible manufacturing systems [FMS], integrated manufacturing systems [IMS], computer integrated manufacturing [CIM]
    • G05B19/41875—Total factory control, i.e. centrally controlling a plurality of machines, e.g. direct or distributed numerical control [DNC], flexible manufacturing systems [FMS], integrated manufacturing systems [IMS], computer integrated manufacturing [CIM] characterised by quality surveillance of production
    • G—PHYSICS
    • G05—CONTROLLING; REGULATING
    • G05B—CONTROL OR REGULATING SYSTEMS IN GENERAL; FUNCTIONAL ELEMENTS OF SUCH SYSTEMS; MONITORING OR TESTING ARRANGEMENTS FOR SUCH SYSTEMS OR ELEMENTS
    • G05B19/00—Programme-control systems
    • G05B19/02—Programme-control systems electric
    • G05B19/418—Total factory control, i.e. centrally controlling a plurality of machines, e.g. direct or distributed numerical control [DNC], flexible manufacturing systems [FMS], integrated manufacturing systems [IMS], computer integrated manufacturing [CIM]
    • G05B19/41885—Total factory control, i.e. centrally controlling a plurality of machines, e.g. direct or distributed numerical control [DNC], flexible manufacturing systems [FMS], integrated manufacturing systems [IMS], computer integrated manufacturing [CIM] characterised by modeling, simulation of the manufacturing system
    • G—PHYSICS
    • G05—CONTROLLING; REGULATING
    • G05B—CONTROL OR REGULATING SYSTEMS IN GENERAL; FUNCTIONAL ELEMENTS OF SUCH SYSTEMS; MONITORING OR TESTING ARRANGEMENTS FOR SUCH SYSTEMS OR ELEMENTS
    • G05B2219/00—Program-control systems
    • G05B2219/30—Nc systems
    • G05B2219/32—Operator till task planning
    • G05B2219/32339—Object oriented modeling, design, analysis, implementation, simulation language
    • 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
    • Y02P90/02—Total factory control, e.g. smart factories, flexible manufacturing systems [FMS] or integrated manufacturing systems [IMS]

Abstract

The invention belongs to the field of real-time optimization control of industrial processes, and discloses a method suitable for online evaluation of global operation conditions in a continuous production process.

Description

Method suitable for online evaluation of global operation state in continuous production process
Technical Field
The invention relates to the field of real-time optimization control of industrial processes, in particular to a method suitable for online evaluation of global operation conditions in a continuous production process.
Background
The real-time optimization control method is an effective means for solving the problem of complex process industrial process optimization and control, and optimizes economic benefits on the premise of comprehensively considering economic information, device physical constraints and device freedom. Generally, loop control and process operation optimization are combined, a two-layer structure is adopted, economic performance indexes are optimized through plan scheduling in the upper layer, set values of a bottom layer control loop are generated, controlled variables are enabled to track the set values through a controller in the bottom layer, therefore, the process is enabled to operate in an economic optimization state as far as possible, and organic association of enterprise operation targets and production operation is achieved. The evaluation aiming at the running condition of the continuous production process is an important link and an indispensable foundation for realizing the effective running optimization and control of the continuous production process.
From the view of the whole flow of continuous production, the operation optimization and control are realized in a layered manner, and mainly relate to three layers of a plan scheduling layer, a real-time optimization layer and a process control layer (for example, the classification is disclosed in 'research on optimization decision method of mineral separation production whole flow operation indexes under dynamic environment' published by doctor thesis of northeast China university in 2012). Currently, there are many achievements for evaluating the running status of the continuous production process, and the achievements are applied to industrial production practice to obtain better effects, but the existing evaluation method has the following defects: firstly, research is often carried out only on the operation condition of a single layer, and the global operation condition cannot be completely reflected. For example, only the performance evaluation of the control system/loop is performed, and only the process control layer is covered, and only the operation performance of the control system/loop itself is reflected. Or only the optimal running condition of the real-time optimization layer is evaluated, and the effect can not completely reflect the global running condition. And secondly, the optimal operation condition of the real-time optimization layer is evaluated, and the feedback influence of the performance of a control system/loop of a process control layer on the real-time optimization layer is not considered in the comprehensive evaluation index, so that the result is possibly not an accurate, practical and feasible optimal operation index. When the production process is subjected to steady-state optimization in the prior art, the performance of a popular assumed control system is in an optimal state, namely, the control system can accurately track the set value of the control system, so that the real-time optimization layer operation index is extremely close to the target value. However, in the actual production process, the performance of the control system cannot be guaranteed to be always in the optimal state, so the assumed condition during steady state optimization is not satisfied, that is, the optimal operation index obtained by performing steady state optimization at this time is only the optimal operation index in an ideal state (which cannot be actually realized). The evaluation of the running condition performance of the real-time optimization layer according to the optimal running index in an ideal state causes the deviation of the running performance evaluation result of the real-time optimization layer, and the running performance of the real-time optimization layer cannot be truly reflected. And thirdly, the running condition of the real-time optimization layer is evaluated, and the inlet working conditions are not distinguished, so that the evaluation method of each running mode is the same. Due to variations in production environment, operating regimes, etc., a production process may have multiple modes of operation. At present, part of evaluation methods firstly classify and evaluate the operation modes offline, and then identify the operation modes online, but when classifying the operation modes, specific entry conditions, operation modes and the like are not clear, that is, the operation modes corresponding to the operation modes are not clear, and the evaluation method for each operation mode is the same. In practice, the entrance working conditions have a great influence on the completion of the real-time optimization layer operation indexes, and the worse the entrance conditions, the more difficult the operation indexes are to be completed, so that the evaluation standards of the operation conditions under different entrance working conditions are to be treated by some means.
In summary, the existing method for evaluating the production running status can only analyze the running status of a single level, cannot completely reflect the global running status, cannot meet all requirements of online evaluation of the global running status in the production process, and is not beneficial to performance degradation diagnosis in the subsequent link and dynamic self-adjustment in the production process, so that the online evaluation of the global running status in the continuous production process needs to be carried out by integrating three levels, and the online evaluation of the global running status in the continuous production process is accurately realized.
Disclosure of Invention
The invention aims to solve the technical problem of overcoming the defect that the prior art can only analyze the single-level operation condition of the production process, and provides a method suitable for online evaluation of the global operation condition in the continuous production process.
In order to solve the problems, the method for online evaluation of the global operation condition in the continuous production process is provided, and comprises the following steps:
a method suitable for online evaluation of global operation conditions in a continuous production process comprises the following steps:
s1, dividing a continuous production process into a planned scheduling layer, a real-time optimization layer and a process control layer from top to bottom, and determining operation indexes of each layer; a feedback relation from bottom to top exists among the process control layer, the real-time optimization layer and the planning and scheduling layer;
s2, screening out a process key parameter control loop based on production process analysis and mechanism analysis, and evaluating the performance of the key parameter control loop:
s21, screening out a process key parameter control loop based on production process analysis and mechanism analysis;
s22, collecting key parameter control loop operation data, and performing outlier rejection and wavelet denoising on the operation data;
s23, acquiring a random stable part of the non-stable time sequence of the operation data of the key parameter control loop based on an EMD method, and taking the random stable part as a stable time sequence;
s24, establishing a time sequence analysis model for the stable time sequence, and identifying model parameters;
s25, estimating the time delay of the operation data of the key parameter control loop;
s26, based on the time sequence analysis model and the time delay, performing control system performance evaluation on a key parameter control loop by adopting a minimum variance reference;
s3, respectively carrying out online evaluation on the operation performance of the real-time optimization layer from two aspects of online evaluation of the operation state based on historical data and online evaluation of the uncertain optimization operation state considering the feedback influence of the key parameter control loop performance on the real-time optimization layer:
s31, collecting continuous production process operation data, and performing outlier rejection and wavelet denoising on the continuous production process operation data; the continuous production process operation data comprise raw material detection data, process variable data and product quality test data;
s32, classifying inlet working conditions in the production process by using a K-means clustering method based on the raw material detection data to obtain inlet working condition classification results;
s33, classifying the product yield in the continuous production process based on the plan scheduling requirement to obtain a product yield classification result;
s34, constructing a comprehensive product quality and yield index system based on multiple products by utilizing a multi-target comprehensive evaluation method based on the production characteristics of the multiple products in the continuous production process, and taking the comprehensive product quality and yield index system as a real-time optimized layer online evaluation index to obtain a comprehensive product quality and yield index;
s35, establishing a real-time optimization layer operation condition evaluation standard based on historical data in the continuous production process based on the inlet working condition classification result, the product yield classification result, the product quality assay data and the comprehensive product quality and yield indexes, and dividing the real-time optimization layer operation performance into 5 grades of 'excellent', 'good', 'medium', 'poor' and 'unqualified';
s36, based on the process variable data, selecting key process variable data according to Pearson correlation analysis results of process variables and product quality and yield and process mechanism analysis;
s37, establishing a prediction model of comprehensive product quality and yield indexes by using a JIT-stacking method based on the key process variable data;
s38, carrying out online evaluation on the running condition of the real-time optimization layer from the aspect of online evaluation of the running condition based on historical data based on the comprehensive product quality and yield index prediction result and the running condition evaluation standard of the real-time optimization layer in the continuous production process;
s39, based on the key parameter control loop performance evaluation result, performing uncertain optimization on the running condition of the real-time optimization layer, and determining the uncertain optimization result as a real-time optimization layer uncertain optimization evaluation benchmark in the continuous production process, wherein an uncertain optimization model specifically comprises the following steps: the method comprises the following steps of taking the minimum forward deviation of product quality and yield required by planned scheduling as an optimization target, and taking production process energy balance, material balance and key process variable production requirements as constraints;
s310, synthesizing the operating condition evaluation standard of the continuous production process real-time optimization layer based on the historical data and the uncertain optimization evaluation standard of the production process real-time optimization layer, determining the operating condition evaluation standard of the continuous production process real-time optimization layer, specifically revising the operating condition evaluation standard of the real-time optimization layer according to the uncertain optimization operation condition evaluation result, and performing online evaluation on the operating condition of the real-time optimization layer according to the revised evaluation standard;
s4, analyzing the feedback influence of the real-time optimization layer operation performance on the operation performance of the planned scheduling layer, carrying out uncertain optimization on the optimal operation index of the planned scheduling layer, and carrying out online evaluation on the operation performance of the planned scheduling layer according to the index:
s41, collecting related data of short-term plan scheduling of the continuous production process, and when sufficient raw material reserve is available, considering that the short-term plan scheduling of the continuous production process is unrelated to the reserve of the raw materials;
s42, performing uncertain optimization on the operation condition of a planned scheduling layer based on the short-term planned scheduling related data of the continuous production process and the operation performance of the real-time optimization layer, and determining an uncertain optimization result as an uncertain optimization evaluation benchmark of the planned scheduling layer of the continuous production process;
the uncertain optimization model specifically comprises the following steps: the method takes the maximum comprehensive commodity yield and the minimum unit energy factor energy consumption as optimization targets and takes the energy balance, material balance, raw material supply and sub-process operation performance in the production process as constraints;
s43, based on the continuous production process operation data of S31, selecting key process variable data according to a Pearson correlation analysis result and process mechanism analysis of process variables, comprehensive commodity yield and unit energy factor energy consumption;
s44, establishing a prediction model for synthesizing commodity yield and unit energy factor energy consumption by using a JIT-stacking method based on the key process variable data;
and S45, carrying out online evaluation on the operation performance of the planning and scheduling layer based on the comprehensive commodity yield and unit energy factor energy consumption prediction result and the uncertain optimization evaluation criterion of the planning and scheduling layer in the continuous production process.
Further, the operation index of each layer described in S1 is determined by:
the planning and scheduling layer determines a target value of daily comprehensive production indexes according to the target of monthly or quarterly comprehensive production indexes, the sub-process operation performance, the energy consumption and the raw material supply in the continuous production process;
the real-time optimization layer determines the target value of the operation index of each sub-process according to the target value of the daily comprehensive production index of the whole production process and simultaneously considering the capacity of sub-process equipment, the utilization efficiency of the sub-process equipment, the performance of a key parameter control loop, the property of raw materials and the process allowable range of the operation index;
the process control layer sets the set value of the key control system according to the target value of the operation index of the sub-process, so that the actual value of the operation index of the real-time optimization layer approaches to the target value of the operation index.
Further, the continuous production process short-term planning and scheduling related data described in S41 includes: raw material reserve, long-term comprehensive production index, intermediate product or waste product requirement and energy consumption requirement.
Further, S4 is followed by:
and S5, when the overall operation condition evaluation result is not good, tracing the reason of the non-good operation condition according to the evaluation index system of the operation condition of each layer.
Further, the uncertainty of the performance of the key control loop where the dependent variable of the key process variable is located in S39 is an uncertain variable, and is described in an uncertain manner.
Further, the uncertainty of the real-time optimization layer operation performance of the sub-process operation performance factor process described in S42 is an uncertain variable, and is described in an uncertain manner.
Further, the continuous production process includes complicated industrial processes including a hydrocracking process and a residue hydrogenation process.
Compared with the prior art, the technical scheme provided by the invention has the beneficial effects that: the comprehensive three-layer cross-layer collaborative online evaluation of the running state is realized, information transmission and feedback among the running states of each layer are integrated, and the overall running state can be evaluated more accurately; secondly, the feedback influence of the operation performance of the analysis process control layer on the operation performance of the real-time optimization layer is added, and the deviation of the real-time optimization layer operation performance evaluation result caused by evaluating the real-time optimization layer operation condition performance according to the optimal operation index in an ideal state is eliminated to the greatest extent; analyzing the feedback influence of the real-time optimized layer operation performance on the operation performance of the scheduling layer, and eliminating the deviation of the operation performance evaluation result of the scheduling layer caused by evaluating the operation condition performance of the scheduling layer according to the optimal operation index in an ideal state to the maximum extent; fourthly, the operation condition evaluation standards under different entrance working conditions are distinguished, so that the overall operation condition evaluation result is more accurate.
Drawings
FIG. 1 is a schematic diagram of a method for online evaluation of global operating conditions in a continuous production process.
FIG. 2 is a schematic overall flow chart of a method for online evaluation of global operating conditions of a hydrocracking process according to an embodiment of the present invention.
FIG. 3 is a flow chart illustrating the operation optimization and control of a hydrocracking process according to an embodiment of the present invention.
FIG. 4 is a simplified schematic diagram of a hydrocracking process including a hydrofinishing, hydrocracking, high and low pressure separation, fractionation system 4 according to an embodiment of the present invention.
FIG. 5 is a schematic diagram of inlet feed oil composition for a hydrocracking process according to an embodiment of the present invention.
FIG. 6 is a flow chart of inlet condition classification based on the K-means algorithm according to an embodiment of the present invention.
FIG. 7 is a schematic diagram of a hydrocracking process integrated product quality and yield indicator system according to an embodiment of the present invention.
Fig. 8 is a flow chart of a JIT-stacking model that integrates product quality and yield predictions in accordance with an embodiment of the present invention.
FIG. 9 is a graph illustrating the results of a real-time optimization layer operating condition evaluation for a hydrocracking process based on historical data, in accordance with an embodiment of the present invention.
FIG. 10 is a graph illustrating the results of an evaluation of the operating conditions of a hydrocracking process real-time optimization layer in accordance with an embodiment of the present invention that combines historical data-based real-time optimization layer operating condition evaluation with an online evaluation of the uncertain optimization operating conditions that take into account the feedback impact of the process control layer on the real-time optimization layer.
FIG. 11 is a schematic general flow chart of a method for online evaluation of global operating conditions of a two-residue oil hydrogenation process according to an embodiment of the invention.
FIG. 12 is a full flow chart of operation optimization and control of a second residue oil hydrogenation process according to an embodiment of the invention.
FIG. 13 is a simplified schematic of a residue hydrogenation scheme comprising a second hydrofinishing, high and low pressure separation, fractionation system 3 section according to an embodiment of the present invention.
FIG. 14 is a schematic diagram of inlet feed oil composition of a two-residue oil hydrogenation process according to an embodiment of the invention.
FIG. 15 is a schematic diagram of an integrated product quality and yield index system for a two-residue oil hydrogenation process according to an embodiment of the invention.
FIG. 16 is a graph showing the results of an evaluation of the operating conditions of the real-time optimization layer of the residue hydrogenation process based on historical data according to the second embodiment of the present invention.
Detailed Description
For further disclosure of the present invention, the following detailed description of the present invention will be made with reference to the accompanying drawings and examples:
in particular, the terms "upper" and "lower" in the technical solutions disclosed in the present invention are defined with reference to the spatial orientation of fig. 1.
The following is a specific example of the global operation condition online evaluation of two continuous production processes, namely a hydrocracking process and a residue hydrogenation process:
the first embodiment is as follows:
the hydrocracking process of a refinery in China is illustrated, and the method for online evaluation of the global operation condition of the continuous production process, which is provided by the invention, is described in detail by combining the drawings of FIG. 1 and FIG. 2, as follows:
the first step, dividing a hydrocracking flow continuous production process into a planned scheduling layer, a real-time optimization layer and a process control layer from top to bottom, and determining operation indexes of each layer; and a feedback relation from bottom to top exists among the process control layer, the real-time optimization layer and the planning and scheduling layer. The hydrocracking process operation optimization and control overall process is shown in fig. 3, and mainly relates to three layers, namely a planning scheduling layer, a real-time optimization layer and a process control layer. The scheduling layer determines the target value of the daily comprehensive production index of the hydrocracking process according to the target of the monthly or quarterly comprehensive production index of the hydrocracking process (comprehensive commodity rate, light oil yield, unit energy factor energy consumption and the like), the sub-process operation performance (hydrofining process operation performance, hydrocracking process operation performance, high-low pressure separation process operation performance and fractionation system operation performance), the energy consumption (electricity, water, hydrogen, steam and the like), the raw material supply (the supply amount of feed oil and the quality of feed oil) and other constraint conditions. The real-time optimization layer considers the sub-process equipment capacity (the hydrotreating process throughput, the hydrocracking process throughput, the high and low pressure analysis process throughput and the fractionation system throughput), the sub-process equipment operation performance (the hydrotreating process equipment utilization efficiency, the hydrocracking process equipment utilization efficiency, the high and low pressure separation process equipment utilization efficiency and the fractionation system equipment utilization efficiency), the sub-process control system performance (the hydrotreating process key parameter control loop performance, the hydrocracking process key parameter control loop performance, the high and low pressure separation process key parameter control loop performance and the fractionation system key parameter control loop performance), the catalyst activity (the hydrotreating process reactor catalyst activity, the hydrocracking process reactor catalyst activity and the like), the raw material properties (the feed oil flow and the feed oil component) according to the target value of daily comprehensive production indexes of the hydrocracking process, and determining the target value of the operation index of each sub-process, such as the process allowable range of the operation index. The process control layer sets the set value of the key parameter control loop according to the target value of the operation index of the sub-process, so that the actual value of the operation index of the real-time optimization layer approaches to the target value of the operation index as much as possible.
And secondly, screening out a process key parameter control loop based on production process analysis and mechanism analysis, and evaluating the performance of the key parameter control loop. The hydrocracking flow diagram is shown in figure 4 and mainly comprises a hydrofining process, a hydrocracking process, a high-low pressure separation process and a fractionation system 4. Through process mechanism analysis, the selected key parameter control loops are mainly distributed in a hydrocracking process reaction system (a hydrofining process and a hydrocracking process) and a fractionation system, as shown in table 1.
Table 1 hydrocracking run key control loop
Serial number Hydrocracking step Key parameter control loop
1 Hydrorefining Feed oil flow loop
2 Hydrorefining Reactor inlet pressure loop
3 Hydrorefining First bed inlet temperature loop
4 Hydrorefining Second bed inlet temperature loop
5 Hydrorefining Third bed inlet temperature loop
6 Hydrocracking First bed inlet temperature loop
7 Hydrocracking Second bed inlet temperature loop
8 Hydrocracking Third bed inlet temperature loop
9 Hydrocracking Fourth bed inlet temperature loop
10 Fractionation system Heating furnace outlet temperature loop
11 Fractionation system Fractionating tower top pressure loop
12 Fractionation system Fractionating tower top temperature loop
13 Fractionation system Tower bottom temperature hoist of diesel stripping tower
14 Fractionation system Extraction rate flow loop of diesel stripping tower
15 Fractionation system Bottom temperature loop of aviation kerosene stripping tower
16 Fractionation system Aviation kerosene stripping tower extraction rate flow loop
And (3) acquiring the operating data of 16 key parameter control loops in the table 1, and performing outlier rejection and wavelet denoising on the operating data. And then, based on an EMD method, acquiring a random stable part of the non-stable time sequence of the operation data of each key parameter control loop as a stable time sequence. And establishing an ARMA time sequence analysis model for each stable time sequence, and identifying model parameters by adopting a least square method. The ARMA model is a time series analysis model, i.e., an autoregressive moving average model, proposed by american statistician g.e.p Box and british statistician g.m.jenkins in the seventies of the twentieth century, and is shown below.
Figure GDA0002715142330000071
Wherein { (n), (n-1),., (n-q) } is a white noise sequence, which can be considered as an input to the model; { x (n), x (n-1),.., x (n-p) } is a stationary random sequence, which can be understood as the output of the model.
And estimating the time delay of the operation data of each key parameter control loop by adopting a cross-correlation function method, wherein a specific formula is shown as follows.
Figure GDA0002715142330000081
And performing performance evaluation on the key parameter control loop based on a Harris performance index in a minimum variance reference, wherein the Harris performance index is as follows:
Figure GDA0002715142330000082
wherein
Figure GDA0002715142330000083
To control the minimum variance of the system output under the control of the minimum variance controller,
Figure GDA0002715142330000084
and actually outputting the variance for the control system.
Performance is divided into four levels: a0.8, 1, B0.7, 0.8), C0.6, 0.7, D (0,0.6), the results of the control system performance evaluation of 16 key parameter control loops by using the output data of a certain day of the hydrocracking process are shown in Table 2.
TABLE 2 evaluation results of key control loop performance in hydrocracking process
Figure GDA0002715142330000085
And thirdly, online evaluation is carried out on the operation performance of the real-time optimization layer respectively from two aspects of online evaluation of the operation state based on historical data and online evaluation of the uncertain optimization operation state considering the feedback influence of the key parameter control loop performance on the real-time optimization layer. Whether the running state online evaluation based on historical data or the uncertain optimization running state online evaluation considering the feedback influence of the control loop performance on the real-time optimization layer is carried out, what the running target of the real-time optimization layer is needs to be determined firstly, and then the running state grade of the real-time optimization layer is divided according to the running target of the real-time optimization layer. Generally, the improvement of product quality and light oil yield, the reduction of energy consumption, the reduction of economic cost of optimization process, safety performance and the like can be taken as targets of real-time optimization, and the improvement of product quality and light oil yield is adopted in most researches. However, in the actual hydrocracking process, compared with the improvement of product quality and light oil yield, enterprises tend to prolong the service life of the catalyst as much as possible on the premise of ensuring the product quality and yield required by the planned scheduling, because the cost for replacing the catalyst is very expensive. If the product quality and the light oil yield are further improved, the heavy oil as the inlet raw material needs to be better cracked into the light oil, the reaction temperature of a reaction system needs to be increased to a certain extent, and the increase of the reaction temperature of the reaction system inevitably accelerates the deactivation of the catalyst activity, thereby losing the service life of the catalyst. Therefore, in order to prolong the service life of the catalyst as much as possible under the condition of ensuring normal production, the real-time optimization layer operation target is set in the patent to not excessively improve the product quality and the yield of the light oil, and the production requirement of the planned scheduling layer is completed only on the premise of ensuring production, namely, the smaller the forward deviation of the product quality (yield) from the planned scheduling requirement is, the better the product quality (yield) is. The mass forward bias is specifically described as follows:
Figure GDA0002715142330000091
wherein Δ yi,j(h) Is the mass positive deviation of the jth ingredient of the ith product,
Figure GDA0002715142330000092
and
Figure GDA0002715142330000093
upper and lower limits of the jth component of the ith product, y, respectively, of the planned scheduling requirementi,j(h) Is the actual value of the jth ingredient of the ith product. During specific calculation, a corresponding formula is selected according to whether the production requirement of the product quality is a double-side specification or a single-side specification. The forward bias in light oil yield is described as follows:
Figure GDA0002715142330000094
wherein Δ pi(h) Is the yield forward deviation of the ith product,
Figure GDA0002715142330000095
is the lower limit of the ith product yield, p, of the planned scheduling requirementi(h) Is the actual value of the yield of the ith product.
The historical data-based online evaluation of the operating conditions mainly comprises 4 parts of inlet working condition classification, light oil yield classification, multi-index comprehensive evaluation and key quality (yield) index prediction. The main reasons for the inlet working condition classification and the light oil yield classification are that the operation performance of the hydrocracking process real-time optimization layer is meaningful only by evaluating under the same background condition, different inlet conditions have no influence on the operation indexes of the real-time optimization layer, but have larger influence on the completeness of the operation indexes, and the worse the inlet conditions are, the harder the operation indexes are to be completed. The feed oil of the hydrocracking process mainly comprises vacuum wax oil in a tank area, light wax oil in the No. 1 and No. 2 atmospheric and vacuum processes, coker wax oil in a residual oil hydrogenation process (the coker wax oil is less and can be generally ignored) and cycle tail oil from fractionation, as shown in FIG. 5. Because the components of the feed oil need to be analyzed through laboratory tests and are generally detected once a day, the change condition of the inlet working condition is not favorably reflected in real time. However, the flow rate of the feed oil is measurable in real time, and the composition of 5 constituent oils is substantially unchanged, so that the inlet condition classification of the hydrocracking process can be realized by using the flow rates of 5 constituent oils instead of the assay data of the composition of the feed oil. The mahalanobis distance shown in the formula (6) is used as a similarity measurement function, and the hydrocracking process inlet working condition classification based on the K-means algorithm is shown in a flow chart shown in FIG. 6.
Figure GDA0002715142330000101
The light oil yield of the hydrocracking process is classified according to the scheduling requirement, and the light oil yield requirement is not changed greatly generally unless the supply and demand relationship of the market and policy to the light oil is changed obviously.
Since the main light oil products of the hydrocracking process are light naphtha, heavy naphtha, aviation kerosene and diesel oil, and each product oil has various components and different dimensions (such as light naphtha C4, C5 mass fraction, etc.), it is not favorable for the on-line evaluation of the real-time optimization layer operation status of the hydrocracking process. Therefore, in order to evaluate and analyze the real-time optimization layer operation performance, a comprehensive product quality and yield index system based on multiple products and multiple product components of the hydrocracking process needs to be constructed. The overall product quality and yield index system is shown in FIG. 7, which is divided into 4 layers. Wherein the 4 th layer is the component of each product oil, the 3 rd layer is the product oil, the 2 nd layer is the product quality and yield, and the 1 st layer is the material production index of the comprehensive product quality and yield. And then dividing a real-time optimization layer operation condition evaluation standard based on historical data according to an inlet working condition classification, a light oil yield classification, a comprehensive product quality and yield index system, and dividing the real-time optimization layer operation condition of the hydrocracking process into 5 grades of 'good', 'medium', 'poor' and 'unqualified', with the aim that the forward deviation of the product quality and the yield required by the plan scheduling is smaller and better.
Because the product composition indexes in the actual hydrocracking process need laboratory test analysis, the change of the product quality is not easily reflected in real time; meanwhile, the whole process of the hydrocracking process is about 3.5 hours, and the current product oil flow cannot reflect the current operation condition of the hydrocracking process in real time. Therefore, a soft measurement model integrating product quality and yield needs to be established. First, by analyzing the Pearson correlation of process variables with product quality and yield and the hydrocracking process flow process mechanism, key process variables were selected as shown in table 3. And then establishing a JIT-stacking prediction model for integrating the product quality and yield through the selected key process variables.
TABLE 3 Critical Process variables for hydrocracking runs
Figure GDA0002715142330000102
Figure GDA0002715142330000111
The stacking integrated model is characterized in that after a plurality of basic prediction models are trained by using initial training data, the prediction results of the basic prediction models are used as a new training set to learn a new prediction model, and compared with a single model, the stacking integrated model can achieve higher accuracy than a single classifier and can inhibit overfitting. The stacking model has better smoothness, can highlight the function of the basic model with better prediction effect in the integrated model, weaken the function of the basic model with poor prediction effect in the integrated model, and is superior to each single model. However, as time and conditions change, the model trained using the same fixed data is gradually distorted, and the prediction effect is gradually deteriorated, so that the model needs to be updated. Aiming at the problem of stack model updating, based on the concept of Just-In-Time (JIT), the method proposes samples aiming at single missing data, selects similar known data samples as a training set, and establishes a model independently, namely establishes a filling model for each missing sample, can utilize information of data around the samples to be filled to the maximum extent, and avoids the problem of unrealistic model, wherein the selected base models are a GBM algorithm, a random forest algorithm, a generalized linear regression model and a feedforward artificial neural network model respectively. A specific JIT-stacking model flow chart (published in master thesis of "research on hydrocracking process steady-state detection method and construction of working condition library system" of southern university in 2018) is shown in fig. 8.
The data filling method based on the JIT-stacking model comprises the following specific steps:
step 1: selecting other characteristics with strong correlation with the characteristics of the data to be filled by utilizing a step-by-step screening method;
step 2: selecting other samples with higher similarity with the samples to be filled as a new sample set, dividing the samples into a training set and a verification set, and using the samples to be filled as a test set;
and step 3: building a stacking model by using the training set and the verification set in the step 2, selecting GBM, RF, GLM and FNN as a first layer base model, and selecting logistic regression as a second layer base model;
and 4, step 4: and filling the missing data according to the stacking model established in the step 3.
And finally, carrying out online evaluation on the operation condition of the real-time optimization layer of the hydrocracking flow according to the JIT-stacking model prediction result of the comprehensive product quality and yield and the grade division standard of the operation condition of the real-time optimization layer of the hydrocracking flow. The online evaluation of the real-time optimized layer operation condition of a hydrocracking process of a certain factory in China from the perspective of historical data is carried out by utilizing the operation data of the hydrocracking process of the certain factory in China for one year, and the evaluation result is shown in figure 9.
According to the performance evaluation result of the key parameter control loop of the hydrocracking process in the table 2, the following uncertain optimization is carried out on the running condition of the real-time optimization layer by taking the minimum forward deviation of the product quality and the light oil yield required by plan scheduling as a target and taking the energy balance of the production process, the material production balance and the production requirement of key process variables as constraints, wherein the uncertainty of the performance of the key parameter control loop in which the key process variable dependent variable is located is described in an uncertain mode.
Figure GDA0002715142330000121
Figure GDA0002715142330000122
Figure GDA0002715142330000123
Figure GDA0002715142330000124
Figure GDA0002715142330000125
Wherein i 1,2, 12 product quality components, and j 1,2, 16 is a control loop，
Figure GDA0002715142330000126
For a positive deviation of the target i from the target value,
Figure GDA0002715142330000129
for a negative deviation of the target i from the target value,
Figure GDA0002715142330000127
as a predicted value of the ith product quality, biIs the predicted value, x, of the ith productjIs the set point for the jth loop,
Figure GDA0002715142330000128
random term variances in the time series data are extracted for the EMD.
And finally, modifying the grade division standard of the operating condition of the real-time optimized layer of the hydrocracking flow according to the uncertain optimization result, and carrying out online evaluation on the operating condition performance of the real-time optimized layer of the hydrocracking flow according to the modified grade standard of the operating condition of the real-time optimized layer of the hydrocracking flow. The real-time optimization layer operation condition evaluation standard dividing principle based on historical data is to divide according to the magnitude of comprehensive product quality (yield) forward deviation, firstly, the comprehensive product quality (yield) forward deviation is normalized, then, on the basis of removing unqualified conditions, the remaining interval is equally divided into 4 parts, and the 4 operation condition evaluation standard dividing intervals are respectively the comprehensive product quality (yield) forward deviation intervals corresponding to the 4 operation condition grades of 'excellent', 'good', 'medium' and 'poor' in sequence. However, after the optimization is uncertain through the optimal operation condition, the division of the forward deviation interval of the comprehensive product quality (yield) corresponding to the operation condition grade is adjusted. The classification rule at this time is changed to that the section from the non-determined optimization result (after standardization) to 1 is equally divided on the basis of removing the unqualified condition, so that the forward deviation section of the comprehensive product quality (yield) corresponding to the 4 determined operation condition grades of "good", "medium" and "poor" is different from the previous classification result based on the historical data, and further the evaluation result of the operation condition is different. For example, the real-time optimization layer operation condition evaluation result based on the historical data is "good", but it is found through uncertain optimization that the operation target corresponding to the "good" operation condition cannot be realized under the condition of the performance of the existing control system, so the real-time optimization layer operation condition ranking standard needs to be revised. By adjusting the target corresponding to the "excellent" operating condition to be the uncertain optimization result and correspondingly adjusting the other operating condition grade division standards, the operating condition that is originally evaluated as "good" may be updated to be evaluated as "excellent" after the operating condition evaluation standard is adjusted. The method comprises the steps of utilizing the running data of 17 years, 7 months and 31 days of a hydrocracking process of a certain factory in China, and carrying out online evaluation on the running condition of the hydrocracking process on the real-time optimization layer by combining the online evaluation of the running condition based on historical data and the online evaluation of the uncertain optimization running condition considering the feedback influence of the performance of a control system on the running condition of the real-time optimization layer, wherein the evaluation result is shown in figure 10. In the figure, the solid line is the online evaluation result of the running condition of the real-time optimization layer based on the historical data, and the dotted line is the online evaluation result of the running condition of the real-time optimization layer combining the historical data and the uncertain optimization, which are different. This is because, under the influence of the performance of the key parameter control loop, the operating condition at this time may not necessarily meet the standard required by the historical contemporaneous optimal operating condition, and the operating condition of the real-time optimization layer at this time can be truly reflected by the readjusted operating condition ranking principle after the optimal operating condition is uncertainly optimized. When the operation condition evaluation result is 'non-excellent', the 'non-excellent' reason tracing can be carried out through the hydrocracking process real-time optimization layer operation condition evaluation comprehensive product quality and yield index system shown in fig. 7, and the index with poor evaluation result in the evaluation index system is searched. And then, taking the tracing result as a reference to diagnose the non-optimal running condition and locate specific links and variables with performance degradation. And finally, self-adjusting the operation condition according to the diagnosis result to return to the optimal operation condition again.
And fourthly, analyzing the feedback influence of the real-time optimization layer operation performance on the operation performance of the planned scheduling layer, carrying out uncertain optimization on the optimal operation index of the planned scheduling layer, and carrying out online evaluation on the operation performance of the planned scheduling layer according to the index. Firstly, according to related data and real-time optimized layer operation performance of short-term planning and scheduling of a hydrocracking process, uncertain optimization is carried out on the operation condition of a planned scheduling layer of the hydrocracking process by taking the maximum comprehensive commodity rate and light oil yield and the minimum unit energy factor energy consumption as optimization targets and taking the energy balance in the production process, the material production balance, the raw material supply and the operation performance of a sub-process as constraints, wherein the uncertainty of the operation performance of the real-time optimized layer of the sub-process operation performance factor process is an uncertain variable and is described in an uncertain mode; and determining the uncertain optimization evaluation benchmark of the hydrocracking process plan scheduling layer according to the uncertain optimization result. And then establishing a prediction model of comprehensive commodity rate, light oil yield and unit energy factor energy consumption based on JIT-stacking. And finally, performing online evaluation on the operation performance of the hydrocracking process plan scheduling layer according to the comprehensive commodity rate, the light oil yield, the unit energy factor energy consumption prediction result and the uncertain optimization evaluation benchmark of the hydrocracking process plan scheduling layer.
And fifthly, when the overall operation condition evaluation result of the hydrocracking process is poor, tracing the reason of 'non-excellent' operation condition according to the evaluation index system of each layer of operation condition.
Firstly, performing 'non-optimal' reason tracing on the operation condition of the scheduling layer through the operation condition evaluation index system of the scheduling layer, and if the reason is found to be 'non-optimal' of the operation condition of the real-time optimization layer of the sub-process, continuing to perform 'non-optimal' reason tracing according to the operation condition evaluation index system of the real-time optimization layer of the sub-process. By querying the evaluation results of each index in the comprehensive product quality and yield index system in the hydrocracking process real-time optimization layer operation status evaluation shown in fig. 7, the product quality component or product yield with performance degradation can be specifically located. Then, the subsequent performance degradation diagnosis link can carry out performance degradation reason diagnosis by taking the product quality component or the product yield 'non-excellent' as the top event.
Example two:
by taking the residual oil hydrogenation process example of a certain refinery in China, the online evaluation method for the global operation condition of the continuous production process provided by the invention is explained in detail as follows:
the method comprises the steps of firstly, dividing a residual oil hydrogenation process into a planned scheduling layer, a real-time optimization layer and a process control layer from top to bottom, and determining operation indexes of each layer; and a feedback relation from bottom to top exists among the process control layer, the real-time optimization layer and the planning and scheduling layer. The operation optimization and control whole flow of the residual oil hydrogenation flow is shown in fig. 12 and mainly relates to three layers of a planning and scheduling layer, a real-time optimization layer and a process control layer. The scheduling layer determines the target value of the daily comprehensive production index of the residual oil hydrogenation process according to the target of the monthly or quarterly comprehensive production index of the residual oil hydrogenation process (light oil yield, unit energy factor energy consumption and the like), the sub-process operation performance (hydrofining process operation performance, high-low pressure separation process operation performance and fractionation system operation performance), the energy consumption (electricity, water, hydrogen, steam and the like), the raw material supply (supply quantity of feed oil and quality of feed oil) and other constraint conditions. The real-time optimization layer determines the target value of the operation index of each sub-process according to the target value of daily comprehensive production index of the residual oil hydrogenation process, by considering the capacity of the sub-process equipment (the handling capacity of a hydrofining process, the handling capacity of a high-low pressure analysis process and the handling capacity of a fractionation system), the operation performance of the sub-process equipment (the utilization efficiency of the hydrofining process equipment, the utilization efficiency of the high-low pressure separation process equipment and the utilization efficiency of the fractionation system equipment), the performance of the sub-process control system (the performance of a hydrofining process key parameter control loop, the performance of a high-low pressure separation process key parameter control loop and the performance of a fractionation system key parameter control loop), the activity of a catalyst (the activity of a hydrofining process reactor), the property of a raw material (the flow rate. The process control layer sets the set value of the key parameter control loop according to the target value of the operation index of the sub-process, so that the actual value of the operation index of the real-time optimization layer approaches to the target value of the operation index as much as possible.
And secondly, screening out a key parameter control loop of the residual oil hydrogenation process based on the process analysis and the mechanism analysis of the residual oil hydrogenation process, and evaluating the performance of the key parameter control loop. The flow chart of the residue hydrogenation process is shown in figure 13 and mainly comprises a hydrofining process, a high-low pressure separation process and a fractionation system 3. Through process mechanism analysis, the selected key parameter control loop is mainly distributed in a hydrorefining process and a fractionation system of a residual oil hydrogenation process.
Collecting operation data of a key parameter control loop of a residual oil hydrogenation process, and performing outlier rejection and wavelet denoising on the operation data. And then, based on an EMD method, acquiring a random stable part of the non-stable time sequence of the operation data of each key parameter control loop as a stable time sequence. And establishing an ARMA time sequence analysis model for each stable time sequence, and identifying model parameters by adopting a least square method. And estimating the time delay of the operation data of each key parameter control loop by adopting a cross-correlation function method. And finally, performing performance evaluation on the key parameter control loop based on the Harris performance index in the minimum variance reference. Performance is divided into four levels: a0.8, 1, B0.7, 0.8), C0.6, 0.7, D0, 0.6, using the output data collected from the residual oil hydrogenation process to evaluate the performance of the control system of the key parameter control loop.
And thirdly, online evaluation is carried out on the operation performance of the real-time optimization layer of the residual oil hydrogenation process respectively from two aspects of online evaluation of the operation state based on historical data and online evaluation of the uncertain optimization operation state considering the feedback influence of the key parameter control loop performance on the real-time optimization layer. Similar to the hydrocracking process, in the residue hydrogenation process of this embodiment, the real-time optimization layer operation target is set not to excessively improve the product quality and the light oil yield, and the production requirement of the scheduling layer is only completed on the premise of ensuring production, that is, the smaller the forward deviation from the product quality (yield) required by the scheduling layer is, the better. The formula of the mass forward deviation is equivalent to the formula (4) and the formula (5) in the first embodiment.
The historical data-based online evaluation of the operating conditions mainly comprises 4 parts of inlet working condition classification, light oil yield classification, multi-index comprehensive evaluation and key quality (yield) index prediction. The feed oil of the residual oil hydrogenation process mainly consists of vacuum wax oil, straight-run heavy wax oil, coker wax oil and catalytic cycle oil, and is shown in fig. 14. Similar to the first embodiment, the classification of the inlet conditions of the second embodiment also adopts the mahalanobis distance as the similarity measurement function, and classifies the inlet conditions of the residual oil hydrogenation process based on the K-means algorithm.
According to the scheduling requirement, the yield of the light oil in the residual oil hydrogenation process is classified, and the requirement on the yield of the light oil is not changed greatly.
Through mechanism analysis, an index system for the comprehensive product quality and yield of the residual oil hydrogenation process shown in FIG. 15 is established. The index system is divided into 4 layers, wherein the 4 th layer is the component of each product oil, the 3 rd layer is the product oil, the 2 nd layer is the product quality and yield, and the 1 st layer is the material production index of the comprehensive product quality and yield. And then dividing a real-time optimization layer operation condition evaluation standard based on historical data according to an inlet working condition classification, a light oil yield classification, a comprehensive product quality and yield index system, and dividing the operation condition of the residual oil hydrogenation process real-time optimization layer into 5 grades of 'good', 'medium', 'poor' and 'unqualified', with the aim that the forward deviation of the product quality and the yield required by the planned scheduling is smaller and better.
And establishing a soft measurement model of comprehensive product quality and yield of a residual oil hydrogenation process based on JIT-stacking. First, by analyzing the Pearson correlation of process variables with product quality and yield and the process mechanism of the residual oil hydrogenation process, key process variables were selected as shown in table 4. And then establishing a JIT-stacking prediction model for integrating the product quality and yield through the selected key process variables.
TABLE 4 Key Process variables for residuum hydrogenation scheme
Figure GDA0002715142330000161
And finally, carrying out online evaluation on the operation condition of the real-time optimization layer of the residual oil hydrogenation process according to the JIT-stacking model prediction result of the comprehensive product quality and yield and the operation condition grade division standard of the real-time optimization layer of the residual oil hydrogenation process. The operation data of the residue oil hydrogenation process of a certain factory in China 8 months before 2018 is utilized, the real-time optimized layer operation condition is evaluated on line from the perspective of historical data, and the evaluation result is shown in figure 16.
According to the performance evaluation result of a key parameter control loop of a residual oil hydrogenation process, the following uncertain optimization is carried out on the running condition of a real-time optimization layer by taking the minimum forward deviation of the product quality and the light oil yield required by a planned scheduling as a target and taking the energy balance of a production process, the material production balance and the production requirement of key process variables as constraints, wherein the uncertainty of the performance of the key parameter control loop of the key process variable dependent variable is an uncertain variable and is described in an uncertain mode.
Figure GDA0002715142330000162
Figure GDA0002715142330000163
Figure GDA0002715142330000164
Figure GDA0002715142330000165
Figure GDA0002715142330000166
Wherein i is 1,2,3 is the product quality component, j is 1,2,.
Figure GDA0002715142330000167
For a positive deviation of the target i from the target value,
Figure GDA0002715142330000168
for a negative deviation of the target i from the target value,
Figure GDA0002715142330000169
as a predicted value of the ith product quality, biIs the predicted value, x, of the ith productjIs the set point for the jth loop,
Figure GDA0002715142330000171
random term variances in the time series data are extracted for the EMD.
And finally, modifying the grade standard of the running condition of the real-time optimization layer of the residual oil hydrogenation process according to the uncertain optimization result, and carrying out online evaluation on the running condition performance of the real-time optimization layer of the residual oil hydrogenation process according to the modified grade standard of the running condition of the real-time optimization layer of the residual oil hydrogenation process. The real-time optimization layer operation condition evaluation standard dividing principle based on historical data is to divide according to the magnitude of comprehensive product quality (yield) forward deviation, firstly, the comprehensive product quality (yield) forward deviation is normalized, then, on the basis of removing unqualified conditions, the remaining interval is equally divided into 4 parts, and the 4 operation condition evaluation standard dividing intervals are respectively the comprehensive product quality (yield) forward deviation intervals corresponding to the 4 operation condition grades of 'excellent', 'good', 'medium' and 'poor' in sequence. However, after the optimization is uncertain through the optimal operation condition, the division of the forward deviation interval of the comprehensive product quality (yield) corresponding to the operation condition grade is adjusted. The classification rule at this time is changed to that the section from the non-determined optimization result (after standardization) to 1 is equally divided on the basis of removing the unqualified condition, so that the forward deviation section of the comprehensive product quality (yield) corresponding to the 4 determined operation condition grades of "good", "medium" and "poor" is different from the previous classification result based on the historical data, and further the evaluation result of the operation condition is different. For example, the real-time optimization layer operation condition evaluation result based on the historical data is "good", but it is found through uncertain optimization that the operation target corresponding to the "good" operation condition cannot be realized under the condition of the performance of the existing control system, so the real-time optimization layer operation condition ranking standard needs to be revised. By adjusting the target corresponding to the "excellent" operating condition to be the uncertain optimization result and correspondingly adjusting the other operating condition grade division standards, the operating condition that is originally evaluated as "good" may be updated to be evaluated as "excellent" after the operating condition evaluation standard is adjusted.
And fourthly, analyzing the feedback influence of the real-time optimized layer operation performance on the operation performance of the planned scheduling layer, carrying out uncertain optimization on the optimal operation index of the planned scheduling layer, and carrying out online evaluation on the operation performance of the planned scheduling layer of the residual oil hydrogenation process according to the index. Firstly, according to related data and real-time optimized layer operation performance of short-term planning and scheduling of a residual oil hydrogenation process, uncertain optimization is carried out on the operation condition of a planned scheduling layer of the residual oil hydrogenation process by taking maximum light oil yield and minimum unit energy factor energy consumption as optimization targets and taking energy balance in a production process, material production balance, raw material supply and sub-process operation performance as constraints, wherein uncertainty of the operation performance of the sub-process operation performance factor process real-time optimized layer is an uncertain variable and is described in an uncertain mode; and determining uncertain optimization evaluation criteria of a scheduling layer of a residual oil hydrogenation process plan according to uncertain optimization results. Then, a prediction model of the light oil yield and the energy consumption per energy factor based on JIT-keeping is established. And finally, performing online evaluation on the operation performance of the residual oil hydrogenation process planning and scheduling layer according to the light oil yield, the unit energy factor energy consumption prediction result and the uncertain optimization evaluation standard of the residual oil hydrogenation process planning and scheduling layer.
And fifthly, when the overall operation condition evaluation result of the residual oil hydrogenation process is poor, tracing the reason of 'non-excellent' operation condition according to the evaluation index system of the operation condition of each layer. Firstly, performing 'non-optimal' reason tracing on the operation condition of the scheduling layer through the operation condition evaluation index system of the scheduling layer, and if the reason is found to be 'non-optimal' of the operation condition of the real-time optimization layer of the sub-process, continuing to perform 'non-optimal' reason tracing according to the operation condition evaluation index system of the real-time optimization layer of the sub-process. By inquiring the evaluation results of each index in the comprehensive product quality and yield index system in the residue oil hydrogenation process real-time optimization layer operation condition evaluation shown in fig. 15, the product quality component or product yield with performance degradation can be specifically located. Then, the subsequent performance degradation diagnosis link can carry out performance degradation reason diagnosis by taking the product quality component or the product yield 'non-excellent' as the top event.
Finally, the method of the present application is only a preferred embodiment and is not intended to limit the scope of the present invention. Any modification, equivalent replacement, or improvement made within the spirit and principle of the present invention should be included in the protection scope of the present invention.

Claims (7)

1. A method for online evaluation of global operation conditions in a continuous production process, comprising the steps of:
s1, dividing a continuous production process into a planned scheduling layer, a real-time optimization layer and a process control layer from top to bottom, and determining operation indexes of each layer; a feedback relation from bottom to top exists among the process control layer, the real-time optimization layer and the planning and scheduling layer;
s2, screening out a process key parameter control loop based on production process analysis and mechanism analysis, and evaluating the performance of the key parameter control loop:
s21, screening out a process key parameter control loop based on production process analysis and mechanism analysis;
s22, collecting key parameter control loop operation data, and performing outlier rejection and wavelet denoising on the operation data;
s23, acquiring a random stable part of the non-stable time sequence of the operation data of the key parameter control loop based on an EMD method, and taking the random stable part as a stable time sequence;
s24, establishing a time sequence analysis model for the stable time sequence, and identifying model parameters;
s25, estimating the time delay of the operation data of the key parameter control loop;
s26, based on the time sequence analysis model and the time delay, performing control system performance evaluation on a key parameter control loop by adopting a minimum variance reference;
s3, respectively carrying out online evaluation on the operation performance of the real-time optimization layer from two aspects of online evaluation of the operation state based on historical data and online evaluation of the uncertain optimization operation state considering the feedback influence of the key parameter control loop performance on the real-time optimization layer:
s31, collecting continuous production process operation data, and performing outlier rejection and wavelet denoising on the continuous production process operation data; the continuous production process operation data comprise raw material detection data, process variable data and product quality test data;
s32, classifying inlet working conditions in the production process by using a K-means clustering method based on the raw material detection data to obtain inlet working condition classification results;
s33, classifying the product yield in the continuous production process based on the plan scheduling requirement to obtain a product yield classification result;
s34, constructing a comprehensive product quality and yield index system based on multiple products by utilizing a multi-target comprehensive evaluation method based on the production characteristics of the multiple products in the continuous production process, and taking the comprehensive product quality and yield index system as a real-time optimized layer online evaluation index to obtain a comprehensive product quality and yield index;
s35, establishing a real-time optimization layer operation condition evaluation standard based on historical data in the continuous production process based on the inlet working condition classification result, the product yield classification result, the product quality assay data and the comprehensive product quality and yield indexes, and dividing the real-time optimization layer operation performance into 5 grades of 'excellent', 'good', 'medium', 'poor' and 'unqualified';
s36, based on the process variable data, selecting key process variable data according to Pearson correlation analysis results of process variables and product quality and yield and process mechanism analysis;
s37, establishing a prediction model of comprehensive product quality and yield indexes by using a JIT-stacking method based on the key process variable data;
s38, carrying out online evaluation on the running condition of the real-time optimization layer from the aspect of online evaluation of the running condition based on historical data based on the comprehensive product quality and yield index prediction result and the running condition evaluation standard of the real-time optimization layer in the continuous production process;
s39, based on the key parameter control loop performance evaluation result, performing uncertain optimization on the running condition of the real-time optimization layer, and determining the uncertain optimization result as a real-time optimization layer uncertain optimization evaluation benchmark in the continuous production process, wherein an uncertain optimization model specifically comprises the following steps: the method comprises the following steps of taking the minimum forward deviation of product quality and yield required by planned scheduling as an optimization target, and taking production process energy balance, material balance and key process variable production requirements as constraints;
s310, synthesizing the operating condition evaluation standard of the continuous production process real-time optimization layer based on the historical data and the uncertain optimization evaluation standard of the production process real-time optimization layer, determining the operating condition evaluation standard of the continuous production process real-time optimization layer, specifically revising the operating condition evaluation standard of the real-time optimization layer according to the uncertain optimization operation condition evaluation result, and performing online evaluation on the operating condition of the real-time optimization layer according to the revised evaluation standard;
s4, analyzing the feedback influence of the real-time optimization layer operation performance on the operation performance of the planned scheduling layer, carrying out uncertain optimization on the optimal operation index of the planned scheduling layer, and carrying out online evaluation on the operation performance of the planned scheduling layer according to the index:
s41, collecting related data of short-term plan scheduling of the continuous production process, and when sufficient raw material reserve is available, considering that the short-term plan scheduling of the continuous production process is unrelated to the reserve of the raw materials;
s42, performing uncertain optimization on the operation condition of a planned scheduling layer based on the short-term planned scheduling related data of the continuous production process and the operation performance of the real-time optimization layer, and determining an uncertain optimization result as an uncertain optimization evaluation benchmark of the planned scheduling layer of the continuous production process;
the uncertain optimization model specifically comprises the following steps: the method takes the maximum comprehensive commodity yield and the minimum unit energy factor energy consumption as optimization targets and takes the energy balance, material balance, raw material supply and sub-process operation performance in the production process as constraints;
s43, based on the continuous production process operation data of S31, selecting key process variable data according to a Pearson correlation analysis result and process mechanism analysis of process variables, comprehensive commodity yield and unit energy factor energy consumption;
s44, establishing a prediction model for synthesizing commodity yield and unit energy factor energy consumption by using a JIT-stacking method based on the key process variable data;
and S45, carrying out online evaluation on the operation performance of the planning and scheduling layer based on the comprehensive commodity yield and unit energy factor energy consumption prediction result and the uncertain optimization evaluation criterion of the planning and scheduling layer in the continuous production process.
2. The method for on-line evaluation of global operation status of continuous production process according to claim 1, wherein the operation indexes of each layer described in S1 are determined by:
the planning and scheduling layer determines a target value of daily comprehensive production indexes according to the target of monthly or quarterly comprehensive production indexes, the sub-process operation performance, the energy consumption and the raw material supply in the continuous production process;
the real-time optimization layer determines the target value of the operation index of each sub-process according to the target value of the daily comprehensive production index of the whole production process and simultaneously considering the capacity of sub-process equipment, the utilization efficiency of the sub-process equipment, the performance of a key parameter control loop, the property of raw materials and the process allowable range of the operation index;
the process control layer sets the set value of the key control system according to the target value of the operation index of the sub-process, so that the actual value of the operation index of the real-time optimization layer approaches to the target value of the operation index.
3. The method for online global operation status evaluation of continuous production process as claimed in claim 1, wherein said data related to short-term planned scheduling of continuous production process in S41 comprises: raw material reserve, long-term comprehensive production index, intermediate product or waste product requirement and energy consumption requirement.
4. The method for online global operational status assessment of continuous production process according to claim 1, wherein said S4 is followed by further comprising:
and S5, when the overall operation condition evaluation result is not good, tracing the reason of the non-good operation condition according to the evaluation index system of the operation condition of each layer.
5. The method for the on-line assessment of the global operating conditions of the continuous production process as claimed in claim 1, wherein the uncertainty of the performance of the key control loop where the dependent variable of the key process variable is located in S39 is described in an uncertain manner as an uncertain variable.
6. The method for the on-line assessment of the global operation status of the continuous production process as claimed in claim 1, wherein the uncertainty of the layer operation performance is optimized in real time by the sub-process operation performance factor process described in S42, and is an uncertain variable, which is described in an uncertain manner.
7. The method for on-line assessment of global operating conditions of continuous production process according to claim 1, wherein said continuous production process comprises complex industrial processes including hydrocracking process and residue hydrogenation process.
CN201910880712.XA 2019-03-25 2019-09-18 Method suitable for online evaluation of global operation state in continuous production process Active CN110456756B (en)

Applications Claiming Priority (2)

Application Number Priority Date Filing Date Title
CN201910226363 2019-03-25
CN201910226363X 2019-03-25

Publications (2)

Publication Number Publication Date
CN110456756A CN110456756A (en) 2019-11-15
CN110456756B true CN110456756B (en) 2020-12-08

Family

ID=68492285

Family Applications (1)

Application Number Title Priority Date Filing Date
CN201910880712.XA Active CN110456756B (en) 2019-03-25 2019-09-18 Method suitable for online evaluation of global operation state in continuous production process

Country Status (1)

Country Link
CN (1) CN110456756B (en)

Families Citing this family (8)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN111428345B (en) * 2020-02-27 2022-07-05 福建华电可门发电有限公司 Performance evaluation system and method of random load disturbance control system
CN111766785B (en) * 2020-07-10 2021-07-13 北京理工大学 Multi-machine scheduling method for minimizing expected early and late expenses
CN111983997B (en) * 2020-08-31 2021-07-20 北京清大华亿科技有限公司 Coupling analysis-based control loop performance monitoring method and system
CN112396344A (en) * 2020-11-30 2021-02-23 天津大学 Chemical process reliability online evaluation method based on product quality
CN115512530B (en) * 2021-06-23 2023-12-22 中国石油化工股份有限公司 Refining device operation condition early warning method, early warning device and early warning system
CN113268936B (en) * 2021-07-03 2022-07-19 石河子大学 Key quality characteristic identification method based on multi-objective evolution random forest characteristic selection
CN115903727A (en) * 2022-10-10 2023-04-04 乌海宝化万辰煤化工有限责任公司 DCS control system-based PID control loop performance evaluation system
CN116880427B (en) * 2023-09-06 2023-11-14 中南大学 Intelligent control method and system based on feeding condition estimation and working condition analysis

Family Cites Families (8)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
TWI315054B (en) * 2006-05-10 2009-09-21 Nat Cheng Kung Universit Method for evaluating reliance level of a virtual metrology system in product manufacturing
JP2009290150A (en) * 2008-06-02 2009-12-10 Renesas Technology Corp System and method for manufacturing semiconductor device
CN102054125B (en) * 2010-11-16 2013-01-23 浙江大学 Method for stabilizing chemical constituents of charging agglomerate
CN104407589B (en) * 2014-11-26 2017-01-25 西北工业大学 Workshop manufacturing process-oriented active sensing and anomaly analysis method of real-time production performance
CN104932488B (en) * 2015-06-30 2017-06-16 南京工业大学 A kind of Model Predictive Control Performance Evaluation and diagnostic method
CN106709654A (en) * 2016-12-28 2017-05-24 中南大学 Global operating condition evaluating and quality tracing method for hydrocracking process
CN108490782B (en) * 2018-04-08 2019-04-09 中南大学 A kind of method and system being suitable for the missing data completion of complex industrial process product quality indicator based on selective double layer integrated study
CN109063412B (en) * 2018-08-27 2020-08-11 浙江大学 Multi-source data fusion system and method for plasma pyrolysis coal-to-acetylene reaction state evaluation

Also Published As

Publication number Publication date
CN110456756A (en) 2019-11-15

Similar Documents

Publication Publication Date Title
CN110456756B (en) Method suitable for online evaluation of global operation state in continuous production process
Akbarian-Saravi et al. Development of a comprehensive decision support tool for strategic and tactical planning of a sustainable bioethanol supply chain: Real case study, discussions and policy implications
Velázquez et al. Development of an energy management system for a naphtha reforming plant: A data mining approach
US20210348066A1 (en) Predictive control systems and methods with hydrocracker conversion optimization
Goyal et al. Optimization of condition-based maintenance using soft computing
Aziz et al. A study on gradient boosting algorithms for development of AI monitoring and prediction systems
CN104484714B (en) A kind of real-time predicting method of catalytic reforming unit yield
Makepa et al. A systematic review of the techno-economic assessment and biomass supply chain uncertainties of biofuels production from fast pyrolysis of lignocellulosic biomass
US20220299952A1 (en) Control system with optimization of neural network predictor
CN104765347A (en) Yield real-time prediction method in residual oil delayed coking process
CN108446358B (en) Optimization method based on MIV and association rule and data modeling method of petrochemical equipment
CN114239321A (en) Oil refining process mode identification and optimization method based on big data
Niño-Adan et al. Soft-sensor design for vacuum distillation bottom product penetration classification
CN111475957A (en) Oil refining process production plan optimization method based on device mechanism
CN101727609B (en) Pyrolyzate yield forecasting method based on support vector machine
CN110021377B (en) Method and device for predicting deactivation of hydrocracking catalyst and storage equipment
Khaldi et al. Artificial intelligence perspectives: A systematic literature review on modeling, control, and optimization of fluid catalytic cracking
Farshchian et al. Stock market prediction with hidden markov model
Anderson et al. A Bayesian hierarchical assessment of night shift working for offshore wind farms
Al-Jlibawi et al. The efficiency of soft sensors modelling in advanced control systems in oil refinery through the application of hybrid intelligent data mining techniques
Sadighi et al. Modeling and optimizing a vacuum gas oil hydrocracking plant using an artificial neural network
Najari et al. Improving maintenance strategy by physical asset management considering the use of MFOP instead of MTBF in petrochemical
Abdulghafour et al. Developing of reliability-centered maintenance methodology in second power plant of south baghdad
Levene et al. How advanced analytics can benefit infrastructure capital planning
EP4303672A1 (en) Evaluation model generating apparatus, evaluation model generating method, and evaluation model generating program

Legal Events

Date Code Title Description
PB01 Publication
PB01 Publication
SE01 Entry into force of request for substantive examination
SE01 Entry into force of request for substantive examination
GR01 Patent grant
GR01 Patent grant