An event-based automatic optimization method for production yield of fine naphthalene

Abstract

An event-based automatic optimization method for refined naphthalene production yield includes the following steps: S1: Establishing a simulation process driven by the sequence of process operations; S2: Establishing a path search based on the simulation process progression, and determining the target path based on the search results; S3: Establishing a refined naphthalene production program based on the determined target path. This invention's event-based automatic optimization method for refined naphthalene production yield first provides a searchable basis for path search by setting up a simulation process driven by trigger events; secondly, based on this, algorithm-based optimization is performed using an exhaustive search method combined with pruning, and refined naphthalene crystallization production is established based on the target path provided by the optimization results, thus completing the automatic optimization of refined naphthalene crystallization production.

Description

Technical Field

[0001] This invention belongs to the field of refined naphthalene production, specifically relating to an event-based automatic optimization method for refined naphthalene production output. Background Technology

[0002] The production process of refined naphthalene is a relatively complex programmed process with seven steps: Concentration II (C2), Concentration I (C1), Purification I (P1), Purification II (P2), Purification III (P3), Purification IV (P4), and Purification V (P5). Each crystallization step includes feeding, crystallization, discharge, sweating, and melting. The concentration process involves two separate crystallization processes, each with its own distinct characteristics. There are three crystallization tanks on site, and each process takes place within a crystallization tank. Only one process can be run at a time, and the processes that can be performed within each crystallization tank are limited.

[0003] RE-7101A / B box (B1) operation: purification step one, concentration step one, concentration step two.

[0004] RE-7102A / B box (B2) operation: purification steps two to five.

[0005] RE-7103A / B box (B3) operation: purification step one to step two.

[0006] Each intermediate product is collected in a corresponding intermediate tank. All processes—feeding, effluent, sweating, and discharge—involve entering designated tanks. The logistics diagram is shown below. Figure 1 As shown.

[0007] Crystallization time is mainly affected by the quality of raw materials, the number of processes required, and temperature. Figure 2 A schematic diagram showing the relationship between the time and temperature required for each process in step C2 at a certain time is given. When the ambient temperature and the quality of the raw materials change, the curve will change accordingly.

[0008] Because the crystallization time varies, the material level in the intermediate tank changes frequently. It is necessary to rationally select which process to run in the crystallization box to ensure that the material in the intermediate tank meets production requirements and that the crystallization process is not interrupted due to the influence of the tank position.

[0009] Because the crystallization parameters are inconsistent, the crystallization time varies at each step. This further increases the complexity of scheduling the crystallization process.

[0010] Currently, in the existing naphthalene production process, the selection of the crystallization step relies on the operator's experience and judgment. This method involves a large amount of manual calculation and may not necessarily select the most efficient crystallization scheme. Therefore, how to organize and coordinate the crystallization tank and intermediate tank during the production process, and optimize the scheduling strategy to maximize their efficiency, has become a problem that needs to be solved.

[0011] The invention application with application number 202011341455.1 discloses "a method, device, equipment, and medium for advancing simulation time of a continuous-discrete hybrid system". The method includes: for advancing the simulation time of a continuous system, the entire system advances the simulation time according to a minimum step size, wherein the time advancement step size of the simulation object is an integer multiple of the minimum step size; for discrete events, the events are divided into synchronous events and asynchronous events. Synchronous events are executed immediately after they are generated, and asynchronous events are sorted in the event queue according to the timestamp size. When the simulation time advances to the timestamp, the asynchronous events are scheduled and executed. The state of the simulation object changes as the simulation time advances. This method combines the advantages of time stepping and event advancement, broadens the application scope, and improves the efficiency of simulation advancement and scheduling.

[0012] Invention application No. 202011514108.4 discloses a multi-task timing conflict detection method based on event flow analysis, including the following steps: 1: Constructing a set of event time slices for the completion time of each event processing; 2: Constructing a set of task execution times; 3: Determining the relationship between each event and any task in the software architecture to form an event-task relationship set; 4: Sort the tasks called in the time periods where the event time slices in the event-task relationship set overlap according to their order to form a task time slice sequence set; 5: Calculating the task time occupied by each event processing in the task time slice sequence set, determining whether a timeout occurs; determining whether a conflict exists, and outputting the conflicting events and tasks.

[0013] Application No. 202011536684.9 discloses a product performance prediction method and system. The method includes: acquiring product training data; obtaining the radial basis function of each neuron node in the hidden layer of a radial basis neural network based on the product training data; selecting one product to be trained as the target training product from multiple products to be trained, and inputting the technical parameter values ​​of the target training product into the radial basis function of each neuron node to obtain the branch performance prediction probability of each neuron node corresponding to the target training product; optimizing the initial weight set using a particle swarm optimization algorithm to obtain an initial optimal weight set; using the initial optimal weight set as the initial value of the weights of the radial basis neural network, setting different learning rates at different iteration numbers, calculating the loss value using the cross-entropy loss function, and adjusting the weights of the radial basis neural network using a gradient descent algorithm to obtain the final weight set. Summary of the Invention

[0014] To address the above problems, this invention proposes an automated and optimized method for the production of refined naphthalene crystallization, the specific technical solution of which is as follows:

[0015] An event-based method for automatically optimizing the production yield of refined naphthalene, characterized by the following steps:

[0016] S1: Establish an event-driven simulation process.

[0017] S2: Establish a path search based on the simulation progress, and determine the target path based on the search results.

[0018] S3: Establish a naphthalene production procedure based on the determined target path.

[0019] According to the present invention, an event-based method for automatic optimization of refined naphthalene production yield is characterized in that:

[0020] Step S1 is based on static parameters.

[0021] According to the present invention, an event-based method for automatic optimization of refined naphthalene production yield is characterized in that:

[0022] Step S1 is as follows:

[0023] In the production of refined naphthalene crystallization, the end of each sub-process or the start of the next sub-process is set as the trigger event.

[0024] The simulation process is advanced based on the set trigger events.

[0025] According to the present invention, an event-based method for automatic optimization of refined naphthalene production yield is characterized in that:

[0026] The process of advancing the simulation based on the set trigger events is specifically as follows:

[0027] Based on the expected completion time of each sub-process of each process, bottleneck points are established, thus forming a bottleneck point sequence for each process.

[0028] Establish a blocking queue based on the sequence of blocking points for each process and the execution order between processes;

[0029] The simulation process is advanced based on the sequence of bottlenecks in each process.

[0030] According to the present invention, an event-based method for automatic optimization of refined naphthalene production yield is characterized in that:

[0031] When constructing the blocking queue, if the next process is a subsequent process of the previous process, a material replenishment blocking point is constructed; otherwise, blocking points are constructed in the order of material discharge followed by material feed.

[0032] According to the present invention, an event-based method for automatic optimization of refined naphthalene production yield is characterized in that:

[0033] When the trigger event occurs, the next sub-process is determined according to the following steps:

[0034] SS1: Determine whether the next process is the first process set for the corresponding crystallizer. If it is the first process, proceed to step SS5; otherwise, proceed to step SS2.

[0035] SS2: If the next process is a subsequent process of the previous process, proceed to step SS5; otherwise, proceed to step SS3.

[0036] SS3: Determine whether the liquid level margin of the buffer tank corresponding to the next sub-process meets the process requirements. If it does not meet the requirements, proceed to step SS6; otherwise, proceed to step SS4.

[0037] SS4: Calculate the material discharge time;

[0038] SS5: Determine whether the liquid level margin of the buffer tank corresponding to the next sub-process meets the process requirements. If it does not meet the requirements, proceed to step SS6; otherwise, proceed to step SS7.

[0039] SS6: Delay the next sub-process and calculate the delay time;

[0040] SS7: Determine whether the crystallizer corresponding to the sub-process is operating at full load. If it is not operating at full load, calculate the delay time based on the loading rate.

[0041] According to the present invention, an event-based method for automatic optimization of refined naphthalene production yield is characterized in that:

[0042] The path search in step S2 is based on a tree structure.

[0043] According to the present invention, an event-based method for automatic optimization of refined naphthalene production yield is characterized in that:

[0044] Each process is defined as a node in a tree, and a process sequence is formed from the root to the leaf. A scheduling sequence path is formed based on the process sequence.

[0045] The scheduling sequence path search is performed using an exhaustive search method combined with a pruning algorithm.

[0046] According to the present invention, an event-based method for automatic optimization of refined naphthalene production yield is characterized in that:

[0047] The process, which combines exhaustive search with pruning algorithms, specifically includes the following steps:

[0048] S11: Calculate the total execution time of the scheduling sequence path in real time, and compare the calculation result with the set time length. If the calculation result is less than the set time length, proceed to step S12; otherwise, proceed to step S13.

[0049] S12: Calculate the number of scheduling sequence paths in real time and compare the calculation result with the set threshold; if the calculation result is less than the set threshold, continue the search by exhaustive search method; otherwise, proceed to step S14.

[0050] S13: Sort the scheduling sequence paths from high to low according to the set scoring rules, and select the scheduling sequence path with the highest score as the target path.

[0051] S14: Sort the scheduling sequence paths from high to low according to the set scoring rules, count the number of scheduling sequence paths according to the sorting results, and delete those exceeding the set threshold according to the set threshold.

[0052] According to the present invention, an event-based method for automatic optimization of refined naphthalene production yield is characterized in that:

[0053] The score is determined based on the total delay time of each scheduling sequence path, specifically set as the inverse of the total delay time;

[0054] The total delay time is determined as follows:

[0055] t total-dela (C(B i ,j))=t delay +t discharg +t non-full ,

[0056] in,

[0057] t delay Delay time, in seconds (s);

[0058] t discharge Discharge time, unit: seconds;

[0059] t non-full : Delay time calculated based on the filling rate for crystallization boxes that are not allowed to operate at full load, under the condition of not operating at full load, in seconds.

[0060] According to the present invention, an event-based method for automatic optimization of refined naphthalene production yield is characterized in that:

[0061] t non-full = (1-f)×t p ,

[0062] in,

[0063] f: Load rate;

[0064] t P : The time required for this operation step, in seconds.

[0065] According to the present invention, an event-based method for automatic optimization of refined naphthalene production yield is characterized in that:

[0066] Before continuing the search using the exhaustive method as described in step S12, first delete infeasible paths that do not meet the process requirements.

[0067] According to the present invention, an event-based method for automatic optimization of refined naphthalene production yield is characterized in that:

[0068] Before sorting the scheduling sequence paths from high to low according to the set scoring rules as described in step S14, infeasible paths that do not meet the process requirements are deleted first.

[0069] According to the present invention, an event-based method for automatic optimization of refined naphthalene production yield is characterized in that:

[0070] The static parameters are determined based on the empirical average values ​​of historical process parameters.

[0071] According to the present invention, an event-based method for automatic optimization of refined naphthalene production yield is characterized in that:

[0072] The phrase "according to process requirements" specifically refers to: according to the requirements of the set scheduling rule library.

[0073] The present invention discloses an automatic optimization method for naphthalene crystallization production. First, a simulation process driven by trigger events is set up to provide a searchable basis for path search. Second, based on this, an algorithm-based optimization is performed using an exhaustive search method combined with a pruning method. The naphthalene crystallization production is established according to the target path provided by the optimization results, thus completing the automatic optimization of naphthalene crystallization production. Attached Figure Description

[0074] Figure 1 This is a flowchart illustrating the process steps for producing refined naphthalene crystallization in this invention.

[0075] Figure 2 for Figure 1 A schematic diagram showing the time progression and temperature relationship of each process in the C2 step;

[0076] Figure 3 This is a schematic diagram illustrating the definition of objects and state space in this invention;

[0077] Figure 4This is a schematic diagram of the event rules, material flow, and flow rate in the B1 box C2 process in this embodiment of the invention;

[0078] Figure 5 This is a schematic diagram of the event rules, material flow, and flow rate in the B1 box C1 process of this invention embodiment;

[0079] Figure 6 This is a schematic diagram of the event rules, material flow, and flow rate in the B1 box P1 process of this invention embodiment;

[0080] Figure 7 This is a schematic diagram of the event rules, material flow, and flow rate in the B2 box P2 process in this embodiment of the invention;

[0081] Figure 8 This is a schematic diagram of the event rules, material flow and flow rate in the B2 box P3 process in this embodiment of the invention;

[0082] Figure 9 This is a schematic diagram of the event rules, material flow, and flow rate in the B2 box P4 process in this embodiment of the invention;

[0083] Figure 10 This is a schematic diagram of the event rules, material flow and flow rate in the B2 box P5 process in this embodiment of the invention;

[0084] Figure 11 This is a schematic diagram of the event rules, material flow, and flow rate in the B3 box P1 process in this embodiment of the invention.

[0085] Figure 12 This is a schematic diagram of the event rules, material flow, and flow rate in the P2 process of box B3 in this embodiment of the invention.

[0086] Figure 13 This is a schematic diagram of the multi-task parallel state update and blocking sequence construction in this invention;

[0087] Figure 14 This is a schematic diagram of the steps of the present invention;

[0088] Figure 15 This is a schematic diagram illustrating the steps of determining the next sub-process when a triggering event occurs in this invention.

[0089] Figure 16 This is a schematic diagram illustrating the steps of the exhaustive search method combined with the pruning algorithm in this invention;

[0090] Figure 17 This is a schematic diagram illustrating the exhaustive search and pruning according to preset rules in an embodiment of the present invention;

[0091] Figure 18 This is a schematic diagram illustrating pruning based on evaluation indicators in an embodiment of the present invention. Detailed Implementation

[0092] The following is a detailed description of an automated method for producing refined naphthalene crystals according to the present invention, based on the accompanying drawings and specific embodiments.

[0093] An event-based method for automatically optimizing the production yield of refined naphthalene, such as Figure 14 As shown, it includes the following steps:

[0094] S1: Establish an event-driven simulation process.

[0095] S2: Establish a path search based on the simulation progress, and determine the target path based on the search results.

[0096] S3: Establish a naphthalene production procedure based on the determined target path.

[0097] in,

[0098] Step S1 is based on static parameters.

[0099] in,

[0100] Step S1 is as follows:

[0101] In the production of refined naphthalene crystallization, the end of each sub-process or the start of the next sub-process is set as the trigger event.

[0102] The simulation process is advanced based on the set trigger events.

[0103] in,

[0104] The process of advancing the simulation based on the set trigger events is specifically as follows:

[0105] Based on the expected completion time of each sub-process of each process, bottleneck points are established, thus forming a bottleneck point sequence for each process.

[0106] Establish a blocking queue based on the sequence of blocking points for each process and the execution order between processes;

[0107] The simulation process is advanced based on the sequence of bottlenecks in each process.

[0108] in,

[0109] When constructing the blocking queue, if the next process is a subsequent process of the previous process, a material replenishment blocking point is constructed; otherwise, blocking points are constructed in the order of material discharge followed by material feed.

[0110] in,

[0111] When the triggering event occurs, such as Figure 15 As shown, the next sub-process is determined according to the following steps:

[0112] SS1: Determine whether the next process is the first process set for the corresponding crystallizer. If it is the first process, proceed to step SS5; otherwise, proceed to step SS2.

[0113] SS2: If the next process is a subsequent process of the previous process, proceed to step SS5; otherwise, proceed to step SS3.

[0114] SS3: Determine whether the liquid level margin of the buffer tank corresponding to the next sub-process meets the process requirements. If it does not meet the requirements, proceed to step SS6; otherwise, proceed to step SS4.

[0115] SS4: Calculate the material discharge time;

[0116] SS5: Determine whether the liquid level margin of the buffer tank corresponding to the next sub-process meets the process requirements. If it does not meet the requirements, proceed to step SS6; otherwise, proceed to step SS7.

[0117] SS6: Delay the next sub-process and calculate the delay time;

[0118] SS7: Determine whether the crystallizer corresponding to the sub-process is operating at full load. If it is not operating at full load, calculate the delay time based on the loading rate.

[0119] in,

[0120] The path search in step S2 is based on a tree structure.

[0121] in,

[0122] Each process is defined as a node in a tree, and a process sequence is formed from the root to the leaf. A scheduling sequence path is formed based on the process sequence.

[0123] The scheduling sequence path search is performed using an exhaustive search method combined with a pruning algorithm.

[0124] in,

[0125] like Figure 16 As shown, the method of exhaustive search combined with pruning algorithm includes the following steps:

[0126] S11: Calculate the total execution time of the scheduling sequence path in real time, and compare the calculation result with the set time length. If the calculation result is less than the set time length, proceed to step S12; otherwise, proceed to step S13.

[0127] S12: Calculate the number of scheduling sequence paths in real time and compare the calculation result with the set threshold; if the calculation result is less than the set threshold, continue the search by exhaustive search method; otherwise, proceed to step S14.

[0128] S13: Sort the scheduling sequence paths from high to low according to the set scoring rules, and select the scheduling sequence path with the highest score as the target path.

[0129] S14: Sort the scheduling sequence paths from high to low according to the set scoring rules, count the number of scheduling sequence paths according to the sorting results, and delete those exceeding the set threshold according to the set threshold.

[0130] in,

[0131] The score is determined based on the total delay time of each scheduling sequence path, specifically set as the inverse of the total delay time;

[0132] The total delay time is determined as follows:

[0133] t total-delay (C(B i ,j))=t delay +t discharge +t non-ful ,

[0134] in,

[0135] t delay Delay time, in seconds (s);

[0136] t discharge Discharge time, unit: seconds;

[0137] t non-full : Delay time calculated based on the filling rate for crystallization boxes that are not allowed to operate at full load, under the condition of not operating at full load, in seconds.

[0138] in,

[0139] t non-full = (1-f)×t p ,

[0140] in,

[0141] f: Load rate;

[0142] t P : The time required for this operation step, in seconds.

[0143] in,

[0144] Before continuing the search using the exhaustive method as described in step S12, first delete infeasible paths that do not meet the process requirements.

[0145] in,

[0146] Before sorting the scheduling sequence paths from high to low according to the set scoring rules as described in step S14, infeasible paths that do not meet the process requirements are deleted first.

[0147] in,

[0148] The static parameters are determined based on the empirical average values ​​of historical process parameters.

[0149] in,

[0150] The phrase "according to process requirements" specifically refers to: according to the requirements of the set scheduling rule library.

[0151] Working principle, process and implementation examples

[0152] (Step 1) Based on the actual production situation of Baowu charcoal materials (e.g.) Figure 1 As shown), the state space and events are defined based on objects: the objects are defined as: m crystallization boxes B that perform crystallization production. i (1≤i≤m) and n buffer slots A for storing intermediate products. j (1≤j≤n), specifically 3 crystallization boxes B1, B2, B3, and 9 buffer slots A. 10 A 11 A 12 A 13 A 14 A 15&20 A 16 A 17 A 18&19 ,like Figure 3 As shown; then the state space at a certain time t includes: the liquid level V(B) of each crystallization tank. i ,t)(1≤i≤3), running state s(B i ,t), running time t ran (B i ,t) and the liquid level V(A) of each buffer tank j ,t)(10≤j≤18&19). Each process is treated as an event P, including: C2, C1, P1, P2, P3, P4, P5. The entry of liquid from the buffer tank of this process into the crystallization box is the start of the event. The discharge or sweating liquid in each process is discharged into the buffer tank of the previous or current process. The remaining crystals are discharged into the buffer tank of the next process after melting or waiting for continuous replenishment, which is the end of the event.

[0153] (Step 2) Establish a material flow and scheduling rule base: Material flow is established based on three dimensions: material flow direction, material flow rate, and time; the scheduling rule base is established based on five dimensions: plant level, crystallizer level, buffer tank level, process level, and process operation sequence level; the material flow direction and flow rate are all carried out according to the rules of the scheduling rule base, serving as the principle for process execution; all rules constitute the rule base G, where the material flow direction of each step in each process is as follows: Figures 4 to 12 As shown; where time (unit: minutes), material flow, and flow rate (unit: tons) are all variable parameters, given by the scheduler. Based on the state space and events defined in (step 1) and the actual production of Baowu carbon materials, the following rule base G and related process parameters are established (this part is only a partial list based on the above five dimensions to serve as an example, and is not a complete list of rules), C is defined as the process scheduling queue, C(B i ,k) represents crystallization box B i The process arranged in step k: (1) The feed buffer tank is equipped with a minimum liquid level protection ( and ),in, and These respectively represent the minimum liquid level protection provided by the feed buffer tank during feeding and the minimum liquid level protection provided by the feed buffer tank during replenishment; the following standards shall be followed: the minimum liquid level during feeding operation is 5% of the buffer tank capacity, and the minimum liquid level during replenishment operation for heating and melting is 15% of the buffer tank capacity; (2) the feed buffer tank is provided with a maximum liquid level protection (U high (A 10 )~U hig (A 18&19 (3) The raw material buffer tank is replenished with raw materials according to the input feeding speed v. The amount of raw material replenished in the raw material buffer tank during the time interval t1 to t2 is V. supple (t1, t2) = (t2 - t1) × v, and V must be satisfied. supple (t1, t2) + V(A) raw ,t)≤U(A raw )-U high (A raw ), where U(A j ) indicates buffer slot A j The capacity; taking buffer tank A13 as an example: Buffer tank A13 is replenished with 95% naphthalene according to the input feed rate. Let the feed rate be v. Then, the amount of material replenished in buffer tank A13 within the time range t1 to t2 is (t2-t1)×v. If V0(t) is used to represent the amount of material replenished in buffer tank A13 at time t, then V0(t) + V(A) must be satisfied. 13 ,t)≤U(A 13 )-U high(A 13 (4)

[0154] For feeding and replenishing materials

[0155] During feeding, the current liquid level in the buffer tank is subtracted from the expected feed amount, and the result of the subtraction is compared with the set low alarm liquid level for the buffer tank feed. If it is less than the set low alarm liquid level for the buffer tank feed, the subprocess is postponed; otherwise, the liquid level in the buffer tank and the status of the crystallizer are updated.

[0156] When replenishing material, the current liquid level in the buffer tank is subtracted from the replenishment amount, and the result is compared with the set low alarm liquid level for replenishing the buffer tank. If it is lower than the set low alarm liquid level for replenishing the buffer tank, the subprocess is postponed; otherwise, the liquid level in the buffer tank and the status of the crystallizer are updated.

[0157] Here, it is also necessary to consider whether the plant-level rules allow for under-load operation; if the crystallization box can be filled, then full-load operation is allowed; if it cannot be filled, under the condition that under-load operation is allowed (M>0), all should be loaded, and the filling rate should be calculated. Subsequent operations (discharge, sweating, discharge, etc.) should calculate the discharge volume based on the filling rate; if under-load operation is not allowed, then feeding should be delayed, and V(b(C(B) should be updated after feeding is completed. i ,k)),t)=0;(5) Define r(P,h) as the material flow rate of the h-th step operation of process P, where feeding or replenishing operation represents the raw material feed amount, discharge, sweating, and discharge operation represents the by-product discharge amount, and melting operation represents the product retention amount. Define t(P,h) as the time of the h-th step operation of process P (calculated from the start of process P). Then, for discharge, sweating, and discharge operations, r(C(B i ,k),h)×V(B i (,t) represents the crystallization chamber B at time t. i For the discharge volume in step h of step k, when step h ends, update the status and level of the buffer tank and crystallizer. Add the discharge volume of the corresponding crystallizer to the level of the buffer tank, and compare the sum with the set high alarm level of the buffer tank. If the sum is greater than the high alarm level of the buffer tank, it means that the remaining liquid level in the buffer tank is insufficient, and this step should be postponed. Otherwise, let r(C(B) = ... i ,k),h)×V(B i ,t)+V(b(C(B i ,k),h),t)→V(b(C(B i ,k),h),t), and update the crystallization tank level V(B) i ,t)-r(C(B i ,j),k)×V(B i ,t)→V(B i ,t); (6) process C(B) at time ti After j), if C(B) ends, i ,j+1)=C(B i If (j)+1, it means the next process after the current process is the process following the current process, and no discharge operation is needed. The feeding operation of the subsequent process is changed to a replenishment operation. Otherwise, discharge is required first to b(C(B)+1. i ,j)+1), update the liquid level and status of the buffer tank and crystallization box V(b(C(Bi,k)+1),t)=V(b(C(Bi,k)+1),t)+V(Bi,t),V(B i ,t)=0, and then perform the feeding operation; (7) In order to improve efficiency, try not to execute the P2 process in the crystallizer box 2; (8) According to the scheduler's requirements, the P5 process can only be connected to the P4 process in sequence, and the P4 process can only be connected to the P3 process in sequence. Generally, it is not started alone. Taking the C2 process as an example, its material flow direction, flow rate and operation time are as follows Figure 4 As shown.

[0158] (Step 3) Use an event-triggered model to handle liquid level and status updates: Based on the scheduling rule base G established in step (2), we treat the end of each step of each process in each crystallizer and the start of the next operation t(P,h) as an event to trigger the status update of associated objects. That is, the liquid level of the crystallizer and the buffer tank corresponding to the current process, the previous process, or the next process needs to be updated. Each time the status is updated, we need to "pause" the process, i.e., block the process, to complete the update. For example... Figure 13 The "Blocking, State Update" point is shown in the diagram. As scheduling progresses, each crystallizer executes a different process, and the execution time of each process is generally known. Based on this, we can construct a blocking sequence. Since the total time of each process step is different, we reconstruct the blocking sequence when any crystallizer completes the "melt" operation. For example... Figure 13 The "Regenerate Blocking Sequence" point is shown in the image.

[0159] (Step 4) Use a tree to represent the search process for process scheduling, defining each node of the tree as representing a process step C(B). i A path from root to leaf (h) represents a process sequence, that is, the process arrangement of the crystallizer C(B). i Given the initial production state V(B) i ,t),s(B i ,t),V(A f ,t); Define the path sequence library as W, the threshold number of paths in the path sequence library as N, and the planning time interval length as T. When the number of paths in the path sequence library When the sequence size is less than the threshold N of the sequence library, exhaustively enumerate the possible next actions C(B). i len(C(B)i ))+1) and make it form a new path C(B) i Based on the rule base G, delete infeasible paths. Update the remaining paths to the path sequence base W(R). i )←C(B i ).like Figure 17 As shown.

[0160] (Step 5) As time progresses, the tree continues to grow, until all possible numbers (number of branches, ...) are reached. When the threshold N is exceeded, first exhaustively enumerate all possible next actions C(B). i len(C(B) i ))+1) and make it form a new path C(B) i Then, based on the rule base G, infeasible paths are deleted, and paths outside the top N in the score ranking are deleted (pruned) according to the scoring rule S. The score is defined as the negative of the total delay time for executing that path, S(C(B)). i ))--l total-delay (C(B i The total delay time is calculated as follows: (1) The time t is delayed due to insufficient liquid level in the buffer tank that needs to be drained from the crystallizer. delay (1) Directly include in the total delay time; (2) In addition to the first process that can be performed in each crystallizer, the time t required for material discharge operation due to non-continuous operation of other processes. discharge (3) When full-load operation is allowed (M>0), for crystallizers operating at low load, the total delay time is directly included; i The deferred delay time is calculated and included in the total delay time. The deferred calculation method is as follows: t non-full =(1-f(B) i ))×tC(B i，k ), where t p For C(B) i k) The total time required for the process. The sum of these three is the total delay time: t total-delay (C(B i ))=∑ k t delay +∑ k t discharge +∑ k t non-full Update the remaining paths to the path sequence library W(B). i )←C(B i ).like Figure 18 As shown.

[0161] (Step 6) Record the total execution time of each path as path time t. path When t pathIf the time interval is less than the planned time interval length T, repeat step (4) or step (5) and update t. path When t path When T is greater than or equal to T, all path sequences in the path sequence library W are sorted according to the evaluation rule S, and the path with the highest score is provided to the production scheduler.

[0162] The present invention discloses an automatic optimization method for naphthalene crystallization production. First, a simulation process driven by trigger events is set up to provide a searchable basis for path search. Second, based on this, an algorithm-based optimization is performed using an exhaustive search method combined with a pruning method. The naphthalene crystallization production is established according to the target path provided by the optimization results, thus completing the automatic optimization of naphthalene crystallization production.

Claims

1. An event-based automatic optimization method for production yield of fine naphthalene production, characterized by The method comprises the following steps: S1: establishing event-driven simulation process advancement, S2: establishing path search according to the simulation process advancement, and determining the target path according to the search result, S3: establishing the production program of refined naphthalene according to the determined target path; The step S1 is established based on static parameters, and the step S1 specifically comprises: setting the end of each sub-process or the start of the next sub-process in the refined naphthalene crystallization production as a trigger event, advancing the simulation process according to the set trigger event, and the advancement of the simulation process according to the set trigger event specifically comprises: establishing a blocking point according to the expected end time of each sub-process of each process, and forming a blocking point sequence of each process according to the blocking point, establishing a blocking queue according to the blocking point sequence of each process and the execution sequence between each process, advancing the simulation process according to the blocking point sequence of each process, when the blocking queue is constructed, if the next process is the subsequent process of the previous process, a make-up blocking point is constructed, otherwise, the blocking points are established in the order of first discharging and then feeding, when the trigger event occurs, the next sub-process is judged according to the following steps: SS1: judging whether the next process is the first process set for the process corresponding to the crystallization tank, if yes, step SS5 is entered, otherwise, step SS2 is entered; SS2: if the next process is the subsequent process of the previous process, step SS5 is entered, otherwise, step SS3 is entered; SS3: judging whether the buffer tank liquid level reserve corresponding to the next sub-process meets the process requirement, if not, step SS6 is entered, otherwise, step SS4 is entered; SS4: calculating the discharging time; SS5: judging whether the buffer tank liquid level reserve corresponding to the next sub-process meets the process requirement, if not, step SS6 is entered, otherwise, step SS7 is entered; SS6: postponing the next sub-process, and calculating the postponing time; SS7: judging whether the crystallization tank corresponding to the sub-process is in full-load operation, if not, the delay time is converted according to the filling rate, the path search in the step S2 is based on a tree structure.

2. The method according to claim 1, characterized in that: each process is defined as each node of the tree structure, and a process sequence is formed from the root to the leaf, and a scheduling sequence path is formed based on the process sequence; the scheduling sequence path search is performed by the exhaustive method combined with the pruning algorithm.

3. The method according to claim 2, characterized in that: the scheduling sequence path search performed by the exhaustive method combined with the pruning algorithm specifically comprises the following steps: S11: calculating the total execution time of the scheduling sequence path in real time, and comparing the calculation result with the set time length, if the calculation result is less than the set time length, step S12 is entered, otherwise, step S13 is entered; S12: calculating the number of scheduling sequence paths in real time, and comparing the calculation result with the set threshold value; if the calculation result is less than the set threshold value, the search is continued according to the exhaustive method, otherwise, step S14 is entered. S13: According to the set scoring rule, the scheduling sequence path is sorted based on the score from high to low, and the scheduling sequence path with the first sorting is selected as the target path; S14: According to the set scoring rule, the scheduling sequence path is sorted based on the score from high to low, and the number of scheduling sequence paths is counted according to the sorting result. According to the set threshold, the number exceeding the set threshold is deleted.

4. The event-based automatic optimization method for production yield of refined naphthalene according to claim 3, characterized in that: The score is determined according to the total delay time of each scheduling sequence path, and is specifically set as the inverse number of the total delay time; The total delay time is determined according to the following formula: t total-delay (C(B i , j)) = t delay + t discharge + t non-fulll , Wherein, C represents the process scheduling sequence, C(B i , j) = B i , j) + t delay , j) where B i , j) is the number of the crystallization tank corresponding to the process, j is the number of the sub-process corresponding to the process, t i , j) is the delay time, unit: S t discharge : Discharge time, unit: S; t non-full : Delay time in seconds for non-full operation of the crystallizer, converted by the filling rate, in case of non-full operation is allowed.

5. The event-based automatic optimization method for production yield of refined naphthalene according to claim 4, characterized in that: t non-full = (1 - f) x t p , Wherein, f: filling rate; t p : Time required for this operation step sequence, unit: S.

6. The event-based automatic optimization method for production yield of refined naphthalene according to claim 3, characterized in that: Before the search according to the exhaustive method in step S12, the infeasible path that does not meet the process requirements is deleted according to the process requirements.

7. The event-based automatic optimization method for production yield of refined naphthalene according to claim 3, characterized in that: Before the sorting of the scheduling sequence path based on the score from high to low according to the set scoring rule in step S14, the infeasible path that does not meet the process requirements is deleted according to the process requirements.

8. The event-based automatic optimization method for production yield of refined naphthalene according to claim 1, characterized in that: The static parameter is determined based on the empirical average value of the historical process parameters.

9. The event-based automatic optimization method for production yield of refined naphthalene according to claim 6, characterized in that: The process requirements are specifically set according to the set scheduling rule library requirements.

Images