An automatic control system and control method for producing lithium carbonate

Abstract

The present application relates to the technical field of digital control, and more particularly to an automatic control system and control method for producing lithium carbonate. The method comprises the following steps: obtaining lithium carbonate production line sensor data and binding to obtain a process section field state factor set; fine-tuning embedded time state migration based on the process section field state factor set to obtain a multi-dimensional control representation flow; encoding control instructions based on the multi-dimensional control representation flow to obtain a multi-channel control factor set; recalculating field load parameters and micro-period feedback modulation based on the obtained real-time lithium carbonate production line sensor data to obtain a control modulation response matrix; and adjusting a process graph structure in reverse based on the control modulation response matrix to obtain a process abnormality feedback body. The present application helps to enhance the real-time response capability of the lithium carbonate production line to raw liquid fluctuations, equipment state changes and abnormal working conditions, effectively improving product purity, yield and production line operation stability.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of digital control, and in particular to an automatic control system and method for producing lithium carbonate. BACKGROUND

[0002] At present, the preparation process of lithium carbonate in industry mainly includes lithium extraction from salt lake brine, lithium extraction from spodumene calcination and lithium extraction from recycled materials. Among them, lithium extraction from salt lake brine has become the mainstream industrial lithium extraction method at home and abroad because of its abundant raw material resources and low production cost. However, the production process of lithium carbonate involves multiple stages of material separation, solution concentration, precipitation reaction, filtration, drying and other operations, and the process chain is long and the reaction conditions are complex. Moreover, it is greatly affected by many factors such as temperature, pH value, reaction time, and composition fluctuation of raw solution. Therefore, it is difficult for the traditional control method to realize stable, continuous and efficient industrial operation. Most of the existing production lines use a distributed control structure based on a PLC system to set threshold values for key process parameters such as concentration, temperature, and flow rate, and use on-off control or proportional integral adjustment. However, this type of control method generally has problems such as response lag, inability to dynamically adapt to fluctuations in raw solution quality, and fixed system parameters, which leads to large fluctuations in product purity and yield. For example, in the lithium precipitation reaction stage, the Na2CO3 dosage or temperature curve cannot be accurately adjusted according to the real-time reaction rate, which may cause excessive precipitation or incomplete crystallization, affecting the subsequent filtration efficiency and product quality. In the concentration and evaporation stage, fluctuations in the concentration of the raw solution may cause premature crystallization or imbalance of the mother liquor, resulting in equipment scaling, frequent maintenance, and other problems. In addition, the difference in the composition of different batches of raw materials often requires manual intervention to set parameters, and there is a lack of intelligent identification and self-adaptive control capability, which greatly limits the process continuity and product standardization level. SUMMARY

[0003] Therefore, it is necessary to provide an automatic control system and method for producing lithium carbonate to solve at least one of the above technical problems.

[0004] To achieve the above-mentioned purpose, an automatic control method for producing lithium carbonate includes the following steps:

[0005] Step S1: Obtain lithium carbonate production line sensor data, and perform process section physical quantity structuring binding on the lithium carbonate production line sensor data to obtain a process section field state factor set;

[0006] Step S2: Perform multi-step prediction vector extraction according to the process section field state factor set, and perform fine-tuning embedded time state migration on the multi-step prediction vector to obtain a multi-dimensional control representation flow;

[0007] Step S3: Perform digital twin control simulation based on the multi-dimensional control representation flow, and perform control instruction coding on the digital twin control simulation result to obtain a multi-channel control factor set;

[0008] Step S4: transmit the multi-channel control factor set to the production line management platform to perform the production line control task; combine the acquired real-time lithium carbonate production line sensing data to perform on-site load parameter recalculation and micro-period feedback modulation, and obtain a control modulation response matrix;

[0009] Step S5: perform process offset tracing based on the control modulation response matrix, and perform process map structure reverse adjustment on the process section on-site state factor set according to the process offset tracing result, to obtain a process exception feedback body;

[0010] Step S6: take the process exception feedback body as the input of a new round of step S2, and repeat steps S2 to S5 until the process offset is less than a preset process offset threshold.

[0011] Optionally, step S1 specifically comprises:

[0012] Step S11: collect physical sensing data of each process section of the lithium carbonate production line, and aggregate the data according to a preset sampling period to obtain lithium carbonate production line sensing data;

[0013] Step S12: perform data cleaning and time sequence alignment on the lithium carbonate production line sensing data, remove missing values and abnormal points, and perform uniform time step completion to generate a preprocessed data set;

[0014] Step S13: perform physical quantity-process section mapping binding according to the preprocessed data set to obtain a physical quantity-process section relationship table;

[0015] Step S14: classify and integrate the preprocessed data set based on the physical quantity-process section relationship table, and output a structured set according to the process section;

[0016] Step S15: convert the structured set into a state factor vector, sequentially splice the vector according to the data collection time, and thus form a process section on-site state factor set.

[0017] Optionally, step S13 specifically comprises:

[0018] Step S131: extract the sensor ID, sensor installation position number and data collection time corresponding to each type of physical quantity in the preprocessed data set, and thus construct a basic metadata index table;

[0019] Step S132: connect the production line management platform, load the lithium carbonate production line process flow configuration file, extract the spatial position range, physical action label and corresponding process section number of each process section, and construct a process section structure table;

[0020] Step S133: Compare the basic metadata index table with the process section structure table, establish a mapping relationship according to the sensor installation position number and the spatial position range of the process section, and mark the corresponding process section number of each physical quantity to obtain a physical quantity-process section mapping relationship;

[0021] Step S134: Structurally express the physical quantity-process section mapping relationship according to a preset process section number-physical quantity type-sensor identifier-data acquisition frequency framework to form a structured mapping relationship table;

[0022] Step S135: Represent the structured mapping relationship table in a graph structure, wherein the nodes represent the process sections, the edges represent the binding paths of the physical quantities, and the node attributes are the physical quantity sets, to generate a physical quantity-process section relationship table.

[0023] Optionally, the multi-step prediction vector extraction in step S2 is specifically as follows:

[0024] Set the window length as T=60 seconds and the sliding step as 10 seconds, time series segment sets are generated by time series slicing on the process section field state factor set;

[0025] The time series segment set is subjected to multi-channel trend deconstruction processing to obtain a standardized state trend trajectory set;

[0026] Based on the standardized state trend trajectory set, multi-layer sliding convolution is performed to predict the trend prediction vectors of the next 6 sliding steps to generate an initial multi-step prediction sequence set;

[0027] The initial multi-step prediction sequence set is subjected to inter-section data correlation cross adjustment to eliminate low-confidence vectors to obtain a refined prediction vector set;

[0028] The refined prediction vector set is encapsulated according to the formatting framework of the process section number, the physical quantity type and the time series to obtain a multi-step prediction vector set.

[0029] Optionally, the fine-tuning embedded time state transition in step S2 is specifically as follows:

[0030] The time series similarity between the multi-step prediction vector set and the current state factor in the process section field state factor set is calculated, and the multi-step prediction vector set is re-weighted and fine-tuned and fused according to the time series similarity to generate a re-tuned prediction vector group;

[0031] According to the process section structure table, the re-tuned prediction vector group is embedded in the upstream and downstream sections synchronously to generate a section-level time transition sequence;

[0032] The change rate, mutation frequency and steady state interval features in the section-level time transition sequence are extracted and encoded into a state transition vector;

[0033] The state transition vector is compressed and mapped to output a low-dimensional control representation code;

[0034] The low-dimensional control representation codes of each section are time-sequentially spliced and labeled with physical action labels to obtain a multi-dimensional control representation stream.

[0035] Optionally, step S3 is specifically:

[0036] Step S31: connecting a digital twin physical simulation platform; loading a corresponding digital twin simulation model from the digital twin physical simulation platform according to the physical action label in the multi-dimensional control representation stream, and initializing simulation input parameters to generate a simulation model configuration set;

[0037] Step S32: inputting the simulation model configuration set into the digital twin physical simulation platform to run a multi-section parallel process response simulation to obtain simulation response data;

[0038] Step S33: performing control target index matching on the simulation response data, and performing control simulation performance evaluation according to the control target index to select a control input trajectory with the highest performance evaluation to obtain an optimal control trajectory set;

[0039] Step S34: parameter discretization, format reorganization and time synchronization of the control variables in the optimal control trajectory set to generate a standard control instruction stream;

[0040] Step S35: control channel splitting and instruction concurrent mapping of the standard control instruction stream according to the control instruction execution relationship in the lithium carbonate production line process configuration file to generate a multi-channel control factor set.

[0041] Optionally, the on-site load parameter recalculation and micro-period feedback modulation in step S4 are specifically:

[0042] Receiving real-time lithium carbonate production line sensing data controlled by the production line and actuator control logs of executed control instructions;

[0043] According to the real-time lithium carbonate production line sensing data and the actuator control logs, logical discrimination is performed to extract feedback values of corresponding control variables to obtain an actuator running state data set;

[0044] Aligning the actuator running state data set with the multi-channel control factor set before the production line control in the field, calculating the response offset of each control channel, and performing leveling processing to generate a load parameter recalculation matrix;

[0045] Based on the load parameter recalculation matrix, and using the inertia compensation time window and equipment dynamic response characteristics of each process section in the lithium carbonate production line process configuration file, disturbance quantities in a preset micro-period feedback interval are extracted to generate a feedback modulation factor group;

[0046] The feedback modulation factor group is channel-mapped and matched with the process section structure table, and a control modulation response matrix is output.

[0047] Optionally, step S5 is specifically:

[0048] Step S51: According to the response offset of each control channel in the control modulation response matrix and the process section field state factor set, the actual state change amplitude is calculated, the real-time control response offset value of the process section is calibrated, and a process control response offset mapping table is obtained.

[0049] Step S52: The process control response offset mapping table is compared with the physical quantity-process section relationship table, the process section and the associated physical quantity where the offset occurs are located according to the process section number, and the offset cause weight is calculated, so as to obtain a process offset associated factor group.

[0050] Step S53: According to the process offset associated factor group, the path reverse tracking is performed in the graph structure of the physical quantity-process section relationship table, and the offset influence path is obtained.

[0051] Step S54: The offset influence path is structurally adjusted to generate a process map reverse adjustment index table.

[0052] Step S55: The process map reverse adjustment index table is applied to the process section field state factor set, the structured binding relationship is updated, and the abnormal response section is marked, and a process exception feedback body is generated.

[0053] Optionally, step S54 is specifically:

[0054] Step S541: The control correlation degree, process action intensity and conduction direction of each physical quantity node in the offset influence path are extracted, and an offset path structure factor matrix is constructed.

[0055] Step S542: According to the offset path structure factor matrix, the influence intensity of the control response offset on each physical quantity node is analyzed, and the path segments with influence intensity exceeding [0.15, 0.25] are marked to obtain a path abnormal segment identification set.

[0056] Step S543: The path segments in the path abnormal segment identification set are executed for node replacement, weight reconstruction and direction optimization, and the structure change operation is recorded to obtain a path structure change record.

[0057] Step S544: The path structure change record is indexed and layered coded according to the process section number and the physical quantity identification to obtain a process map reverse adjustment index table.

[0058] The present application is directed to the problems of control response lag, large process fluctuation, manual intervention, low process intelligence level and other problems existing in traditional lithium carbonate production, a multi-stage coordinated closed-loop intelligent control method is proposed, which has a significant technical upgrading effect. Through the process section physical quantity structured binding of various types of sensing data of the lithium carbonate production line in step S1, the running state of each process section can be accurately restored, forming a set of field state factors that can be used for control modeling, avoiding the problem of "physical quantity separation from process semantics" in traditional data collection; wherein the mapping relationship table is constructed by process section number, physical quantity type, sensor identifier and other contents in S13, which can enhance the structure clarity of data and the accuracy of subsequent process mapping, which is beneficial to multi-region segmented modeling in downstream models. In step S2, the window length T=60 seconds and the sliding step 10 seconds are set to slice the time series, ensuring that the state trend can cover the key process change period, and the real-time and trend integrity can be considered in the prediction vector extraction; the combination of multi-channel trend deconstruction processing and multi-layer sliding convolution can effectively extract short-term trend changes and enhance the local accuracy of prediction, avoiding the omission of inter-zone linkage characteristics in traditional single variable modeling; the time series similarity calculation mechanism and upstream and downstream embedding strategies introduced in the fine-tuning embedded time state migration improve the adaptability of the prediction results to the process path, making each step of prediction more consistent with the actual control response path; the low-dimensional control representation flow of the state migration vector after compression mapping takes into account the feature expression ability and system calculation efficiency. Step S3 connects the digital twin simulation platform, and based on the multi-dimensional control representation flow, the simulation model corresponding to the process section is called, which can simulate the response effect of various control instructions on the production line operation in advance, avoiding the waste of raw materials and the increase of equipment load caused by actual trial and error; control simulation performance evaluation ensures that the final issued instruction trajectory is the most optimal in performance, improving the effectiveness and target consistency of the control instruction; strictly referring to the process flow configuration file during control channel splitting helps to ensure the process consistency during the instruction issuing process. In step S4, the aligning mechanism of the actuator control log and the sensing data is introduced, which ensures that the feedback source is real and traceable, and the control deviation can be quantified in real time through the load parameter recalculation matrix; the disturbance quantity is extracted by using the inertia compensation time window parameter in the process flow file combined with the dynamic response characteristics of the equipment, and the feedback modulation factor group is generated, so as to realize small dynamic adjustment in a short period, ensure that the system has fine-grained and high-frequency fine-tuning ability, and improve the response sensitivity and working condition self-adaptability.Step S5 can accurately locate the offset source area by controlling the response offset mapping of the channel, and realize the cause analysis and influence range restoration of the offset by means of the process map structure for reverse path tracking and structure adjustment; the path segment influence intensity is selected in the interval [0.15, 0.25] in the structure adjustment, which is beneficial to identify the key path of moderate intensity disturbance, avoiding both excessive reaction to weak disturbance and loss of control due to inertia of strong disturbance, and the control is more reasonable and reasonable; the path structure change operation such as node replacement and weight reconstruction can dynamically update the process model structure, providing input conditions with adaptability under the current production scene for subsequent feedback body generation. Finally, the present scheme forms a closed-loop control architecture with data-driven, self-correcting and continuous optimization by circulating the process abnormal feedback body into the control modeling process, which has strong robustness and high self-adaptive ability, significantly enhances the real-time response capability of the lithium carbonate production line to the raw liquid fluctuation, equipment state change and working condition abnormality, and effectively improves the product purity, yield and production line running stability.

[0059] Optionally, the present specification also provides an automatic control system for producing lithium carbonate for executing the automatic control method for producing lithium carbonate as described above, and the automatic control system for producing lithium carbonate comprises:

[0060] A field state identification module is configured to acquire lithium carbonate production line sensing data, and perform process section physical quantity structural binding on the lithium carbonate production line sensing data to obtain a set of process section field state factors.

[0061] A multi-step prediction module is configured to perform multi-step prediction vector extraction according to the set of process section field state factors, and perform fine-tuning embedded time state migration on the multi-step prediction vector to obtain a multi-dimensional control representation flow.

[0062] A control simulation module is configured to perform digital twin control simulation based on the multi-dimensional control representation flow, and encode control instructions based on the digital twin control simulation result to obtain a set of multi-channel control factors.

[0063] A production line control module is configured to transmit the set of multi-channel control factors to a production line management platform to perform a production line control task; and perform field load parameter recalculation and micro-period feedback modulation based on the acquired real-time lithium carbonate production line sensing data to obtain a control modulation response matrix.

[0064] A process offset optimization module is configured to perform process offset tracing based on the control modulation response matrix, and perform reverse adjustment of the process map structure on the set of process section field state factors based on the process offset tracing result to obtain a process abnormal feedback body.

[0065] The iterative parameter adjustment module is configured to feed the process abnormality feedback body as a new round of input of the multi-step prediction module, and repeatedly iterate the multi-step prediction module to the process offset optimization module until the process offset is less than a preset process offset threshold.

[0066] The automatic control system for producing lithium carbonate can realize any one of the automatic control methods for producing lithium carbonate, and is used as a medium for combining operations and signal transmission between modules to complete the automatic control method for producing lithium carbonate. BRIEF DESCRIPTION OF DRAWINGS

[0067] Other features, objects, and advantages of the application will become more apparent from the following detailed description of non-limiting embodiments thereof, read in conjunction with the accompanying drawings:

[0068] Fig. 1 A step flowchart of the automatic control method for producing lithium carbonate of the application;

[0069] Fig. 2 A detailed step flowchart of step S1 in the application;

[0070] Fig. 3 A detailed step flowchart of step S13 in the application;

[0071] The implementation of the object of the application, functional features, and advantages will be further described with reference to the embodiments and the accompanying drawings. DETAILED DESCRIPTION

[0072] The technical method of the application will be described clearly and completely below in conjunction with the accompanying drawings. Obviously, the described embodiments are part of the embodiments of the application, rather than all the embodiments. Based on the embodiments in the application, all other embodiments obtained by those skilled in the art without creative labor fall within the protection scope of the application.

[0073] In addition, the accompanying drawings are only schematic illustrations of the application, and are not necessarily drawn to scale. The same reference numerals in the drawings represent the same or similar parts, and thus repeated descriptions thereof will be omitted. Some block diagrams shown in the drawings are functional entities, which do not necessarily correspond to physically or logically independent entities. The functional entities can be implemented in the form of software, or in one or more hardware modules or integrated circuits, or in different network and / or processor methods and / or microcontroller methods.

[0074] It should be understood that, although the terms "first", "second", etc. can be used herein to describe various elements, these elements should not be limited by these terms. These terms are only used to distinguish one element from another. For example, a first element could be termed a second element, and, similarly, a second element could be termed a first element, without departing from the scope of the example embodiments. The term "and / or" as used herein includes any and all combinations of one or more of the associated associated items.

[0075] To achieve the above object, the present application provides an automatic control method for producing lithium carbonate, comprising the following steps: Figs. 1 to 3 The present application provides an automatic control method for producing lithium carbonate, comprising the following steps:

[0076] Step S1: Obtain lithium carbonate production line sensor data, and perform process section physical quantity structured binding on the lithium carbonate production line sensor data to obtain a process section field state factor set;

[0077] In this embodiment, industrial sensor terminals (including conductivity sensors, pH probes, thermocouples, flow meters, etc.) are deployed at each key process section of the lithium carbonate production line, and various physical quantity data is collected at a sampling period of 2 seconds and is connected to an edge computing module for preliminary caching. Using a data classification service deployed on the edge side, the original sensor data is labeled and classified according to the "process section number-physical quantity type" according to the corresponding relationship of the section physical quantity defined in the process flow configuration file. Subsequently, after the structured data cleaning and time alignment process, including linear interpolation filling for missing values and median replacement processing for mutation values, the data is uniformly processed into continuous state time series with a step of 5 seconds. Finally, through the binding mapping rule of physical quantity and process section, the classified physical quantity data is encapsulated as a structured state factor vector, forming a process section field state factor set covering the process sections of lithium precipitation, concentration, filtration, and drying.

[0078] Step S2: Extract a multi-step prediction vector according to the process section field state factor set, and perform fine-tuning embedded time state migration on the multi-step prediction vector to obtain a multi-dimensional control representation flow;

[0079] In this embodiment, the process section field state factor set is used as input to extract the trend value and variation direction within the next 5 time steps (i.e. 50 seconds) from the time series of each process section, and combined into a multi-step prediction vector. During the extraction process, the prediction sequence is updated at a 20-second interval through a sliding window mechanism. After generating the initial prediction vector, it is compared item by item with the current state factor to calculate the dynamic Euclidean distance as a time series similarity index. According to the similarity size, different prediction vectors are given different adjustment weights, and a linear fusion is performed to generate a re-adjusted prediction vector group. Then, according to the action path between the upstream and downstream sections in the process map, the prediction results are embedded into the state prediction of the upstream and downstream sections in a synchronous manner to obtain a time migration sequence covering multiple process sections. Then, the gradient change rate, mutation frequency, and steady-state holding time in the sequence are extracted and uniformly encoded into a state migration vector. Finally, the state vector is reduced in dimension using principal component compression, and a low-dimensional representation is output. After labeling each control item with a physical quantity, a multi-dimensional control representation stream is generated.

[0080] Step S3: Perform digital twin control simulation based on the multi-dimensional control representation stream, and encode the digital twin control simulation results to obtain a multi-channel control factor set;

[0081] In this embodiment, after receiving the multi-dimensional control representation stream, a lithium carbonate production line digital mirror system reflecting the actual physical structure is built in the virtual environment. The system loads the actual process flow configuration file, including the number, spatial position, medium type, action mechanism, and dynamic response time of each process section. For example, the response delay of the precipitation section is set to 12 seconds, and the drying section is 8 seconds. The multi-dimensional control representation stream is input into the simulation system one by one to simulate the state evolution path after the control action is triggered, and to evaluate its influence on the target productivity, purity, and energy consumption indicators. By comparing the control input and simulation feedback at each time, the optimal control action set is extracted and encoded according to the physical quantity, action point number, and time delay to form a transmissible multi-channel control factor set. Each control factor contains five fields: control variable name, application range, control intensity, target process section number, and application time point.

[0082] Step S4: Transmit the multi-channel control factor set to the production line management platform to perform production line control tasks; combine the acquired real-time lithium carbonate production line sensor data to recalculate the field load parameters and perform micro-cycle feedback modulation to obtain a control modulation response matrix;

[0083] In this embodiment, the multi-channel control factor set is sent to the production line management platform through the standard control interface, and the platform maps it to the corresponding execution unit according to the preset control protocol. When the control action is completed, the real-time sensor data 10 seconds before and after the control is collected, and the control execution log is obtained. By comparing the target change of the control factor with the actual change range of the sensor data, the response offset value of the control variable is extracted. For example, the target flow adjustment is to increase by 15%, but the actual increase is only 10%, and the offset value is 5%. After sorting all the offset values of the control channels, a load parameter recalculation matrix is constructed. Combined with the inertia compensation time window set in the process flow configuration file (such as 6 seconds for the sedimentation section) and the equipment dynamic response characteristics (such as the valve adjustment speed of 0.2 seconds / unit flow), the control perturbation is extracted within the 5-second micro-period feedback interval to form a feedback modulation factor group. Map this factor group with the process section structure to generate a structured control modulation response matrix as the basis data for subsequent tracing and adjustment.

[0084] Step S5: Based on the control modulation response matrix, trace the process offset, and adjust the process section field state factor set according to the process offset tracing result to obtain the process abnormal feedback body;

[0085] In this embodiment, based on the control modulation response matrix, the quantitative relationship between the response offset of each control channel and the state change of the process section is analyzed item by item to form a control response offset mapping table, which records the process section number and the corresponding offset amplitude. On this basis, compare the physical quantity-process section binding table to confirm the specific process section and its related physical quantity that causes the offset. By comparing the degree of influence of each factor in different path nodes, calculate which physical quantities may cause each offset, and assign offset cause weights. For example, if the pH value is highly related to the offset amplitude, the physical quantity is given a higher weight. Then perform reverse path tracking in the process map, search for possible source physical nodes from the offset point along the map edge in reverse, extract the path structure, and record the influence direction and intensity of each segment. Analyze these paths, extract path segments with an influence intensity greater than 20%, and perform node replacement and path structure adjustment. Finally, encode these adjustment operations according to the process section number into a structured table, and use it to update the state factor set to generate a new process abnormal feedback body for the next round of adjustment input.

[0086] Step S6: Take the process abnormal feedback body as the input of step S2 in the next round, repeat steps S2 to S5 until the process offset is less than the preset process offset threshold.

[0087] In this embodiment, the generated process anomaly feedback body is used as the input for the next round of prediction, replacing the original state factor set, and re-entering the prediction process. The mapping relationship between the state factors and the control factors is updated once every complete cycle from prediction to control to feedback adjustment, and the maximum deviation value of the current process deviation is recorded. The basis for determining whether to enter the next cycle is whether the current maximum deviation value is lower than the preset threshold value, 0.5%. If the requirement is not met, the adjusted state feedback is used as the input to enter the next iteration. During the cycle, the stability of the state of multiple key process sections, the consistency of the response, and the prediction accuracy are monitored at the same time to ensure that the adjustment effect is optimized while maintaining process stability. Finally, when all the deviation values are within the threshold value, the control iteration process is ended, and the final control result is used as the basis for updating the real-time control instruction.

[0088] Optionally, step S1 is specifically:

[0089] Step S11: Collecting physical sensor data of each process section of the lithium carbonate production line, and collecting the data according to a preset sampling period to obtain lithium carbonate production line sensor data;

[0090] In this embodiment, in the lithium carbonate production line, temperature, pH, conductivity, flow, pressure, and liquid level six types of physical sensors are arranged in the extraction, precipitation, neutralization, centrifugation, and drying key process sections, industrial-grade RS485 communication protocol is selected to connect with the PLC control end, and data collection tasks are performed with a fixed sampling period of 5 seconds. The numbers of various sensors are combined according to the process section numbers and equipment type codes, for example, “D-03-TEMP” represents the third temperature sensor in the drying section. The collection task is triggered by the central control system at regular intervals, and after each batch of data collection, the data is uniformly collected into the production data center server, and the storage structure adopts a process section-time stamp double-indexing method. In order to ensure the synchronization of the data of each section, the data collection program uses a data caching mechanism, when the data of a section is not reported due to sensor delay, it is not written into the database temporarily, and after all the data of the current sampling period are ready, batch data is written into the database.

[0091] Step S12: Data cleaning and time sequence alignment are performed on the lithium carbonate production line sensor data, missing values and abnormal points are removed, and uniform time step completion is performed to generate a preprocessed data set;

[0092] In this embodiment, the lithium carbonate production line sensor data collected in the previous step is taken as input. First, time stamp alignment is performed to unify the time step of 5 seconds, and a complete timeline sequence is constructed. Missing time points on the time axis are filled by linear interpolation. If a sensor does not return valid data for three consecutive times, it is marked as "NA" in the data set and recorded as a missing segment. During data cleaning, threshold intervals are set, such as pH value range of 3.5 to 8.5 and temperature range of 15°C to 95°C. When the monitoring data exceeds the physical reasonable range, it is deleted. To avoid false deletion of transient fluctuations, the abnormal value needs to be maintained for 3 consecutive periods before being removed. The final preprocessed data set is composed of three fields of sensor ID, time stamp and numerical value, and the original data identifier is retained for subsequent comparison.

[0093] Step S13: According to the preprocessed data set, the physical quantity-process section mapping binding is performed to obtain the physical quantity-process section relationship table;

[0094] In this embodiment, after preprocessing, the sensor and process section binding table established in the configuration file is used to map and match the physical quantity and process section for each data. The binding table is constructed in three columns, namely "sensor number", "belonging process section number" and "physical quantity type". For example, the sensor with number "C-04-PH" can be corresponded to "precipitation section" and "pH value" physical quantity through the table item. In the matching process, if a sensor is not defined in the configuration table, it is recorded in the unbound list for manual supplement. To improve the mapping efficiency, the system hashes the sensor number to establish a quick index table, and the matching result is directly appended to the original data set to form new data fields "bound process section number" and "physical quantity name". Through this mapping, unstructured sensor data can be quickly classified into actual process sections, providing a basis for subsequent data structure reconstruction.

[0095] Step S14: Based on the physical quantity-process section relationship table, the preprocessed data set is classified and integrated, and the structured set is output according to the process section;

[0096] In this embodiment, after obtaining the data set with physical quantity and process segment identifier, all data is partitioned and integrated according to the process segment number. For example, after extracting all the data identified by the "drying segment", the data is arranged in ascending order of timestamp, classified and arranged according to the "temperature", "flow", "pressure" fields, and the structured data set of the drying segment is constructed. The data set of each process segment contains a time column, multiple physical quantity columns and a data source identifier column. In the integration process, an in-memory buffer partitioning strategy is used. First, a cache area is created according to the process segment number, and then the corresponding data is filled in sequence. After filling, the result structure is written uniformly. For cross-segment data (such as temperature sensors arranged across two segments), the physical center point is assigned to the process segment with relatively high proportion. The final output result is a structured collection containing all process segments, and each collection encapsulates the data details of all effective time ranges of each segment.

[0097] Step S15: Convert the structured collection into a state factor vector, sequentially splice according to the data collection time, and thus form a process segment field state factor set.

[0098] In this embodiment, the physical quantities in the output structured collection are spliced in time sequence to form a time-continuous state factor vector stream. Taking the drying segment as an example, according to the data sampling frequency of every 5 seconds, the temperature, flow and pressure data are arranged in sequence according to the field to construct a single time point state factor vector. All time point vectors are arranged in ascending order of time to form a drying segment state factor sequence. In order to handle the asynchronous sampling of different segments, interpolation and normalization processing is performed on each segment state factor sequence to ensure that there is a complete factor vector at each time point. The final result is a field state factor set containing all process segments, which is structured as a two-layer structure of process segment number and corresponding time sequence. Each sequence node contains a group of state factor data, which is used for subsequent prediction modeling and control response analysis.

[0099] Optionally, step S13 is specifically:

[0100] Step S131: Extract the sensor ID, sensor installation position number and data collection time corresponding to each type of physical quantity in the preprocessed data set, thereby constructing a basic metadata index table;

[0101] In this embodiment, before structurally managing the pre-processed data set, the basic key information fields therein are first extracted, including the sensor ID corresponding to each piece of data, the physical quantity type, the collection time, and the specific location number where the sensor is installed. Taking actual data as an example, if the record of a certain temperature data is “TEMP-06-123.4-20240508 08:15:10”, wherein “TEMP-06” is the sensor number, “123.4” is the measured value, and “20240508 08:15:10” is the timestamp, the system parses the record as the sensor ID being “TEMP-06”, the location number being “D-04-Z3” (representing the 4th work station in the 3rd block in the dry section) through configuration file parsing, and the data time being “20240508 08:15:10”. Through batch processing, the system structurally deconstructs the records under all different physical quantities, sorts them according to the triplets of time, location, and physical quantity type, and finally forms a metadata index table for subsequent binding and comparison, which has high efficient retrieval and spatial section positioning capabilities.

[0102] Step S132: Connect the production line management platform, load the lithium carbonate production line process configuration file, extract the spatial location range, physical action label, and corresponding process section number of each process section, and construct a process section structure table;

[0103] In this embodiment, by establishing communication with the management platform of the lithium carbonate production line, the REST interface is called to load the “lithium carbonate process flow configuration file” stored in the backend. The configuration file is in JSON format, and the content includes the number, physical action category, geographical distribution boundary (defined by plane coordinate block), and upstream and downstream connection information of each process section. For example, the definition of the “precipitation section” contains its spatial range “[X1=12.0, Y1=4.5, X2=16.0, Y2=7.5]”, the action label “precipitation reaction”, and the section number “SEC-02”. After parsing all paragraphs, a process section structure table is constructed, arranged in ascending order of number, and accompanied by the action category and spatial boundary fields. The table is structured as a five-field data structure of “section number—spatial range—physical function—connected upstream section number—connected downstream section number”, providing a spatial and logical matching basis for subsequent mapping.

[0104] Step S133: Compare the basic metadata index table with the process section structure table, establish a mapping relationship according to the sensor installation location number and the spatial location range of the process section, and mark the corresponding process section number of each physical quantity to obtain the physical quantity-process section mapping relationship;

[0105] In this embodiment, the established metadata index table and the constructed process section structure table are spatially matched with the sensor installation position number as the core. The specific process is to parse the installation position number of each sensor into the corresponding coordinate area, and then perform regional inclusion judgment with the spatial boundary defined by each process section. If a temperature sensor position number is "D-03-Z4", the coordinate point [X=14.5, Y=6.0] is obtained through spatial decoding, and the inclusion verification is performed with the boundary of the process section "SEC-02". Once it is confirmed to fall within the defined range, it is marked as belonging to "SEC-02". This process is processed for all sensors one by one, while recording the corresponding physical quantity type and attribution section number of each sensor, thereby generating a complete "physical quantity-process section mapping relationship". This result is used for directional binding and functional linkage judgment of process data in the subsequent.

[0106] Step S134: The physical quantity-process section mapping relationship is structured and expressed according to the preset process section number-physical quantity type-sensor identifier-data acquisition frequency framework, forming a structured mapping relationship table.

[0107] In this embodiment, the mapping result obtained is arranged into a structured expression format. The specific framework is set to a four-level structure, i.e. "process section number-physical quantity type-sensor identifier-acquisition frequency". Taking the drying section (number SEC-05) as an example, its subordinate temperature sensors include "TEMP-05-A" and "TEMP-05-B", and their acquisition frequencies are both 5 seconds. The mapping is "SEC-05-temperature-TEMP-05-A-5s" and "SEC-05-temperature-TEMP-05-B-5s". Similarly, all physical quantities and their belonging sensors are constructed into a full-quantity structured mapping, which is output in a tabular manner. Each record is a unique identifier combination, with high indexability and engineering traceability. The structured mapping relationship table can also be exported as a CSV or database table for use by the field control system or data modeling.

[0108] Step S135: The structured mapping relationship table is represented in a graph structure, where the node represents the process section, the edge represents the binding path of the physical quantity, and the node attribute is the physical quantity set, generating a physical quantity-process section relationship table.

[0109] In this embodiment, the structured mapping relationship is further abstracted into a graph structure. In the graph construction process, the process section number is taken as the node in the graph, and the sensor mapping path of each type of physical quantity is taken as the edge in the graph. The direction of the edge is defined according to the material flow direction of the production line, and the node attribute is set as the physical quantity type set in the current section, such as “{temperature, pressure, pH}”. If there is a binding relationship of physical quantity flow between two process sections, such as the pH sensor data of a certain precipitation section simultaneously affecting the downstream neutralization section, a “SEC-02→SEC-03” edge is drawn in the graph, and “pH” is marked as an edge attribute. The physical quantity-process section relationship table constructed in this way presents a directed attribute graph structure, which has good visualization ability and cross-section data dependence expression ability, and can be used to track the data source and flow path in engineering analysis.

[0110] Optionally, the multi-step prediction vector extraction in step S2 is specifically:

[0111] The window length is set to T=60 seconds, the sliding step is 10 seconds, the process section field state factor set is time series sliced, and a time series fragment set is generated;

[0112] In this embodiment, before constructing the multi-step prediction input data, the collected process section field state factor set is cut according to a unified time window. The length of each time window is set to 60 seconds, and the sliding step is 10 seconds. The entire time series data set is traversed in a non-overlapping manner. Taking a temperature factor as an example, if its sampling frequency within one minute is 1 Hz, each window will contain 60 data points. By sliding 10 seconds forward from the starting point each time, the system forms multiple continuous but partially overlapping time periods, and the number of state factors retained in each fragment remains consistent, thereby ensuring the coverage continuity of the fragments on the time axis. The finally formed time series fragment set not only retains the complete time sequence characteristics, but also ensures sufficient context information support for subsequent trend extraction.

[0113] The time series fragment set is processed by multi-channel trend deconstruction to obtain a standardized state trend trajectory set;

[0114] In this embodiment, based on the previously generated time series fragment set, the data fluctuation trend inside each fragment is normalized and trend deconstructed by physical quantity channel. Each type of physical quantity such as temperature, pressure, conductivity, etc. may have amplitude differences in different time segments, and must first perform maximum-minimum value normalization to map each channel data uniformly to the interval [0, 1] to eliminate the influence of dimension and amplitude. Subsequently, local extreme point extraction and change slope calculation are performed on each channel, and the trend trajectory code is generated based on this structure. For example, after extracting 5 main rising sections and 4 stable sections in the temperature trend to form a "U5-S4" coding structure as the basis for the trend trajectory. The trend trajectories of all physical quantity channels are combined through time axis alignment, and finally a standardized state trend trajectory set is formed for the next stage.

[0115] Based on the standardized state trend trajectory set, multi-layer sliding convolution is performed to predict the trend prediction vector of the future 6 sliding steps, and an initial multi-step prediction sequence set is generated;

[0116] In this embodiment, the standardized trend trajectory set is subjected to continuous sliding window processing, and short-term dynamic patterns and periodic fluctuation features are extracted by performing moving convolution operation layer by layer, and these features are used to estimate the future trend direction. Based on the parameters of sliding step 10 seconds and prediction length 6 steps (i.e. future 60 seconds), the system expands the prediction process at the end position of each fragment. In the structure used, the convolution kernel length is set to 5, the convolution layer number is 3, and the step is 1, and the edge information is preserved during convolution to prevent trend truncation. At each sliding, the system outputs 6 groups of continuous trend estimation vectors, each containing the trend values of multiple physical quantity channels. All output vectors are combined in time sequence to obtain an initial multi-step prediction sequence set, reflecting the trend change direction and intensity of each physical quantity in the future period of time.

[0117] The initial multi-step prediction sequence set is subjected to inter- segment data correlation cross adjustment to eliminate low-confidence vectors to obtain a refined prediction vector set;

[0118] In this embodiment, after generating the preliminary prediction vector, cross- segment correlation analysis and confidence screening are performed. First, the prediction results between upstream and downstream segments are compared in units of process segments, and if the prediction results significantly deviate from the upstream and downstream trend coordination, the prediction vector is marked as a potential unreliable item. For example, if the pH is predicted to drop sharply in the precipitation segment and the downstream neutralization segment does not show the corresponding pH trend, the confidence of this vector is rated as less than the threshold value 0.6. All vectors with insufficient confidence are eliminated, and the remaining part enters the reordering stage, and the priority is redefined according to the trend continuity and intra-segment consistency, and finally a refined prediction vector set with uniform structure and clear direction is formed, providing a high-confidence data basis for subsequent packaging processing.

[0119] The set of refined prediction vectors is encapsulated in a unified format according to the process section number, physical quantity type, and time sequence, to obtain a multi-step prediction vector set.

[0120] In this embodiment, after confidence filtering and structure arrangement, the refined prediction results need to be encapsulated in a unified format for subsequent analysis and control calling. The encapsulation structure adopts a three-layer naming format, namely "process section number-physical quantity type-time label". The time label is encoded according to the prediction starting time and the prediction step length, such as "T+10s", "T+20s", etc., to ensure traceability. Taking the example vector "SEC-04-temperature-T+30s", the corresponding data record is "0.78", representing the normalized trend value of the physical quantity. All encapsulated data are stored in a dictionary structure, supporting calling by section or expanding by time sequence.

[0121] Optionally, the fine-tuning embedding time state transition in step S2 is specifically:

[0122] The time sequence similarity between the multi-step prediction vector set and the current state factor in the process section field state factor set is calculated, and the multi-step prediction vector set is re-weighted and fine-tuned according to the time sequence similarity to generate a re-tuned prediction vector group;

[0123] In this embodiment, in order to improve the adaptability of the prediction results to the actual state, the change characteristics of each group of multi-step prediction vectors and the current field state factor in the same time range are compared. The comparison window is set to 60 seconds, and the matching standards include the similarity of change slope, the similarity of fluctuation frequency, and the coincidence degree of extreme value position. Taking the temperature factor as an example, in a certain process section, its prediction vector shows a continuous rising trend, while the field state factor shows a slight fluctuation rising trend. Then, the similarity score is calculated according to the three indexes, and 0.85 is set as the re-weighting threshold. For the prediction vectors with a score higher than the threshold, a high weight is directly assigned; for the prediction vectors with a score lower than the threshold but not lower than 0.6, the correlation of adjacent physical quantities is referred to for supplementary weight correction. Finally, all prediction vectors are fine-tuned according to the corresponding weights to generate a re-tuned prediction vector group that is more suitable for the current state.

[0124] According to the process section structure table, the re-tuned prediction vector group is embedded in the upstream and downstream sections synchronously to generate a section-level time migration sequence;

[0125] In this embodiment, after completing the re-weighting fine-tuning, in order to establish a complete dynamic process expression, the re-tuned prediction vector groups of each section are sequentially integrated according to the upstream and downstream topological relationship defined in the process section structure table. In specific operation, the prediction vector at the prediction starting time point is first extracted from the upstream section, and is gradually embedded according to the time extension downstream, to ensure the continuity of the time label and the completeness of the physical quantity coverage. For example, if the precipitation section (section number SEC-03) and the filtration section (section number SEC-04) constitute a sequential relationship, the temperature, pH value, conductivity, etc. corresponding to "T+30s" in SEC-03 will be embedded into the prediction input of SEC-04 at "T+40s" in each group of predictions, to realize cross-section time transition feature fusion. Finally, each process section has a section-level time migration sequence containing the historical state of its upstream and downstream, which strengthens the contextual understanding ability of the future change trend.

[0126] The change rate, mutation frequency and steady state interval features in the section-level time migration sequence are extracted and encoded as state migration vectors.

[0127] In this embodiment, after obtaining the complete section-level time migration sequence, in order to further refine the core dynamic features, the physical quantity curves in each sequence are analyzed by multiple indicators. The change rate is represented by the value increase / decrease ratio per unit time, the mutation frequency is counted by the number of times that the value difference between two consecutive time points exceeds a threshold (such as 10%), and the steady state interval is defined as the length of the continuous time period with a change rate less than 0.01. Taking the conductivity of the filtration section as an example, if there are three mutations between "T+10s" and "T+60s", the average change rate is 0.03 / s, and the steady state duration is 30 seconds, then the corresponding index code is (R=0.03, J=3, S=30). All indicators are arranged in sequence according to the physical quantity dimension to form a state migration vector, which is an important input feature set for subsequent dimension reduction processing.

[0128] The state migration vector is compressed and mapped to output a low-dimensional control representation code;

[0129] In this embodiment, in order to facilitate the use of state transition information in the control process, the multi-dimensional state transition vector is projected into a fixed low-dimensional space. In the mapping process, a plurality of reference control samples are used to form a projection basis, and the compressed dimension is set to 8 dimensions, which ensures that the main trend information is retained and sufficient compression ratio is achieved. The reference control sample refers to a set of process state data with clear process label, stable state identification or known control effect selected from the historical lithium carbonate production process. These samples are used as a control or basis in the compression mapping process, and are used to establish the projection relationship of the state transition vector in the low-dimensional space. Taking the precipitation section as an example, the length of its state transition vector is 45 dimensions, and after mapping processing, it is converted into a control code sequence containing 8 floating point numbers, such as (0.12, 0.45, 0.03, …, 0.68). The encoding vector will be used as a simplified description of the current and predicted future process state, used for integrated analysis with other process section control data, to ensure control accuracy on the basis of information condensation.

[0130] The low-dimensional control representation codes of each section are time-sequentially spliced and labeled with physical action labels to obtain a multi-dimensional control representation flow.

[0131] In this embodiment, in order to complete the control input structure, the low-dimensional control codes of each process section are time-sequentially spliced to form a complete control representation flow. Each group of codes is prefixed with an identifier, including section number (such as SEC-05), physical action label (such as “separation”, “heating”), and time point label (such as “T+60s”), to ensure data traceability and application scalability. The splicing process is carried out according to the time axis to form a two-dimensional matrix structure, each row representing the control state of all process sections at a certain time point, and each column corresponding to a certain type of physical action dimension. The final output result can be used as input for digital twin simulation or control decision-making basis, forming the input core of the entire dynamic control system - multi-dimensional control representation flow.

[0132] Especially important is that the re-weighting fine-tuning fusion specifically includes:

[0133] Extract the current state factor sequence in the process section field state factor set, and set a sliding interval according to the time window length to construct a current state reference sequence set;

[0134] In this embodiment, in the on-site state factor extraction link, first, all state factor data in the time window of T=60 seconds from the current time point forward is selected from the process section on-site state factor set, and a sliding step of 10 seconds is set to form a sequence group of several overlapping time periods. Each sequence is indexed by time and contains multiple physical quantity channels such as temperature, pressure, flow rate, and conductivity to ensure complete process response dimensions. In the construction process, grouping is performed according to process section numbers to avoid cross interference of data from different sections, forming a current state reference sequence set with section characteristics, providing a structurally unified reference object for subsequent similarity measurement.

[0135] For each prediction vector in the multi-step prediction vector set, a comparison is made with the current state reference sequence set item by item, and a prediction-state similarity matrix is output.

[0136] In this embodiment, in the similarity calculation link, each sequence in the reference sequence set obtained in the previous step is sequentially aligned with the prediction trajectory from the multi-step prediction vector set, and the sum of the squared differences of each physical quantity dimension at the same time position is calculated, and finally a prediction-state similarity matrix is output. To control the interference degree of abnormal channels, the channel weight filtering threshold is set to 0.05, and low-contribution channels are masked. At the same time, to avoid matching misjudgment caused by too short prediction trajectories, only prediction vectors with a length of no less than the window T are retained to participate in the matching process, ensuring that the similarity matrix has a physical meaning of comparability.

[0137] According to the prediction-state similarity matrix, a similarity weighting factor is assigned to each prediction vector by combining a preset time offset weight function, generating a re-weighted coefficient vector.

[0138] In this embodiment, in the weighting coefficient construction process, the similarity score is extracted according to each row value of the prediction-state similarity matrix, and a group of predefined time offset decay weight functions are processed. The time offset weight function adopts a Gaussian decay form, the center is aligned with the current time point, the standard deviation is set to 15 seconds, and the influence degree of similarity away from the current time window is controlled. Finally, each prediction vector is assigned a corresponding re-weighting factor, and a re-weighted coefficient vector group with the same length as the prediction vector number is formed. The entire weight generation process maintains structural alignment while enhancing the time adaptability of the prediction value.

[0139] Apply the re-weighted coefficient vector to the multi-step prediction vector set to perform a weighted reconstruction operation, fine-tune and fuse the prediction vectors, and generate a re-adjusted prediction vector group.

[0140] In the weighting fine-tuning operation in this embodiment, the re-weighting coefficient vector group output in the previous stage is applied to the original multi-step prediction vector set, and weighted reconstruction is performed by multiplying each item according to the prediction time point, and then the multiple weighted prediction values in the same section and the same time position are normalized and averaged to eliminate abnormal disturbance. This operation ensures that the fine-tuned prediction result not only retains the original prediction trend, but also enhances its fit with the current process state. The finally generated re-tuned prediction vector group is output in a structured matrix form, maintaining the original channel dimension and time span, while improving its actual reference value in the process control scene.

[0141] Optionally, step S3 is specifically:

[0142] Step S31: connect the digital twin physics simulation platform; load the corresponding digital twin simulation model from the digital twin physics simulation platform according to the physical action label in the multi-dimensional control characteristic flow, and initialize the simulation input parameters to generate a simulation model configuration set;

[0143] In this embodiment, the lithium carbonate production line digital twin physics simulation environment deployed on the private industrial cloud platform is connected through the HTTPS protocol, the identity information is verified, and the related model resources are obtained. The multi-dimensional control characteristic flow contains multiple physical action labels, such as "heat exchange", "ion migration", "gas-liquid phase transfer", etc. These labels are extracted through pre-set label analysis rules. By analyzing each physical action label in the multi-dimensional control characteristic flow, the corresponding process characteristics are first identified, such as heat exchange is usually associated with reaction furnace, cooling system, etc. section, ion migration is associated with electrolytic cell, reaction kettle, etc. equipment. According to the nature and actual application of these physical actions, these labels are compared with the pre-set digital twin simulation model in the simulation platform. The registration information of the simulation model contains the physical action type, the process section applicable to the model, the parameter setting range, etc. Therefore, the system queries the model library of the simulation platform according to the label analysis result, and loads the simulation model consistent with the label. For example, if the label is "heat exchange", the reaction furnace and cooling system related heat conduction model will be queried and loaded; if the label is "ion migration", the electrolytic cell and the corresponding mass transfer model will be loaded. This process ensures that the loaded simulation model is consistent with the physical action label extracted from the multi-dimensional control characteristic flow, and can accurately reflect the control object required. Then, the simulation platform automatically initializes the input parameters of each simulation model according to the control target and the input parameter requirements of the simulation model. For example, for the "heat exchange" label, the platform will set the temperature, fluid rate, etc. parameters of the reaction furnace, and obtain their values from the control system. For the "ion migration" label, the platform loads the solution concentration, electrolysis voltage, etc. related data in the electrolytic cell as the initial input parameters. Through these preset and initialization operations, the simulation model can accurately simulate according to the real-time data, and provide reliable data support for subsequent control decisions.

[0144] Step S32: input the simulation model configuration set into the digital twin physical simulation platform, run the multi-zone parallel process response simulation, and obtain simulation response data;

[0145] In this embodiment, in the multi-zone process simulation stage, the simulation model configuration set generated in the previous step is input into the task scheduling interface of the simulation platform, and the parallel running module is activated. The simulation task is split by process zone, each zone is executed in an independent container, the simulation time span is set to 180 seconds, and the time step is 2 seconds, ensuring that the response details are fully recorded. The output parameters recorded by each simulation process include real-time reaction temperature change, material conversion rate, electrolytic cell voltage response, etc. The simulation platform uses preloaded physical equations and historical measured data for real-time calculation, and finally outputs a set of simulation response data covering all zones with time alignment characteristics, and packs it as a structured data set as the basis for subsequent analysis.

[0146] Step S33: perform control target index matching on the simulation response data, and perform control simulation performance evaluation according to the control target index, select the control input trajectory with the highest performance evaluation, and obtain the optimal control trajectory set;

[0147] In this embodiment, in the simulation result analysis link, the key index values related to the control target in each simulation response sequence are extracted, such as temperature control stability (temperature variance), target concentration achievement rate, voltage curve fluctuation rate, etc., and are matched with the specified control target interval. Then, from all simulation tasks, select the response result that meets the error tolerance range (±3%) in the control target index and has the highest comprehensive score, and the corresponding control input trajectory is considered as the optimal path. The control trajectory retains the change curve of the control variable with time, such as solution pump speed change, voltage gain adjustment curve, etc., forming an optimal control trajectory set containing zone number, time label, and control variable value, providing input basis for actual control execution.

[0148] Step S34: parameterize the control variables in the optimal control trajectory set, reorganize the format, and synchronize the time to generate a standard control instruction stream;

[0149] In this embodiment, for the selected optimal control trajectory, each control variable (such as pump speed, voltage, current intensity) is discretized, the unit precision is unified to two decimal places, and it is classified and packaged according to the zone number. All control points are aligned on the time line based on the simulation time step, ensuring that the control instructions can be recognized by the PLC device. Then, the parameter items in all control trajectories are reorganized into a fixed field format, such as [control variable type, target value, time stamp], and finally a standard control instruction stream with clear structure and convenient scheduling and execution is output. The instruction stream is used for subsequent automatic control system dispatching and execution.

[0150] Step S35: According to the execution relationship of the control instructions in the process configuration file of the lithium carbonate production line, the standard control instruction stream is controlled, the control channel is split, and the instruction concurrent mapping is performed, and a multi-channel control factor set is generated.

[0151] In this embodiment, in the control instruction issuing configuration link, the process configuration file of the lithium carbonate production line is read, and the execution channel information corresponding to each control instruction is extracted, such as the binding relationship with the PLC controller port, the synchronization identifier of the feedback sensor, etc. According to the target control point, the physical action type and the execution priority of the instruction, the standard control instruction stream is split into multiple control channels. For sections with concurrent control requirements, such as the stirring rate and heating temperature in the reaction kettle, the instructions will be sent in parallel within a minimum time difference of 0.1 seconds to form the final multi-channel control factor set. Each control factor clearly marks the channel number, execution delay time, instruction valid period and other fields to ensure accurate response of downstream equipment.

[0152] Optionally, the on-site load parameter recalculation and micro-period feedback modulation in step S4 are specifically:

[0153] The real-time lithium carbonate production line sensor data controlled by the production line and the actuator control log of the executed control instructions are received.

[0154] In this embodiment, after the production line control, the real-time sensor data stream and the actuator control system connected to the lithium carbonate production line are connected. Sensor data (such as temperature, pressure, concentration, etc.) from each process section of the production line and the control log of the actuator are received through the industrial protocol. The sensor data is transmitted in real time at a set frequency (for example, updated once per second), and the actuator control log contains the control instruction execution time, control parameters and actual execution results of the equipment. Data transmission is secured through encryption protocols to ensure its security and real-time performance, ensuring accurate reflection of the state of the production line.

[0155] According to the real-time lithium carbonate production line sensor data and the actuator control log, the execution feedback value of the corresponding control variable is extracted, and the actuator running state data set is obtained.

[0156] In this embodiment, after receiving the real-time sensor data and control logs, first, data synchronization is performed, and matching of various data is ensured through predetermined rules such as timestamp alignment. Then, logical discrimination of the execution of instructions in the control log is performed. For example, for the heating control system of the reaction kettle, if the control log shows that the "heating instruction has been executed", the actual feedback data executed by the actuator is extracted, such as the heating temperature change value, and is corrected to ensure that the feedback value corresponds to the actual change. Finally, according to these extracted feedback data, a state data set containing the running state of each actuator (such as motor running state, heating device temperature, etc.) is constructed.

[0157] The actuator running state data set is field-aligned with the multi-channel control factor set before the production line control, the response offset of each control channel is calculated, and leveling processing is performed to generate a load parameter recalculation matrix;

[0158] In this embodiment, the actuator running state data set is field-aligned with the multi-channel control factor set. The field of each control channel (such as pressure, temperature, flow) is matched with the corresponding historical control data. According to the control factor format of the multi-channel control factor set (for example, the unit of temperature value is Celsius, and the unit of flow is liter / minute), all data fields are standardized to ensure consistent data format. After data alignment, the actual feedback data of the actuator is compared with the expected target value in the optimal control trajectory set, and the response offset of each control channel is calculated. Specifically, the offset refers to the difference between the current feedback value and the target set value. For example, if the temperature of the reaction kettle is set to 90°C, and the actual feedback value is 88°C, the response offset is -2°C. Next, leveling processing is performed on these offsets to eliminate long-term deviations caused by equipment differences or operation errors. The leveling process uses a dynamic adjustment method based on a time window. For example, within a 15-minute time window, the temperature fluctuation range is observed, and the control variable is adjusted according to these fluctuations. The leveling process gradually reduces the offset through linear interpolation or proportional adjustment, and adjusts it to the vicinity of the target value. Finally, the adjustment parameters of each control channel are summarized to form a load parameter recalculation matrix.

[0159] Based on the load parameter recalculation matrix, and using the inertia compensation time window of each process section and the dynamic response characteristics of the equipment in the lithium carbonate production line process configuration file, the control perturbation in the preset micro-period feedback interval is extracted, and a feedback modulation factor group is generated;

[0160] In this embodiment, the load parameter recalculates the matrix, combined with the inertia compensation time window of each process section in the process flow configuration file and the dynamic response characteristics of the equipment, and extracts the preset micro-period feedback interval control perturbation. For example, if the equipment response of a certain process section (such as a concentration pool) is slow, a time window (for example, 30 seconds) will be set to predict the dynamic response of the equipment based on historical data, and the current control strategy will be adjusted combined with the feedback perturbation. For each control channel, these perturbations (such as pressure fine adjustment, temperature fine adjustment, etc.) will be extracted and form a feedback modulation factor group. These modulation factors will be used to accurately adjust various control factors in the production line to achieve precise control of the production process.

[0161] The feedback modulation factor group is matched with the process section structure table to output a control modulation response matrix.

[0162] In this embodiment, after obtaining the feedback modulation factor group, the modulation factors are mapped to the corresponding process section by referring to the process section structure table. For example, if a certain section (such as a reaction kettle) contains two control channels of temperature control and pressure control, according to the micro-perturbations in the feedback modulation factor group, they will be assigned to the corresponding control channels, and independent modulation response values will be generated for each channel. This process ensures that each control channel is properly adjusted to accurately fine-tune the production line. Finally, a control modulation response matrix is output, where each row represents the control response of a process section, and each column represents the corresponding feedback modulation factor.

[0163] Especially important is that the logical discrimination is specifically:

[0164] Extract the time sequence sampling time stamp, physical quantity type and corresponding value in the real-time lithium carbonate production line sensor data to generate a sampling state mapping table;

[0165] In this embodiment, each physical quantity type (for example: reaction kettle outlet temperature, sedimentation tank liquid level, pressure filtration system pressure) and its corresponding time sequence sampling time stamp and actual measurement value are extracted from the data stream collected by the lithium carbonate production line sensor. The data of each sensor type is sampled at a uniform sampling period (set to 2 seconds here) and is accompanied by ISO-8601 standard format timestamp information. In order to guarantee the uniformity of the data structure, all sampling values are converted to standard units (such as temperature unified to ℃, pressure unified to kPa), and abnormal sudden change data caused by sensor drift is removed. The extracted content generates a structured sampling state mapping table according to the fields of process section number, channel number, timestamp, physical quantity type and value, which is convenient for subsequent matching and analysis.

[0166] Extract the target set value, execution time point and channel number of each control channel in the actuator control log to generate a control instruction execution table;

[0167] In this embodiment, each control log from the executor control log is parsed to extract the control channel number, control target set value, instruction issuance time point, and corresponding control type (such as speed control, temperature setting, liquid inlet valve opening, etc.). The control log is uploaded by the control master station through the industrial Ethernet protocol, and all records carry high-precision (less than 1 millisecond error) time tags to ensure time synchronization with sensor data. Each control instruction is encoded in the format of "channel number-time point-set value" and is classified into the control instruction execution table. The table is indexed by the logical order of control instructions to establish a clear data structure and lay a data foundation for subsequent executor response tracking and control closed-loop feedback.

[0168] According to the sampling state mapping table and the control instruction execution table, the channel number is taken as the key value, the execution time point is taken as the matching window, and alignment matching is performed within the preset maximum response time delay range to obtain an initial control feedback pair;

[0169] In this embodiment, the sampling state mapping table and the control instruction execution table are combined, the control channel number is taken as the index key, and the maximum response time delay is set to 20 seconds as the time matching window. From the execution time of each control instruction, the continuous sampling records of the channel within 20 seconds after the time point are found, and the data in which the first change rate reaches the set threshold (such as the temperature change rate being greater than 0.2°C / s) is selected as the response point to form an initial control feedback pair with the instruction. If there is no effective response within 20 seconds, it is recorded as an empty pair. This alignment process strictly ensures that the response sampling time is not earlier than the control instruction time, and ensures that the feedback is indeed caused by the control action.

[0170] Logical consistency discrimination is performed on the initial control feedback pair to eliminate logically abnormal pairs to obtain an effective control feedback pair;

[0171] In this embodiment, logical consistency verification is performed on the formed initial control feedback pair to eliminate logically abnormal data pairs. Specifically, the following rules are included: if the control type is temperature rise and the feedback value continuously decreases within the response time window, it is marked as reverse response; if the physical quantity unit of the response sampling point is inconsistent or NaN or null, it is marked as data abnormality; if the control instruction set value changes less than the set tolerance (such as less than 0.1%) and there is no obvious physical quantity change, it is considered as invalid response. All the above marked items are eliminated from the feedback pair, and the remaining part is reserved as an effective control feedback pair. The logical consistency discrimination also includes synchronous judgment of the process section state, for example, the reaction kettle should feedback the temperature rise effectively in the "reaction" state, otherwise it is eliminated.

[0172] Actual response values of each control variable are extracted from the effective control feedback pair, classified and aggregated according to the control channel, and an executor running state data set is output.

[0173] In this embodiment, the actual response value of each control variable is extracted from the verified effective control feedback pair. The feedback values under the same channel are aggregated and counted in time sequence with the control channel as the aggregation primary key, to calculate the average response level, the maximum response amplitude and the time required to actually reach the set value (i.e. the response delay). For example, for the speed control channel of the stirring motor, the feedback speed sequence is extracted and the change trend is calculated to obtain key indicators such as the start and stable time of the equipment after receiving the instruction. Finally, the actuator running state data set is constructed, which is stored according to the channel number, time sequence, control variable type, response value and response state for subsequent load adjustment and feedback modulation calculation.

[0174] Optionally, step S5 is specifically:

[0175] Step S51: According to the response offset of each control channel in the control modulation response matrix and the set of on-site state factors of the process section, the actual state change amplitude is calculated, the real-time control response offset value of the process section is calibrated, and thus the process control response offset mapping table is obtained;

[0176] In this embodiment, the response offset of each control channel is extracted from the control modulation response matrix, combined with the process section number and the corresponding physical quantity set mapped in the established "physical quantity-process section relationship table", and the actual state change amplitude is calculated channel by channel by comparing the currently collected on-site state factor data with the standard fluctuation interval of the process section on-site state factor set in the past 24 hours. Taking the drying section temperature channel TEMP-05-A as an example, if the standard fluctuation is ±0.8℃, and the current temperature offset is +2.1℃, the response offset of this channel in this process section is "high temperature deviation threshold range". The offset result is bound to the section (SEC-05) and recorded in the real-time control response offset table, to construct a unified format process control response offset mapping table, which includes channel ID, process section number, physical quantity type, offset value and calibration label.

[0177] Step S52: Comparing the process control response offset mapping table with the physical quantity-process section relationship table, locating the process section and associated physical quantity where the offset occurs according to the process section number, and calculating the offset cause weight, so as to obtain the process offset associated factor group;

[0178] In this embodiment, the "process control response offset mapping table" and the "physical quantity-process section relationship table" are compared at the field level, focusing on matching the physical quantity type and the process section number, to find all the process sections where the offset occurs and the associated physical quantities. After matching, the cause weight of the influence factor is calculated for each offset record, which is based on the frequency of similar offset events in the historical data, the stability coefficient of the associated physical quantity and the calibration state of the on-site sensor. The specific form is: Among them W i F represents the weight value (range 0 to 1) for the i-th offset cause; i F represents the average daily frequency of historical offset events causing the i-th offset (in times per day); max The highest offset frequency among all causes within the period (used for normalization); S i C represents the stability coefficient (range 0-1, higher values ​​indicate greater stability) of the physical quantity corresponding to the i-th offset cause; i C represents the actual calibration cycle (in days) for the device related to the i-th offset cause; max The maximum allowable calibration period (for normalization) is defined as follows: α, β, and γ are the weighting coefficients of each influencing factor, with default values ​​of α = 0.4, β = 0.3, and γ = 0.3. For example, when a pH value shifts in a neutralization stage (SEC-03), it is compared with historical pH-related shift events. A comprehensive score is assigned based on the following conditions: occurrence frequency of 6 times / day, highest shift frequency among all causes within the period of 10 times / day, stability coefficient of 0.82, calibration period of 9 days, and maximum allowable calibration period of 14 days. The final comprehensive weight of this shift cause is 0.487, allowing for horizontal comparison with other causes and the construction of a "process shift correlation factor group." Statistical analysis reveals that this cause mainly originates from both sensor drift and raw material fluctuations. Each cause is assigned a corresponding weight, ultimately forming a structured "process shift correlation factor group."

[0179] Step S53: Based on the process offset correlation factor group, perform reverse path tracing in the graph structure of the physical quantity-process section relationship table to obtain the offset influence path;

[0180] In this embodiment, based on the generated "process offset correlation factor group," the source of offset influence is retrieved in reverse within the graph structure of the "physical quantity-process segment relationship table." Each edge in the graph structure represents a physical quantity flow path, and nodes represent process segments. Input paths with the same physical quantity label are traced upstream from the node where the offset occurred. For example, for a pH offset in the neutralization segment, if its upstream is the precipitation segment (SEC-02), and there exists a connecting path with the edge attribute "pH," the data dependency source will be traced back along that path. The maximum tracing depth is set to 3 segments, and the path reliability is scored based on timestamp and physical quantity synchronization. Finally, an offset influence path from SEC-02 to SEC-03 is output, and its path node sequence, edge attributes, and time window matching degree are recorded for subsequent structural analysis.

[0181] Step S54: Perform structural adjustments on the offset influence path and generate a reverse adjustment index table for the process diagram;

[0182] In this embodiment, the identified offset influence path is structurally adjusted, that is, the association of nodes or the path strength of edges in the graph structure is adjusted, to generate a "process map reverse adjustment index table". In the implementation process, first, compare the current physical quantity fluctuation state of all nodes on the path with the historical stable value interval. If the state fluctuation of a node exceeds 3 times the standard deviation, the priority of the node in the index table is raised, and it is marked as "first-level adjustment candidate". Then, according to the physical quantity type in the edge attribute, the edges with connection strength lower than the threshold (such as signal correlation coefficient <0.6) are down-weighted or shielded. The finally formed reverse adjustment index table structure includes: priority adjustment section number, corresponding physical quantity, adjustment strength level and recommended processing action (such as sensor re-inspection, control logic delay adjustment, etc.), which is used to guide the operation and maintenance or automatic control system to respond quickly.

[0183] Step S55: Apply the process map reverse adjustment index table to the process section field state factor set, update the structured binding relationship, and mark the abnormal response section to generate the process exception feedback body.

[0184] In this embodiment, the generated "process map reverse adjustment index table" is applied to the current collected "process section field state factor set" one by one. The application process includes updating the binding relationship in the original state factor data structure to replace or strengthen the dependence path between the offset physical quantity and the process section. For example, if the original binding path of a process section points to three physical quantity nodes, after adjustment by the index table, only two main physical quantity paths are retained, and the rest are marked as abnormal data input. The updated binding structure will be visualized in a graph structure, and the abnormal response section will be highlighted in the graph (such as highlighted edges, red nodes), and a "process exception feedback body" will be generated, including the updated binding structure, the retained main physical quantity path, the abnormal process section number, the abnormal physical quantity, the response strength score, etc.

[0185] Optionally, step S54 is specifically:

[0186] Step S541: Extract the control association degree, process action strength and conduction direction of each physical quantity node in the offset influence path, and construct an offset path structure factor matrix;

[0187] In this embodiment, first, according to the previously obtained physical quantity-process section relationship diagram, all physical quantity nodes appearing in the offset influence path are extracted. For each node, relying on the control linkage configuration table built in the field control system, the control correlation parameter (set as a decimal between 0 and 1, such as “0.32” indicating medium correlation) is extracted; at the same time, combined with the process configuration file of the process section, the process action intensity corresponding to the node is read (for example, the action coefficient of temperature control in the drying section is “0.58”, and the action coefficient of pH control in the precipitation section is “0.81”). In addition, according to the installation sequence of the physical sensor on site and the material flow direction, the offset conduction direction information is extracted from the process map and coded as “→” or “←” two directional indicators. The above three information is combined in structure, filled into a two-dimensional matrix structure according to the node order, and each row of the matrix represents a node of the offset path, and each column is “control correlation”, “process action intensity” and “conduction direction code”, respectively, to form the offset path structure factor matrix. This matrix is used to identify which sections in the path may play a key role in offset propagation in the subsequent.

[0188] Step S542: According to the offset path structure factor matrix, the influence intensity of the control response offset on each physical quantity node is analyzed, and the path sections with influence intensity exceeding [0.15, 0.25] are marked to obtain a path abnormal section identification set;

[0189] In this embodiment, the generated offset path structure factor matrix is called to analyze the influence intensity of each path node. First, the control correlation and process action intensity of each node are weighted and averaged, and the influence intensity is defined as 0.6x control correlation + 0.4x process action intensity, and the obtained intensity value is compared with the set influence threshold interval [0.15, 0.25]. If the intensity value of the path section corresponding to the node is within the interval, it means that it has a medium offset diffusion risk, so it is marked as “potential abnormal section”. For example, if the temperature node of a certain drying section has a control correlation of 0.20 and an action intensity of 0.35, its influence intensity is 0.6x0.20+0.4x0.35=0.26, which exceeds the upper limit and is marked as “abnormally high risk section”. In this way, all sections in the offset path are traversed, and all sections that meet the conditions are extracted, and finally the “path abnormal section identification set” is output in the form of a structure table, and the record fields include section number, physical quantity name, influence intensity value and marking label (such as “high risk” and “medium risk”).

[0190] Step S543: Perform node replacement, weight reconstruction and direction optimization on the path sections in the path abnormal section identification set, and record the structure change operation to obtain the path structure change record;

[0191] After identifying the "path abnormal section identification set", the embodiment performs structure reconstruction operation on the path sections involved in it one by one. First, for the physical quantity nodes in each abnormal path section, the adjacent node with the weakest correlation strength is selected as the replacement node, and the node replacement operation is performed. For example, in SEC-04, if the pH node has too high influence strength, the conductivity node with a strength below 0.05 near the pH node will be automatically selected as the replacement, and the conduction link will be updated. Secondly, after completing the node replacement, the weight reconstruction operation is performed on the values of all edges on the path section, that is, the control weight is recalculated, the new weight is generated in the form of "new node control correlation degree x action strength", and the old value in the original path is replaced. Finally, for the path section with direction contradiction (i.e., the original direction is "→", but the upstream and downstream section numbers are changed to form "←"), the direction adjustment is performed, the connection direction in the graph structure is updated, and the conduction vector in the graph data structure is rewritten. Each replacement, reconstruction or direction adjustment operation is recorded, and the original section number, modification time, change content, executor identification and other fields are marked to form a path structure change record, which serves as a basis for subsequent backtracking or reverse derivation.

[0192] Step S544: Indexing and hierarchical coding the path structure change record according to the process section number and physical quantity identifier to obtain a process map reverse adjustment index table.

[0193] In the embodiment, the path structure change record generated in the foregoing steps is indexed for subsequent reverse modification operation of the process map. The specific method is as follows: first, the process section number and physical quantity identifier in each change record are indexed and coded in two fields, and "section number + physical quantity identifier" is set as the first index key, for example, "SEC-04-pH" is generated for SEC-04 and "pH"; then, the records are sorted according to the change time sequence, and a time stamp and a change serial number are added as the second index. Secondly, in order to ensure the traceability of the multi-level mapping, the type of the change operation (such as "node replacement" and "direction adjustment") is taken as an index classification label, and a "operation type-process section-physical quantity" three-level index structure is established. Finally, the system exports the index structure as a reverse adjustment index table, and the table fields include index key, operation type, original path information, new structure description, execution time, responsible person identification and other contents, which supports JSON and database table dual-format output. The table can be used by the upstream map management engine to dynamically modify the process section control logic and visually render the path graph.

[0194] Alternatively, the present specification also provides an automatic control system for producing lithium carbonate for executing the automatic control method for producing lithium carbonate as described above, and the automatic control system for producing lithium carbonate comprises:

[0195] The on-site state identification module is configured to acquire lithium carbonate production line sensing data, and perform process section physical quantity structured binding on the lithium carbonate production line sensing data to obtain a process section on-site state factor set.

[0196] The multi-step prediction module is configured to perform multi-step prediction vector extraction according to the process section on-site state factor set, and perform fine-tuning embedded time state migration on the multi-step prediction vector to obtain a multi-dimensional control representation flow.

[0197] The control simulation module is configured to perform digital twin control simulation based on the multi-dimensional control representation flow, and perform control instruction coding on a digital twin control simulation result to obtain a multi-channel control factor set.

[0198] The production line control module is configured to transmit the multi-channel control factor set to a production line management platform to perform a production line control task, and perform on-site load parameter recalculation and micro-period feedback modulation based on acquired real-time lithium carbonate production line sensing data to obtain a control modulation response matrix.

[0199] The process offset optimization module is configured to perform process offset tracing based on the control modulation response matrix, and perform process map structure reverse adjustment on the process section on-site state factor set according to a process offset tracing result to obtain a process exception feedback body.

[0200] The iterative parameter adjustment module is configured to take the process exception feedback body as an input of a new round of the multi-step prediction module, and repeatedly iterate the multi-step prediction module to the process offset optimization module until a process offset amount is less than a preset process offset amount threshold.

[0201] Therefore, from any viewpoint, the embodiments should be considered as exemplary and non-limiting, the scope of the present application being defined by the appended claims and not by the above description, and it is intended to embrace all variations falling within the meaning and range of equivalents of the elements of the claims.

[0202] The above description is merely one specific implementation of the application, which enables a person skilled in the art to understand or implement the application. Various modifications to these embodiments will be apparent to those skilled in the art, and the general principles defined herein can be implemented in other embodiments without departing from the spirit or scope of the application. Therefore, the application shall not be limited to these embodiments shown herein, but shall conform to the broadest scope consistent with the principles and novel features disclosed herein.

Claims

1. An automatic control method of producing lithium carbonate, characterized by, The method comprises the following steps: Step S1: Obtain lithium carbonate production line sensor data, and perform process section physical quantity structure binding on the lithium carbonate production line sensor data to obtain a process section field state factor set; Step S2: Perform multi-step prediction vector extraction according to the process section field state factor set, and perform fine-tuning embedded time state migration on the multi-step prediction vector to obtain a multi-dimensional control representation flow; the fine-tuning embedded time state migration in step S2 is specifically as follows: Calculate the time sequence similarity between the multi-step prediction vector set and the current state factor in the process section field state factor set, and re-weight and fine-tune the multi-step prediction vector set according to the time sequence similarity to generate a re-tuned prediction vector group; According to the process section structure table, the re-tuned prediction vector group is synchronously embedded in the upstream and downstream sections to generate a section-level time migration sequence; Extract the change rate, mutation frequency and steady-state interval characteristics in the section-level time migration sequence, and encode them into a state migration vector; Compress and map the state migration vector to output a low-dimensional control representation code; The low-dimensional control representation codes of each section are sequentially spliced, and a physical action label is marked to obtain a multi-dimensional control representation flow; Step S3: Perform digital twin control simulation based on the multi-dimensional control representation flow, and encode the digital twin control simulation result to obtain a multi-channel control factor set; step S3 is specifically as follows: Step S31: Connect a digital twin physical simulation platform; load a corresponding digital twin simulation model from the digital twin physical simulation platform according to the physical action label in the multi-dimensional control representation flow, and initialize simulation input parameters to generate a simulation model configuration set; Step S32: Input the simulation model configuration set into the digital twin physical simulation platform, run multi-section parallel process response simulation, and obtain simulation response data; Step S33: Perform control target index matching on the simulation response data, and perform control simulation performance evaluation according to the control target index to select a control input trajectory with the highest performance evaluation to obtain an optimal control trajectory set; Step S34: Discretize the control variables in the optimal control trajectory set, reorganize the format, and synchronize the time to generate a standard control instruction stream; Step S35: According to the control instruction execution relationship in the lithium carbonate production line process flow configuration file, perform control channel splitting and instruction concurrent mapping on the standard control instruction stream to generate a multi-channel control factor set; Step S4: Transmit the multi-channel control factor set to a production line management platform to perform a production line control task; perform field load parameter recalculation and micro-period feedback modulation on the obtained real-time lithium carbonate production line sensor data to obtain a control modulation response matrix; the field load parameter recalculation and micro-period feedback modulation in step S4 are specifically as follows: Receive real-time lithium carbonate production line sensor data and actuator control logs of executed control instructions after production line control; According to the real-time lithium carbonate production line sensor data and the actuator control logs, perform logical discrimination, extract feedback values of corresponding control variables, and obtain an actuator running state data set; The actuator running state dataset is field-aligned with the multi-channel control factor set before the production line control, the response offset of each control channel is calculated, and leveling processing is performed to generate a load parameter recalculation matrix; Based on the load parameter recalculation matrix, and using the inertia compensation time window of each process section in the lithium carbonate production line process flow configuration file and the equipment dynamic response characteristics, the control perturbation in the preset micro-period feedback interval is extracted to generate a feedback modulation factor group; The feedback modulation factor group is mapped and matched with the process section structure table to output a control modulation response matrix; Step S5: Based on the control modulation response matrix, process offset tracing is performed, and the process section field state factor set is adjusted in reverse according to the process offset tracing result to obtain a process abnormality feedback body; Step S6: The process abnormality feedback body is taken as the input of a new round of step S2, and steps S2 to S5 are repeated and iterated until the process offset is less than the preset process offset threshold.

2. The automatic control method of producing lithium carbonate according to claim 1, characterized by, Step S1 specifically comprises: Step S11: Collecting physical sensing data of each process section of the lithium carbonate production line, and collecting the data according to a preset sampling period to obtain lithium carbonate production line sensing data; Step S12: Data cleaning and time sequence alignment are performed on the lithium carbonate production line sensing data, missing values and abnormal points are removed, and uniform time step completion is performed to generate a preprocessed dataset; Step S13: Physical quantity-process section mapping binding is performed according to the preprocessed dataset to obtain a physical quantity-process section relationship table; Step S14: The preprocessed dataset is classified and integrated based on the physical quantity-process section relationship table, and a structured set is output according to the process section; Step S15: The structured set is converted into a state factor vector, sequentially spliced according to the data collection time, and thus the process section field state factor set is composed.

3. The automatic control method of producing lithium carbonate according to claim 2, characterized by, Step S13 specifically comprises: Step S131: Extracting the sensor ID, sensor installation position number and data collection time corresponding to each type of physical quantity in the preprocessed dataset to construct a basic metadata index table; Step S132: Connecting the production line management platform, loading the lithium carbonate production line process flow configuration file, extracting the spatial position range, physical action label and corresponding process section number of each process section, and constructing a process section structure table; Step S133: Comparing the basic metadata index table with the process section structure table, establishing a mapping relationship according to the sensor installation position number and the spatial position range of the process section, and marking the corresponding process section number of each physical quantity to obtain a physical quantity-process section mapping relationship; Step S134: Structurally expressing the physical quantity-process section mapping relationship according to a preset process section number-physical quantity type-sensor identifier-data collection frequency framework to form a structured mapping relationship table; Step S135: Representing the structured mapping relationship table in a graph structure, wherein the nodes represent the process sections, the edges represent the binding paths of the physical quantities, the node attributes are the physical quantity sets, and a physical quantity-process section relationship table is generated.

4. The automatic control method of producing lithium carbonate according to claim 2, characterized by, The multi-step prediction vector extraction in step S2 specifically comprises: The window length is set as T=60 seconds, the sliding step is 10 seconds, the process section field state factor set is time-sliced, and a time series fragment set is generated; The time series fragment set is processed by multi-channel trend deconstruction to obtain a standardized state trend trajectory set; Based on the standardized state trend trajectory set, multi-layer sliding convolution is performed to predict a trend prediction vector for the next 6 sliding steps, and an initial multi-step prediction sequence set is generated; The initial multi-step prediction sequence set is cross-adjusted for inter-zone data correlation to eliminate low-confidence vectors, and a refined prediction vector set is obtained; According to the formatting framework of process section number, physical quantity type and time series, the refined prediction vector set is formatted and encapsulated to obtain a multi-step prediction vector set.

5. The automatic control method of producing lithium carbonate according to claim 1, characterized by, Step S5 is specifically: Step S51: According to the response offset of each control channel in the control modulation response matrix and the process section field state factor set, the actual state change amplitude is calculated correspondingly, the real-time control response offset value of the process section is calibrated, and a process control response offset mapping table is obtained; Step S52: The process control response offset mapping table is compared with the physical quantity-process section relationship table, the process section and the associated physical quantity where the offset occurs are located according to the process section number, and the offset cause weight is calculated, so as to obtain a process offset associated factor group; Step S53: According to the process offset associated factor group, the path reverse tracking is performed in the graph structure of the physical quantity-process section relationship table, and an offset influence path is obtained; Step S54: The offset influence path is adjusted in structure to generate a process graph reverse adjustment index table; Step S55: The process graph reverse adjustment index table is applied to the process section field state factor set to update the structured binding relationship and mark the abnormal response section, and a process abnormality feedback body is generated.

6. The automatic control method of producing lithium carbonate according to claim 5, characterized by, Step S54 is specifically: Step S541: The control correlation degree, process action intensity and conduction direction of each physical quantity node in the offset influence path are extracted to construct an offset path structure factor matrix; Step S542: The influence intensity of the control response offset on each physical quantity node is analyzed according to the offset path structure factor matrix, and the path segments with influence intensity exceeding [0.15, 0.25] are marked to obtain a path abnormal segment identification set; Step S543: The path segments in the path abnormal segment identification set are executed for node replacement, weight reconstruction and direction optimization, and the structure change operation is recorded to obtain a path structure change record; Step S544: The path structure change record is indexed and layered encoded according to the process section number and the physical quantity identification to obtain a process graph reverse adjustment index table.

7. An automatic control system for producing lithium carbonate, characterized by, An automatic control system for producing lithium carbonate is used to perform the automatic control method for producing lithium carbonate as claimed in claim 1, comprising: A field state identification module is configured to acquire lithium carbonate production line sensing data, and to perform process section physical quantity structured binding on the lithium carbonate production line sensing data to obtain a process section field state factor set; A multi-step prediction module is configured to extract a multi-step prediction vector according to the process section field state factor set, and to perform fine-tuning and embedding time state migration on the multi-step prediction vector to obtain a multi-dimensional control representation flow; The control simulation module is configured to perform digital twin control simulation based on the multi-dimensional control representation flow, encode control instructions based on a simulation result of the digital twin control simulation, and obtain a set of multi-channel control factors. The production line control module is configured to transmit the set of multi-channel control factors to a production line management platform to perform a production line control task, re-calculate a field load parameter and perform micro-period feedback modulation based on acquired real-time lithium carbonate production line sensing data, and obtain a control modulation response matrix. The process offset optimization module is configured to perform process offset tracing based on the control modulation response matrix, and perform reverse adjustment of a process map structure based on a set of field state factors of a process section according to a result of the process offset tracing, and obtain a process abnormality feedback body. The iterative parameter adjustment module is configured to take the process abnormality feedback body as an input of a new round of the multi-step prediction module, and repeatedly iterate the multi-step prediction module to the process offset optimization module until a process offset amount is less than a preset process offset amount threshold.

Images