A data fusion-based real-time optimization method and system for three-acid purification process parameters

Abstract

The application discloses a kind of real-time optimization method and system of three-acid purification process parameters based on data fusion, belong to phosphating intelligent automatic control field.For surface filtration, plate and frame independent control, data lag, no filter material aging self-adaptation, multi-objective imbalance, static safety threshold and other industry defects, four innovation systems of entropy weight multi-source fusion engine, four-layer pre-safety optimization, three-level time sequence interlocking negative feedback and slurry conservation decoupling are constructed.Fusion slurry pH, aluminum and magnesium ion working condition data, segmented interpolation is combined with entropy weight objective weighting, permanent safety hard red line and dynamic adjustable threshold are divided, double-variable filter material attenuation, differential pressure layered blockage identification are matched, DCS lightweight bottom and upstream interlock linkage are equipped.Adapt to four parallel surface filtration + two-stage plate and frame coupled phosphoric acid three-acid purification section, full-condition self-adaptive control, significantly reduce the clear liquid overproof, equipment overload, reduce water and electricity filter material loss, industrial operation stability is excellent.

Description

Technical Field

[0001] This invention relates to the field of industrial automation control technology, specifically to a method and system for real-time optimization of process parameters for the purification of three acids (phosphoric acid, sulfuric acid, and nitric acid) in chemical production processes. Background Technology

[0002] In the phosphorus chemical and inorganic acid production industries, "three-acid purification" is a process that typically includes two main steps: surface filtration (such as rotary table filter) and plate and frame filter press.

[0003] Existing control technologies have the following significant drawbacks: 1. Severe control island phenomenon: Existing technologies typically treat surface filtration systems and plate and frame filter press systems as two independent control units. Operators often adjust the feed only based on the load of the surface filter, neglecting the processing capacity of the plate and frame filter press, resulting in a mismatch between upstream and downstream loads and frequent "stocking overload" or "feedout" phenomena.

[0004] 2. Data lag: Traditional DCS (Distributed Control System) can only process real-time data such as pressure and flow rate. However, key indicators that determine product quality (such as the solid content of the clarified liquid and the moisture content of the filter cake) usually rely on manual analysis in the laboratory, resulting in a data lag of 2-4 hours. By the time quality abnormalities are detected, a large number of defective products have already been produced.

[0005] 3. Lack of consideration for the entire equipment life cycle: Existing control programs are mostly based on fixed logic (such as fixed pressing time and fixed backwashing cycle). As filter cloths and membranes are used, their water permeability and interception performance will gradually decline (equipment deterioration). Fixed programs cannot automatically compensate for this decline, resulting in inflated energy consumption or reduced filtration effect.

[0006] Therefore, there is an urgent need for a real-time optimization control method that can integrate multi-source data, resolve upstream and downstream coupling conflicts, and adapt to equipment aging. Summary of the Invention

[0007] To address the shortcomings of the existing technologies, this invention proposes a real-time optimization method and system for tri-acid purification process parameters based on data fusion.

[0008] The technical solution adopted in this invention is as follows: In one implementation of the first aspect, this application proposes a real-time optimization method for tri-acid purification process parameters based on data fusion, applied to a tri-acid purification device including a surface filtration system and a plate and frame filter press system. The method includes the following steps: Acquire multi-source heterogeneous data of the entire process of tri-acid purification, including DCS time series data, offline test data, equipment degradation data and common media data; The multi-source heterogeneous data is spatiotemporally aligned and hierarchically weighted and fused to generate a fused working condition matrix; The fusion working condition matrix is ​​input into a preset four-layer hierarchical progressive collaborative optimization model for calculation to obtain a set of candidate process parameters. The four-layer hierarchical progressive collaborative optimization model includes a hard constraint layer for equipment safety, a constraint layer for clear liquid quality, a sludge treatment efficiency layer, and a water and electricity energy consumption optimization layer arranged in order of priority from high to low. The lower layer constraint serves as the insurmountable boundary for the upper layer optimization. The optimal control command is generated based on the calculation results and sent to the DCS system to adjust the action parameters of the field actuators.

[0009] In one implementation of the second aspect, this application proposes a real-time optimization system for the parameters of a three-acid purification process based on data fusion, comprising: The data fusion module is used to perform the data acquisition and fusion steps in the method provided by the first aspect or any possible implementation of the first aspect; The collaborative optimization calculation module is used to execute the four-level hierarchical progressive calculation steps in the method provided by the first aspect or any possible implementation of the first aspect; The instruction issuing module is used to convert the calculation results into control signals that the DCS can recognize; The data fusion module, collaborative optimization calculation module, and instruction issuance module are connected through an industrial communication network.

[0010] Thirdly, this application provides an electronic device, comprising: A processor, a memory, and a communication bus, wherein the processor and the memory communicate with each other via the communication bus; The memory stores program instructions that can be executed by the processor, which can invoke the program instructions to perform the method provided by the first aspect or any possible implementation thereof.

[0011] Fourthly, this application provides a computer-readable storage medium, characterized in that it stores computer program instructions thereon, wherein the computer program instructions, when executed by a processor, implement the method provided by the first aspect or any possible implementation of the first aspect.

[0012] Fifthly, this application provides a computer program product, The computer program product includes a computer program that, when executed by a processor, implements the method provided by the first aspect or any possible implementation of the first aspect.

[0013] In summary, due to the adoption of the above technical solution, the beneficial effects of the present invention are: This invention addresses the phosphoric acid triacid purification process in a phosphate chemical plant, which uses four parallel surface filters coupled with two-stage plate and frame filters. It establishes four major systems: entropy-weighted multi-source data fusion, four-layer hierarchical safety-priority optimization, three-level time-series negative feedback, and slurry conservation decoupling. It also adds adaptive identification of filter media attenuation, frequency-based distribution of fast and slow variables, and a DCS local fallback interlocking mechanism. Unlike existing technologies that suffer from independent control of two filtration systems, 2-4 hour lag in laboratory data, inability to adapt fixed parameters to filter media aging, statically fixed safety thresholds, conflicting optimization objectives, and significant load impacts on upstream and downstream processes, this invention aligns real-time DCS data with offline laboratory data and dynamically assigns weights based on information entropy. It strictly follows progressive constraints based on safety, quality, efficiency, and energy consumption for optimization. Relying on three-level closed-loop feedback to dynamically adjust process parameters as filter media deteriorates, and using the material conservation equation to decouple the load of series processes, it achieves adaptive intelligent control under all operating conditions. This effectively reduces issues such as excessive slurry levels and equipment overload, lowers water, electricity, and filter media consumption, and improves the long-term operational stability and overall profitability of the triacid purification process. Attached Figure Description

[0014] The present invention will be described by way of example and with reference to the accompanying drawings, wherein: Figure 1 This is a flowchart of the method in this invention; Figure 2 This is a block diagram of the system architecture in this invention; Figure 3 This is a block diagram of the hardware composition of the electronic device of the present invention. Detailed Implementation

[0015] To make the objectives, technical solutions, and advantages of the embodiments of this application clearer, the technical solutions of the embodiments of this application will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only a part of the embodiments of this application, and not all of the embodiments. The components of the embodiments of this application described and shown in the accompanying drawings can be arranged and designed in various different configurations. Therefore, the following detailed description of the embodiments of this application provided in the accompanying drawings is not intended to limit the scope of the claimed application, but merely represents selected embodiments of this application. All other embodiments obtained by those skilled in the art based on the embodiments of this application without inventive effort are within the scope of protection of this application.

[0016] Example 1 This embodiment provides a real-time parameter optimization method for triacid purification process based on data fusion, which is applied to the phosphoric acid-associated triacid purification workshop section. The standardized hardware configuration of this workshop section includes 4 parallel JY-120NBP large surface filters and two sets of XMZGFQ300 two-stage plate-and-frame integrated unit with two pressing and one washing processes. The entire production link completely covers the fully coupled continuous process flow of slurry feeding from the clarification tank, PPA raffinate acid feeding, surface filtration separation, primary pressing dewatering by plate-and-frame, primary size mixing homogenization, secondary deep pressing, secondary fine size mixing, and closed conveying of slag slurry to the gypsum repulping tank. See Figure 1 , the complete execution steps of the whole optimization control method are as follows: Step S1: Collect multi-source heterogeneous data of the whole triacid purification process in the whole domain. The multi-source heterogeneous data specifically includes seven complete categories: second-level high-frequency time-series operation data of DCS system, manual sampling offline laboratory test data every 2 hours, basic parameters of slurry raw material (collect influencing indexes such as slurry pH value, concentration of aluminum, magnesium and heavy metal impurity ions synchronously), degradation and loss data of filter cloth / membrane in the whole life cycle of filter material, metering data of consumption of classified public media such as water, electricity and compressed air, fault interlock trigger data such as overpressure and overload of equipment, pump failure, and working condition label classification data distinguishing normal / high-load / maintenance states; all data are collected uniformly through industrial isolation gateway to eliminate mutual interference of signals from different data sources.

[0017] Step S2: Perform segmented spatio-temporal coordinate alignment, entropy weight adaptive layered weighted fusion, intelligent cleaning of abnormal dirty data, process logic linkage verification of manual test values, and three-level hierarchical data fault-tolerant compensation processing on the collected multi-source heterogeneous data in sequence. After all indexes complete standardized dimensionless conversion, output a standardized fusion working condition matrix in uniform format, where each row in the matrix corresponds to a one-second working condition, and each column corresponds to a type of process and equipment indexes.

[0018] Step S3: Input the normalized fused working condition matrix into a four-layer hierarchical progressive collaborative optimization model that has completed offline training in advance and continuously iterates and optimizes online to carry out multi-constraint solution calculation, and output multiple groups of feasible candidate process parameter sets; the four-layer hierarchical progressive collaborative optimization model is strictly ranked according to priority from high to low as dynamic equipment safety hard constraint layer, clear liquor permanent quality constraint layer, slurry treatment efficiency layer, and Pareto energy consumption optimal layer for classified water and electricity. It is clearly stipulated that all optimization objectives of the lower layer cannot break through the constraint boundary given by the upper layer. The safety layer has three sets of safety check logic built-in: the simultaneous equations of material conservation in the whole process of slurry, the hard constraint of critical time of plate-and-frame pressing process, and the full equipment load saturation clipping operator. Any parameter combination that violates the safety equations is directly eliminated.

[0019] Step S4: Based on the multi-objective non-dominated optimal solution set obtained from the model solution, split the fast variable adjustment command and slow variable setpoint command according to the difference in response speed, and match the multi-variable step-by-step and time-based adjustment anti-oscillation control mechanism. The commands are distributed to the plant's DCS distributed control system in layers to automatically adjust the real-time action parameters of various actuators such as feed pumps, return valves, plate and frame presses, and backwashing devices. The single adjustment amplitude is limited to avoid drastic fluctuations in operating conditions.

[0020] Step S5: Establish a three-level physical isolation read / write interlock negative feedback closed-loop correction mechanism to feed back the actual on-site operating condition prediction deviation to the front-end data fusion stage, continuously iterating and updating the internal calculation coefficients of the optimization model. The closed-loop correction includes three independent parallel sub-processes: instantaneous second-level feedback, hourly cached feedback, and daily exclusive global iteration. At the same time, introduce the filter material dual-variable time-series decay function and differential pressure slope layered blockage discrimination logic, and simultaneously match the lightweight backup interlock control logic of the DCS system. Under extreme operating conditions, the safe operation of the device can be guaranteed without relying on the upper-level optimization server.

[0021] This invention overcomes inherent industry defects such as independent control of surface filtration and plate and frame filter press in traditional three-acid purification processes, lack of adaptive compensation for lagging laboratory data, inability to dynamically adjust parameters due to equipment filter material aging, imbalance of multiple optimization objectives, fixed and rigid safety thresholds, and frequent upstream and downstream load shocks. It constructs four major systems: entropy weight multi-source heterogeneous data hierarchical fusion, four-layer hierarchical progressive safety priority collaborative optimization, three-level time-series interlocking negative feedback correction, and sludge conservation decoupling. These systems are simultaneously equipped with adaptive identification of filter material attenuation, frequency-based command issuance of fast and slow variables, and a lightweight DCS backup interlocking mechanism. Information entropy enables objective and dynamic weighting, strictly adhering to the optimization logic of step-by-step constraints on equipment safety, filtrate quality, processing efficiency, and energy consumption costs. It relies on three-level closed-loop feedback (second / hour / day) to correct model parameters in real time and dynamically adjust the backwashing and pressing process durations based on filter media wear. At the same time, it relies on the material conservation equation to decouple upstream and downstream loads and issue speed adjustment commands in layers to avoid operating condition oscillations. Ultimately, it achieves full-process adaptive intelligent control of four parallel surface filters coupled with two-stage plate and frame filters, significantly reducing the frequency of filtrate exceeding standards and equipment overload failures, significantly reducing water, electricity, and filter media consumption, and improving the overall stability of continuous operation and comprehensive production economic benefits of the three-acid purification section.

[0022] In an optional implementation, the fusion processing of multi-source heterogeneous data specifically includes: S2.1: Multi-condition segmented spatiotemporal alignment. Piecewise linear interpolation functions are established for three typical conditions: normal slurry feeding, high-concentration slurry overload, and equipment shutdown for maintenance. The time coordinate axis of DCS second-level high-frequency time series and offline test data with a lag of 2 hours is unified to eliminate the quality prediction deviation caused by the time series misalignment of test data. S2.2: Entropy weight adaptive hierarchical weighted operation. The system automatically reads all data every hour to calculate the entropy of various indicators. It achieves objective dynamic weighting based on the entropy value. It forces the allocation of weights for safety and quality data to be permanently higher than that for energy consumption data. The fixed and unchangeable weight priority is: safety alarm indicators > clear liquid quality indicators > sludge raw material composition indicators > equipment deterioration and loss indicators > water and electricity consumption indicators. S2.3: A three-tiered data fault tolerance and compensation mechanism. When the data loss duration for a single type of indicator is less than 1 hour, the corresponding parameters under the same historical operating conditions are retrieved and piecewise interpolation is used to complete the data loss. When the data loss duration is between 1 and 4 hours, all thresholds of the current optimization model are automatically locked, and model iteration updates are paused. When the data loss duration is greater than 4 hours, the system automatically switches to semi-automatic operation mode and simultaneously triggers on-site audible and visual alarms to remind operators to intervene in manual monitoring. S2.4: Automatic cleaning and verification of abnormal dirty data in conjunction with laboratory values. The program automatically identifies and removes abnormal dirty data such as pump idling, hydraulic pipeline leakage, and sensor zero drift. Manually entered slurry and clear liquid concentration values ​​are simultaneously verified with the equipment real-time differential pressure and front-end feed concentration. Abnormal values ​​that exceed the reasonable range of the process are directly intercepted and do not participate in subsequent matrix fusion calculations. S2.5: All process indicators are uniformly normalized to a dimensionless standard for single-ton slurry output, eliminating interference from calculations of indicators with different dimensions such as flow rate, pressure, concentration, and energy consumption. The final output is a fusion working condition matrix with a unified format that can be directly input into the model.

[0023] In an optional implementation, the execution logic of the four-layer hierarchical progressive collaborative optimization model includes: Step 3.1: Pre-screening of dynamic equipment safety hard constraint layer, setting two levels of control thresholds. The first level is a permanent and unmodifiable hard red line (solid content of clear liquid ≤3%, critical time of plate and frame pressing process, allowable limit of hydraulic pressure / differential pressure of equipment, rated overload current of motor, which cannot be manually modified under all production conditions). The second level is a dynamically adjustable safety range that changes in conjunction with the slurry concentration. All model output candidate parameters are first checked by the material conservation equation system. Solution sets that violate the material balance are directly eliminated. At the same time, a global slurry load saturation trimming operator is configured. When the feed flow rate exceeds the rated capacity of the device, the adjustment range of various parameters is forcibly limited to prevent the equipment from overloading. Step 3.2: Permanent quality constraint layer for clear liquid, with clear liquid solid content ≤3% as an insurmountable hard process indicator, and simultaneously embedding the critical time constraint of plate and frame pressing to prohibit the program from extending the pressing time indefinitely, resulting in ineffective power and water resource consumption; Step 3.3: Sludge treatment efficiency layer, with built-in load balancing distribution sub-algorithm for 4 parallel surface filters, and a two-way decoupled coupling equation for clarified sludge-surface filter sludge-plate and frame feed is established. The three types of variables are synchronously linked: surface filter backwashing cycle, plate and frame pressing and holding time, and front-end slurry conditioning and water addition. The algorithm eliminates the load impact and material crosstalk between upstream and downstream serial processes. Step 3.4: Water and electricity Pareto energy consumption optimization layer. All production power consumption, cleaning water consumption, and filter material consumption costs are calculated in a dimensionless normalized manner to generate a multi-objective Pareto non-dominated optimal solution set. This allows managers to manually adjust the preference weights of the three types of indicators (water, electricity, and filter material consumables) on a monthly basis, avoiding the damage to quality and equipment safety caused by the infinite compression of a single energy consumption indicator.

[0024] In a further scheme, the bidirectional decoupling equation of the sludge is fully embedded in the hard constraint equation set of material conservation throughout the entire process: total amount of clarified sludge feed = output of surface filtration clear liquid + output of surface filtration sludge; output of surface filtration sludge = total amount of feed to plate and frame filter press; total amount of feed to plate and frame filter press = output of clear liquid from plate and frame filter press + total amount of solid phase in plate and frame filter cake. All parameter combinations that do not satisfy the material balance equation are directly discarded and not included in the optimal solution set.

[0025] In a further embodiment, the three-level physical isolation read / write interlock negative feedback closed-loop correction mechanism specifically includes: 1) 1-second response to differential pressure and hydraulic instantaneous sudden disturbances, real-time lock of the current complete working condition snapshot, independent calculation of temporary compensation valve position adjustment, only control of fast-response variables such as pumps and valves, and setting upper and lower limits for single adjustment amplitude to prevent continuous adjustment from causing system operating condition oscillations; 2) The offline test results of the clear liquid quality deviation data are obtained every 2 hours and stored in an independent isolated cache pool. The global optimization model coefficients will not be rewritten in real time. Only the backwashing cycle and plate and frame pressing target duration are slightly adjusted. The permanent safety and quality red line parameters remain unchanged throughout the process. 3) The system automatically acquires exclusive read and write permissions for the model at 0:00 every day. The accumulated working condition deviation data of the whole day is written into the model in batches to complete the global iteration. During the iteration process, the system automatically switches to semi-automatic operation mode and sets a 1-hour smooth transition period. During the transition period, large-scale parameter adjustments are prohibited. The mechanism incorporates a built-in bivariate coupled attenuation function for the filter media. The independent variables of the function include the total cumulative operating time of the filter cloth / membrane and the average solids content of the slurry during a single day. It distinguishes between two types of clogging states: reversible clogging on the surface and irreversible clogging in the deep layer, based on the slope of the pressure difference change. When the clear liquid index exceeds the standard twice consecutively, the system automatically pushes an offline chemical cleaning warning for the filter media, rather than simply relying on extending the online backwashing time to solve the clogging. During equipment maintenance and unloading shutdown, the program automatically freezes the model weight update process to avoid ineffective iterations.

[0026] In a further embodiment, the issuance of optimal control commands along with corresponding fault fallback logic includes: The optimal solution set obtained from the model is divided into fast variables (feed pump operating frequency, backwash control valve opening, issued and executed at a second-level frequency) and slow variables (plate and frame pressing time, surface filtration backwash cycle, updated DCS setpoints at an hour-level frequency). Different variables are issued step by step at 30-second intervals within the same adjustment cycle, effectively suppressing the oscillation of operating conditions caused by continuous adjustment. The entire system is set up with a three-level progressive backup operating architecture: fully automatic parameter optimization operation mode → semi-automatic manual fine-tuning mode → DCS local lightweight two-layer optimization backup program; when the edge gateway and the upper-layer optimization server are offline and disconnected for more than 30 seconds without communication signal, the DCS local system automatically activates the lightweight security + quality two-layer optimization logic instead of switching to the fixed timing old control program. When the unit experiences interlocking failures such as clear liquid mixing or hydraulic over-extension, the system synchronously links the upstream slurry feed pump to reduce its operating frequency, thereby reducing the overall load on the unit from the source of materials and preventing continuous system oscillation caused by repeated start-stop cycles of the interlocking.

[0027] In an optional implementation, after the entire method is executed in one complete run, the system simultaneously performs four-dimensional quantitative process condition archiving and model iteration support: from four dimensions, the frequency of safety failures, the overall pass rate of clear liquid products, the sludge processing capacity per unit time, and the unit consumption of water, electricity and consumables by category, the data are independently archived according to three types of working conditions: normal sludge, high sludge load, and filter media aging; a standardized verification dataset is automatically generated daily as the input data source for the global model iteration of the day, and a built-in sub-module for tracing out out-of-standard issues is included, which can automatically locate the root cause of abnormal indicators as three types of problems: raw material composition fluctuation, filter media aging and loss, and improper adjustment of process parameters.

[0028] Example 2 This embodiment provides a real-time optimization system for the purification process parameters of three acids based on data fusion. The system adopts a modular and layered architecture design, with each module deployed independently and functions decoupled. (See reference...) Figure 2 It specifically includes six major functional modules: The data fusion module is used for unified access of seven types of heterogeneous production data in the plant area, multi-condition segmented interpolation time sequence alignment, entropy weight method adaptive hierarchical weighting, hierarchical data fault tolerance compensation, abnormal dirty data cleaning, and dimensionless normalization of all indicators, and finally outputs a standardized fusion condition matrix. The collaborative optimization calculation module includes a built-in dynamic safety load trimming operator, a slurry full-process material conservation decoupling algorithm, a load balancing distribution logic for four parallel filters, and an independent calculation unit for multi-objective Pareto solutions in hydropower. The three-level time-series negative feedback correction module includes a second-level instantaneous disturbance correction subunit, an hourly laboratory buffer feedback subunit, and a daily exclusive global iteration subunit. It also integrates a filter media dual-variable attenuation calculation unit and a differential pressure stratification blockage intelligent discrimination unit. The instruction issuance module enables time-sharing issuance of isolated fast and slow response variables, interlocking and linkage control of upstream and downstream equipment, lightweight local DCS safety control, and automatic archiving and storage of full-process operation logs. The working condition archiving and traceability module is responsible for statistical analysis of four major dimensions of operating indicators: safety, quality, production capacity, and energy consumption, and automatically generates daily verification datasets and anomaly traceability reports. The human-machine interaction monitoring module, equipped with a large visual screen, displays real-time operating conditions, model optimization parameters, equipment degradation curves, energy consumption statistics, and alarm and early warning information. It supports operators to view online and manually fine-tune settings within their authorized limits. All functional modules communicate with each other through an industrial isolation edge gateway, a local industrial server on the factory intranet, and the existing on-site DCS control system. All existing on-site measuring devices, such as differential pressure, oil pressure, flow rate, and online testing equipment, can be directly reused, eliminating the need for a large number of additional high-precision testing devices and resulting in lower implementation costs.

[0029] Example 3 This embodiment provides an electronic device that can be deployed in an industrial server room within a factory area to run the real-time optimization method for the three-acid purification process parameters described in Embodiment 1. (See attached document.) Figure 3 The hardware components include: a processor, a high-speed memory, and a device communication bus, wherein the processor and the memory communicate with each other at high speed through the communication bus; The memory is internally divided into four major areas: program storage area, real-time operating condition cache area, historical data archive area, and model iteration cache area. The memory stores a complete program instruction package that can be called and executed by the processor. When the processor calls and runs the program instruction package, it can completely execute the entire process of the real-time optimization method for three-acid purification process parameters based on data fusion as described in any optional implementation of Embodiment 1. The equipment is additionally equipped with an industrial Ethernet communication interface, an audible and visual alarm output interface, and a local display interaction interface, and is compatible with the deployment standards of chemical explosion-proof computer rooms.

[0030] Example 4 This embodiment provides a computer storage medium, which can be selected from various industrial storage carriers such as industrial-grade solid-state drives, USB flash drives, and disk arrays. The storage medium has a complete computer program instruction code package permanently stored on it. When the program instruction code package is loaded and executed by any processor with computing power, it can completely implement all the steps of the real-time optimization method for three-acid purification process parameters based on data fusion described in Embodiment 1 and all its optional embodiments. The storage medium supports three program deployment methods: offline copying, local deployment on site, and remote distribution from the cloud, and is suitable for various network environments in chemical industrial sites, such as those without external networks or with isolated internal networks.

[0031] The above-described embodiments are only used to illustrate the technical solutions of the present invention, and are not intended to limit it. Although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some of the technical features. Such modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the spirit and scope of the technical solutions of the embodiments of the present invention.

Claims

1. A method for real-time optimization of process parameters for the purification of three acids based on data fusion, characterized in that, A method for purifying three acids, applied to a surface filtration system and a plate and frame filter press system, includes the following steps: Acquire multi-source heterogeneous data of the entire process of tri-acid purification, including DCS time series data, offline test data, equipment degradation data and common media data; The multi-source heterogeneous data is spatiotemporally aligned and hierarchically weighted and fused to generate a fused working condition matrix; The fusion working condition matrix is ​​input into a preset four-layer hierarchical progressive collaborative optimization model for calculation to obtain a set of candidate process parameters. The four-layer hierarchical progressive collaborative optimization model includes a hard constraint layer for equipment safety, a constraint layer for clear liquid quality, a sludge treatment efficiency layer, and a water and electricity energy consumption optimization layer arranged in order of priority from high to low. The lower layer constraint serves as the insurmountable boundary for the upper layer optimization. The optimal control command is generated based on the calculation results and sent to the DCS system to adjust the action parameters of the field actuators.

2. The method according to claim 1, characterized in that, The spatiotemporal alignment and hierarchical weighted fusion of the multi-source heterogeneous data specifically includes: Align the testing times of offline test data with the time axis of DCS time series data; The safety alarm data is assigned a first weight, the clear liquid quality data is assigned a second weight, and the energy consumption data is assigned a third weight, wherein the first weight > the second weight > the third weight; The integrated operating condition matrix is ​​generated by calculating a comprehensive score for the current operating condition based on the weighted data.

3. The method according to claim 2, characterized in that, Specifically, the steps include the following: S2.1: Segmented linear interpolation based on three working conditions: normal sludge, high sludge, and equipment maintenance, to unify the DCS second-level time sequence and the 2-hour lag test data time axis; S2.2: Calculate the entropy of various data information every hour to achieve objective dynamic weighting, and force the weight of safety and quality data to be permanently higher than that of energy consumption data. Fixed weight priority: safety alarm > clear liquid quality > slurry raw material > equipment deterioration > water and electricity consumption; S2.3: Data missing for less than 1 hour is filled by interpolation based on historical data under the same operating conditions; data missing for 1-4 hours is frozen at the current model threshold; data missing for more than 4 hours is automatically switched to semi-automatic mode and an audible and visual alarm is triggered. S2.4: Automatically remove data related to pump idling, hydraulic leakage, and dirt; manually enter sludge / clean liquid concentration and verify it in conjunction with real-time differential pressure and feed concentration; abnormal values ​​are blocked from entering the fusion matrix. S2.5: All process parameters are normalized to dimensionless values ​​per ton of slurry, and a fused operating condition matrix is ​​output.

4. The method according to claim 3, characterized in that, The execution logic of the four-layer hierarchical progressive collaborative optimization model is as follows: First, the current operating parameters are compared with the equipment's safety hard constraint layer. If the safety boundary is triggered, the control parameters are locked and optimization is stopped. If the safety boundary is not triggered, the predicted parameters are compared with the clear liquid quality constraint layer. If the predicted parameters exceed the quality threshold, a quality compensation command is generated. Under the premise of satisfying the equipment safety hard constraint layer and the clear liquid quality constraint layer, the maximum matching processing capacity of upstream and downstream equipment is calculated with the sludge treatment efficiency layer as the optimization target. Under the premise of satisfying all the aforementioned hierarchical constraints, the optimal control command is obtained by taking the optimal hydropower energy consumption layer as the objective function.

5. The method according to claim 4, characterized in that, When the sludge treatment efficiency layer is taken as the optimization target, the backwashing cycle of the surface filtration system and the pressing time of the plate and frame filter press system are used as coupling variables. By establishing an upstream and downstream load matching model, the load impact between series processes is eliminated.

6. The method according to claim 4, characterized in that, Before inputting the fused operating condition matrix into the preset four-layer hierarchical progressive collaborative optimization model, the method further includes: A three-level negative feedback closed-loop correction mechanism is constructed, wherein the three-level negative feedback closed-loop correction mechanism includes: Instantaneous second-level negative feedback: Real-time monitoring of pressure mutation values ​​in DCS timing data. When the mutation value exceeds the first threshold, a fast variable command is generated to adjust the feed valve opening. Hourly-level negative feedback: Receive the offline test data, calculate the actual quality deviation, and when the deviation exceeds the second threshold, correct the regression coefficients in the collaborative optimization model; Daily-level cyclical negative feedback: The performance degradation index is calculated based on equipment degradation data. When the degradation index exceeds the third threshold, the baseline control parameters in the collaborative optimization model are automatically updated.

7. The method according to claim 6, characterized in that, The automatic updating of the baseline control parameters in the collaborative optimization model specifically includes: calculating the permeability decay coefficient based on the cumulative running time of the filter cloth or filter membrane; dynamically extending the pressing and holding time of the plate and frame filter press system based on the permeability decay coefficient, and simultaneously adjusting the backwash water pulse frequency.

8. The method according to claim 1, characterized in that, The process of generating the optimal control command and sending it to the DCS system specifically includes: decoupling the optimal control command into fast variable commands and slow variable commands; sending the fast variable commands to the field actuators at a frequency of seconds, and sending the slow variable commands to the setpoint interface of the DCS system at a frequency of hours.

9. A real-time optimization system for tri-acid purification process parameters based on data fusion, characterized in that, include: The data fusion module is used to perform the data acquisition and fusion steps in the method according to any one of claims 1-8; A collaborative optimization calculation module is used to execute the four-level hierarchical progressive calculation steps in the method of any one of claims 1-8; The instruction issuing module is used to convert the calculation results into control signals that the DCS can recognize; The data fusion module, collaborative optimization calculation module, and instruction issuance module are connected through an industrial communication network.

10. An electronic device, characterized in that, include: A processor, a memory, and a communication bus, wherein the processor and the memory communicate with each other via the communication bus; The memory stores program instructions that can be executed by the processor. When the processor calls the program instructions to execute the program, it implements the real-time optimization method for tri-acid purification process parameters based on data fusion as described in any one of claims 1-8.

Images