An aluminum electrolysis cell power distribution system regionalized independent discharging control method based on pseudo-resistance monitoring

Abstract

The application is suitable for the technical field of electrical measurement and state monitoring, and provides an aluminum electrolysis cell power distribution system regionalized independent discharging control method based on pseudo-resistance monitoring, which comprises the following steps: collecting anode currents and cell voltages of each measuring point; dividing the multiple measuring points into several control regions, and calculating pseudo-resistance values of each control region; selecting control regions meeting conditions from all control regions to form a reference region set; determining alumina concentration reference states of each control region in the reference region set through a preset concentration state mapping relationship; taking the reference region set as a starting point, and iteratively matching and propagating alumina concentration states to all control regions according to spatial proximity relationships and pseudo-resistance value similarities of control regions with undetermined states and the reference region set; generating independent discharging control instructions of each control region according to alumina concentration states of all control regions; and realizing regionalized independent control of alumina concentration in the aluminum electrolysis cell.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of electrical measurement and condition monitoring, and particularly relates to a regionalized independent discharging control method for an aluminum electrolysis cell power distribution system based on pseudo-resistance monitoring. BACKGROUND

[0002] In aluminum electrolysis production, accurate measurement of the anode current of the electrolysis cell is the basis for process monitoring and optimization. Currently, the measurement of the series current or local current of the aluminum electrolysis cell mainly uses the equidistance voltage drop method or the Hall current sensor. These technologies measure the voltage drop or magnetic field on the conductive bus, indirectly calculate the current value flowing through, and take this current data as one of the input bases for the cell control system to judge the working state of the electrolysis cell and perform unified feeding control.

[0003] However, the measurement accuracy of the equidistance voltage drop method is easily affected by factors such as contact resistance and uneven bus temperature, and the actual accuracy is low. The Hall current sensor is easily disturbed by stray magnetic fields and temperature drift, and has poor measurement stability in the unstable environment of strong magnetic fields and high temperatures in the electrolysis workshop. In addition, as the electrolysis cell develops towards large-scale, the internal physical and chemical field distribution is uneven, resulting in obvious regional differences in the concentration of aluminum oxide. The existing method can only provide overall average state information and cannot determine the concentration changes in the local region, and cannot make targeted adjustments when an abnormal concentration occurs in a certain region, causing local underfeeding or overfeeding, reducing current efficiency, increasing power consumption, and even possibly causing local anode effect or failure problems. SUMMARY

[0004] The present application provides a regionalized independent discharging control method for an aluminum electrolysis cell power distribution system based on pseudo-resistance monitoring, which solves the problem that the existing method cannot determine the concentration changes in the local region, and even possibly causes local anode effect or failure problems.

[0005] The present application provides a regionalized independent discharging control method for an aluminum electrolysis cell power distribution system based on pseudo-resistance monitoring, comprising:

[0006] Through the optical fiber current sensor deployed at multiple measurement points on the anode bus of the aluminum electrolysis cell, the anode current data of each measurement point is collected in real time; at the same time, the cell voltage of the aluminum electrolysis cell is measured;

[0007] According to the spatial topological relationship of each measurement point, the multiple measurement points are divided into several control regions; for each control region obtained by division, the pseudo-resistance value of the control region is calculated according to the sum of the anode current data of all measurement points in the control region and the cell voltage;

[0008] According to the pseudo-resistance values of the control regions and the historical change trend of the pseudo-resistance values of the control regions, control regions meeting a reference determination condition are screened out from all the control regions to form a reference region set; the reference determination condition includes that the pseudo-resistance value of a control region is in a preset clear state interval for a preset number of continuous periods, and a fluctuation coefficient of the pseudo-resistance value is less than a preset fluctuation threshold;

[0009] According to the pseudo-resistance values of the control regions in the reference region set, an alumina concentration reference state of each control region in the reference region set is determined through a preset concentration state mapping relationship;

[0010] Taking the reference region set as a starting point, alumina concentration states are propagated to all the control regions through iterative matching according to the spatial proximity relationship and the similarity of the pseudo-resistance values between the control regions with an undetermined state and the reference region set, until the alumina concentration state of each control region is determined;

[0011] According to the alumina concentration states of all the control regions, independent unloading control instructions of the control regions are generated in combination with material diffusion effects between the control regions and action constraints of unloading equipment.

[0012] Further, the control regions are divided from the plurality of measurement points according to the spatial topological relationship of the measurement points, including:

[0013] A multi-dimensional feature vector is constructed for each measurement point, and the multi-dimensional feature vector includes a spatial position coordinate of the measurement point, a unloader identifier to which the measurement point belongs, and a current response intensity of the measurement point to an adjacent unloading action based on historical data statistics;

[0014] A two-stage clustering division is performed, including:

[0015] Based on the spatial position coordinate and the unloader identifier, the plurality of measurement points are initially coarsely divided to form a plurality of initial candidate regions;

[0016] For each initial candidate region, the initial candidate region is internally merged or split optimized based on the current response intensity of each measurement point in the region, to obtain a plurality of control regions.

[0017] Further, the initial candidate region is internally merged or split optimized, including:

[0018] A variance of the current response intensity feature between all the measurement points in each initial candidate region is calculated as a consistency index of the initial candidate region;

[0019] calculate a difference degree of the current response intensity features between two adjacent measurement points at a boundary of any two adjacent initial candidate regions as a boundary coupling degree indicator between the two adjacent initial candidate regions;

[0020] when the consistency indicator of an initial candidate region is lower than a first preset threshold, keep the initial candidate region intact;

[0021] when the consistency indicator of an initial candidate region is higher than the first preset threshold, split the initial candidate region into at least two new control regions;

[0022] when the boundary coupling degree indicator between two adjacent initial candidate regions is higher than a second preset threshold, merge the two adjacent initial candidate regions into one control region.

[0023] Further, the filtering of the control regions that meet the reference determination condition from all the control regions comprises:

[0024] for each control region, calculating a state determination confidence score according to a depth at which the pseudo-resistance value of the control region falls into the explicit state interval within the continuous preset number of periods and a matching degree between a pseudo-resistance variation trend of the control region and a preset typical variation mode of the state interval;

[0025] determining all the control regions with the state determination confidence score higher than a preset confidence threshold as a reference region set.

[0026] Further, the preset concentration state mapping relationship is defined by preset alumina concentration state division thresholds; the alumina concentration state division thresholds comprise pseudo-resistance value thresholds corresponding to under-concentration state, normal state and over-concentration state.

[0027] Further, the determination of the alumina concentration reference state of each control region in the reference region set according to the pseudo-resistance value of the control region in the reference region set through the preset concentration state mapping relationship comprises:

[0028] continuously comparing the pseudo-resistance value of each control region in the reference region set with the alumina concentration state division thresholds;

[0029] determining the alumina concentration reference state of the control region as one of under-concentration state, normal state or over-concentration state according to a specific interval into which the pseudo-resistance value falls in a pseudo-resistance value interval defined by the division thresholds.

[0030] Further, the alumina concentration state is propagated to all control regions by iterative matching according to the spatial proximity and the similarity of pseudo-resistivity values between the control regions in the undetermined state and the reference region set, until the alumina concentration state of each control region is determined, including:

[0031] A control region adjacency network is constructed, each control region is taken as a node in the control region adjacency network, and if two control regions are adjacent in space, a connection edge is established between the corresponding two nodes;

[0032] The nodes corresponding to each control region in the reference region set are marked as determined state nodes, and the alumina concentration reference state of each control region in the reference region set is taken as the known state of the corresponding node;

[0033] The following steps are repeatedly executed until all nodes in the control region adjacency network are marked as determined state nodes, or the preset maximum number of iterations is reached:

[0034] All nodes adjacent to the determined state nodes and not yet in the determined state are selected from the control region adjacency network to form a set of nodes to be matched;

[0035] For each node to be matched in the set of nodes to be matched, the pseudo-resistivity similarity between the node to be matched and each adjacent determined state node is calculated;

[0036] For each node to be matched in the set of nodes to be matched, based on the pseudo-resistivity similarity between the node to be matched and each adjacent determined state node and the alumina concentration state of each adjacent determined state node, the state confidence of the node to be matched belonging to each alumina concentration state is calculated;

[0037] For each node to be matched in the set of nodes to be matched, if the maximum state confidence of the node to be matched is greater than a preset propagation confidence threshold, the node to be matched is marked as a determined state node, and the alumina concentration state corresponding to the maximum state confidence is taken as the alumina concentration state of the node;

[0038] If the maximum state confidence of all nodes to be matched in the set of nodes to be matched is not greater than the preset propagation confidence threshold, the iteration is stopped.

[0039] Further, the calculation of the state confidence of the node to be matched belonging to each alumina concentration state includes:

[0040] According to the alumina concentration states of each adjacent determined state node of the to-be-matched node and the pseudo-resistance value similarity between the to-be-matched node and each adjacent determined state node, a comprehensive support degree of each alumina concentration state of the to-be-matched node is calculated.

[0041] The comprehensive support degree is normalized to obtain a state confidence degree of each alumina concentration state of the to-be-matched node.

[0042] Further, when the repeated execution step is terminated because the preset maximum iteration number is reached, if there are still undetermined state nodes in the control region adjacent network, the following steps are performed:

[0043] For each undetermined state node, the alumina concentration state of all adjacent nodes of the undetermined state node in the control region adjacent network is obtained.

[0044] The alumina concentration state with the highest occurrence frequency in the adjacent nodes is taken as the alumina concentration state of the undetermined state node.

[0045] The undetermined state node is marked as a determined state node.

[0046] Further, the independent feeding control instructions of each control region are generated according to the alumina concentration states of all control regions, combined with the material diffusion effect between control regions and the action constraints of the feeding device, including:

[0047] Based on the alumina concentration states of each control region, the target feeding amount required by each control region is determined.

[0048] A feeding instruction optimization model is constructed, the optimization model taking the minimization of the whole tank alumina concentration distribution imbalance degree and the total action number of the feeding device as an objective function, and taking the target feeding amount, the material diffusion effect model and the feeding device action constraints as constraint conditions.

[0049] The feeding instruction optimization model is solved to obtain the feeding time and feeding amount of each control region in the next control period, forming the independent feeding control instructions.

[0050] From the above technical solutions, the present application has the following advantages:

[0051] The application acquires current data of each measuring point on the anode bus of the aluminum electrolysis cell and cell voltage by deploying a fiber current sensor network, divides control areas based on the spatial topological relationship of the measuring points and calculates pseudo-resistance values of each area, selects a set of reference areas with clear and stable states, and determines the alumina concentration reference state thereof through a preset concentration state mapping relationship, then takes the reference area as a starting point, combines spatial proximity and pseudo-resistance similarity, and propagates the concentration state to all areas through iterative matching, and finally generates independent discharge control instructions for each area according to the accurate concentration state of each area, considering material diffusion and equipment constraints. BRIEF DESCRIPTION OF DRAWINGS

[0052] Figure 1 An embodiment flowchart of the aluminum electrolysis cell power distribution system regional independent discharge control method based on pseudo-resistance monitoring in the application;

[0053] Figure 2 An embodiment flowchart of the aluminum electrolysis cell power distribution system regional independent discharge control method based on pseudo-resistance monitoring in the application;

[0054] Figure 3 An embodiment flowchart of the aluminum electrolysis cell power distribution system regional independent discharge control method based on pseudo-resistance monitoring in the application. DETAILED DESCRIPTION

[0055] The terms "first", "second", "third", "fourth" and the like in the description of the application and in the claims of the foregoing drawings, if any, are used for distinguishing between similar objects and not necessarily for describing a particular sequential or chronological order. It is to be understood that the use of the terms so construed can be interchanged, under appropriate circumstances, to describe the embodiments of the application described herein in a sequence other than the one illustrated or described herein. Furthermore, the terms "comprise" and "comprising" and any variations thereof, are intended to cover a non-exclusive inclusion, for example, a process, method, system, product or apparatus that comprises a list of steps or units can not necessarily be limited to those steps or units that are clearly listed, but can include other not clearly listed steps or units or other steps or units inherent to such process, method, product or apparatus.

[0056] Embodiment one

[0057] The method implemented in this embodiment can be implemented in a system, and can be implemented in a server or a terminal, and the specific implementation is not limited. The method in this application will be introduced from the perspective of system implementation. Please refer to Figure 1 The method provided by the embodiment of the application comprises the following steps:

[0058] S1. By using fiber optic current sensors deployed at multiple measuring points on the anode busbar of the aluminum electrolysis cell, the anode current data of each measuring point is collected in real time; at the same time, the cell voltage of the aluminum electrolysis cell is measured.

[0059] On the anode busbar of the target aluminum electrolytic cell, several representative locations are selected as measurement points based on the process layout and the distribution of feeding points. At each measurement point, a high-precision fiber optic current sensor is deployed around the anode busbar guide rod or busbar. This sensor operates based on the Faraday magneto-optical effect principle. Specifically, the sensing fiber forms a closed loop. When a large DC current flows through the busbar, the generated surrounding magnetic field causes the polarization plane of the linearly polarized light transmitted in the fiber to rotate. The rotation angle is proportional to the current intensity. By detecting this rotation angle, the real-time anode current value flowing through the measurement point can be calculated. The system synchronously collects current data from all measurement points and simultaneously measures the cell voltage of the electrolytic cell using a high-precision voltage transmitter. The collected raw data includes at least: the channel number corresponding to each measurement point, the real-time anode current value, the cell voltage value, and a precise timestamp.

[0060] S2. Based on the spatial topology of each measurement point, divide the multiple measurement points into several control regions; for each control region, calculate the pseudo resistance value of the control region based on the sum of the anode current data of all measurement points in the control region and the tank voltage.

[0061] Due to the complex electromagnetic, flow, and temperature field distributions within large aluminum electrolysis cells, the alumina concentration is not uniformly distributed within the cell. Therefore, this step aims to aggregate physically discrete measurement points into several control regions with relatively consistent internal states based on their spatial correlation and electrical behavior similarity. After region division, for each control region, the anolyte current data of all measurement points within that region are summed in real time to obtain the equivalent total current for that region. Simultaneously, the synchronously acquired slot voltage is read. The pseudo resistance value of this region was calculated based on Ohm's law. Pseudo-resistance is a key equivalent parameter characterizing the cell condition in aluminum electrolysis, reflecting state information closely related to alumina concentration, such as electrolyte resistance and polarization voltage. A decrease in alumina concentration in the electrolyte leads to a decrease in its conductivity, which in turn causes an increase in the pseudo-resistance value; conversely, an excessively high concentration may lead to a decrease in the pseudo-resistance value. Therefore, by calculating the pseudo-resistance value of each control region, a real-time quantitative index reflecting the alumina concentration level in that local area can be obtained. Please refer to [link to relevant documentation]. Figure 2 The process of dividing the control area includes the following:

[0062] 201. Construct a multi-dimensional feature vector for each measurement point, including the spatial position coordinates of the measurement point, the identifier of the feeder to which the measurement point belongs, and the current response intensity of the measurement point to the adjacent feeding action based on historical data statistics;

[0063] The spatial position coordinates are three-dimensional coordinates or one-dimensional linear coordinates along the anode busbar, accurately describing the physical position of the measurement point. The feeder identifier indicates to which specific feeder the measurement point belongs in the process, and one feeder usually affects the electrolyte concentration within a certain range around it. The adjacent feeding action refers to the feeding operation performed by the feeders corresponding to the control regions directly adjacent to the control region to which the measurement point belongs in space; the current response intensity is a dynamic process feature, which is used to quantify the sensitivity of the measurement point to the feeding action of the adjacent region; the acquisition method is to analyze historical data, and when a feeder performs a feeding action, the fluctuation amplitude and response time of the anode current of each measurement point within its influence range are counted, and a response intensity coefficient representing the response of the point to the feeding disturbance is obtained through normalization processing; the region with high response intensity is more susceptible to local concentration changes.

[0064] 202. Perform two-stage clustering division, including:

[0065] 202a. Based on the spatial position coordinates and the feeder identifier, the plurality of measurement points are initially coarsely divided to form a plurality of initial candidate regions;

[0066] Specifically, the first stage is defined as a coarse division based on rules and positions. First, according to the feeder identifier, the measurement points are initially grouped into the influence range of the respective feeders. Then, within each group, according to the spatial position coordinates of the measurement points, secondary grouping is performed based on physical proximity, such as grouping every N adjacent guide rods. This stage forms a plurality of initial candidate regions, which are mainly divided based on static prior process layout information, ensuring the basic rationality of the division.

[0067] 202b. For each initial candidate region, based on the current response intensity characteristics of the measurement points in the region, the initial candidate region is internally merged or split optimized to obtain a plurality of control regions.

[0068] The second stage is defined as a fine optimization based on dynamic response, so that the final control regions have high electrical behavior consistency. The process specifically includes the following steps:

[0069] 1. Calculate the variance of the current response intensity characteristics between all measurement points within each initial candidate region as the consistency index of the initial candidate region;

[0070] 2. Calculate the difference degree of the current response intensity characteristics between adjacent measurement points at the boundary of any two adjacent initial candidate regions as the boundary coupling degree index between the two adjacent initial candidate regions;

[0071] Specifically, the consistency index is used to evaluate whether the responses of the measurement points in the region to the disturbance of the feed are synchronous and uniform. The smaller the variance, the more consistent the current response behavior of the points in the region, and the more suitable it is as a whole control unit. The boundary coupling degree index is used to evaluate whether the electrical behavior of the two adjacent regions at the boundary is continuous and gradual or there is a sudden change. The smaller the difference degree, the more similar the response characteristics of the measurement points on both sides of the boundary, which implies that the two regions may be affected by similar concentration fields and are closely coupled.

[0072] 3. When the consistency index of an initial candidate region is lower than a first preset threshold, the initial candidate region is kept intact;

[0073] 4. When the consistency index of an initial candidate region is higher than the first preset threshold, the initial candidate region is split into at least two new control regions;

[0074] 5. When the boundary coupling degree index between two adjacent initial candidate regions is higher than a second preset threshold, the two adjacent initial candidate regions are merged into one control region.

[0075] The first preset threshold is set by analyzing the statistical distribution of the current response intensity variance in a typical control region during a stable running stage and combining process experience. If the consistency index of a region is too high, it indicates that the internal response is uneven, and there may be unrecognized sub-regions, so it is split to ensure the consistency of the internal state of each final control region. The second preset threshold is set by statistically analyzing the difference degree of the response intensity at the boundary of two adjacent regions identified as the same concentration region in history. If the boundary coupling degree of the two regions is extremely high, it indicates that the electrical behavior is continuous, and artificial division may have split the same physical process, so they are merged to form a control unit with more physical meaning.

[0076] Through the above two stages of division and optimization, the final control region set not only respects the physical layout of the electrolytic cell and the feed process, but also fully considers the dynamic electrical response characteristics of each region, achieving an improvement from fixed physical zoning to dynamic intelligent zoning.

[0077] S3. According to the pseudo-resistance values of each control region and the historical change trend of the pseudo-resistance values of each control region, control regions that meet the reference determination condition are selected from all control regions to form a reference region set; the reference determination condition includes that the pseudo-resistance value of the control region is in a preset clear state interval for a continuous preset period of time, and the fluctuation coefficient of the pseudo-resistance value is less than a preset fluctuation threshold;

[0078] This step is to identify the control regions with clear state and reliable data from all control regions as reference regions, thereby providing a high-confidence starting point for the subsequent state propagation process. Due to the existence of complex factors such as electromagnetic interference and material transportation fluctuations during the operation of the electrolytic cell, not all pseudo-resistance signals of the regions can clearly reflect the alumina concentration. By setting the reference judgment condition, those regions with stable and clear pseudo-resistance signals can be selected, ensuring the accuracy of the subsequent state propagation. The historical change trend here refers to the change characteristics of the pseudo-resistance value of each control region in the past period of time, which is obtained by analyzing the time series of the pseudo-resistance value, and specifically includes the change direction, change rate, and volatility. This trend reflects the dynamic evolution process of the alumina concentration in the region.

[0079] 1. For each control region, a state determination confidence score is calculated according to the depth of the pseudo-resistance value of the control region falling into the clear state interval within a continuous preset number of periods and the matching degree of the pseudo-resistance change trend of the control region with the typical change mode of the state interval.

[0080] 2. All control regions with a state determination confidence score higher than the preset confidence threshold are determined as the reference region set.

[0081] Specifically, the preset number of periods refers to the length of the time window for continuous sampling, which is determined according to the process response speed and control period, for example, 10 consecutive sampling periods; this parameter ensures the persistence of the state and avoids false judgments caused by transient interference. The preset clear state interval refers to the pseudo-resistance value range that is pre-divided according to process knowledge and historical data and corresponds to a specific alumina concentration state; for example, set the pseudo-resistance value greater than the threshold for the understock state interval, between and for the normal state interval, and less than for the overstock state interval. The interval boundary values and are obtained by statistical analysis of the pseudo-resistance value and the measured value of the alumina concentration in the historical data, for example, taking the pseudo-resistance critical value corresponding to the concentration state switching. The fluctuation coefficient is a statistical indicator for measuring data stability, which is the standard deviation of the pseudo-resistance value sequence within a continuous preset number of periods divided by its average value; a small fluctuation coefficient indicates that the pseudo-resistance signal of the region is stable and less affected by random noise. The preset fluctuation threshold is used to define the upper limit of the acceptable fluctuation level, which is determined by statistical analysis of the pseudo-resistance fluctuation coefficients of all regions during the stable operation stage, for example, taking the 90% quantile value of the cumulative distribution of all region fluctuation coefficients, or setting a typical value according to process experience; below this threshold means that the signal is stable enough.

[0082] The depth falling into the explicit state interval refers to a value for quantifying the degree of deviation of the current pseudo-resistance value from the boundary of the state interval to which it belongs; for example, for the under-quantity state interval, the pseudo-resistance value is not only greater than the lower limit , but also exceeds more, the depth value is greater, indicating that the state characteristics are farther away from the ambiguous zone. The preset typical change pattern of the state interval refers to a typical rule capable of representing the change of the pseudo-resistance value with time under a certain specific alumina concentration state. The pattern is used to evaluate whether the change trend of the pseudo-resistance of the region to be judged conforms to the expectation; for example, the typical change pattern of the under-quantity state can be a straight line with a gentle positive slope, representing the slow rise of the pseudo-resistance caused by the continuous consumption of the concentration; the typical change pattern of the normal state can be a smooth sequence with small fluctuations around the mean value. These typical change patterns can be extracted by analyzing the pseudo-resistance time series in the stable state in a large amount of historical data using cluster analysis. The state judgment confidence score is a quantitative index calculated by comprehensively considering the depth and the matching degree, the higher the score, the more sufficient and reliable the evidence that the region is judged as the current state. The calculation formula of the state judgment confidence score is as follows: , wherein is the weight coefficient corresponding to the depth, is the weight coefficient corresponding to the trend matching degree, is a normalization function, which scales the values of and to the interval [0, 1], is the depth, is the matching degree. The preset confidence threshold is the final threshold value for determining whether a region can be used as a reference region, which is set to 0.75 here; only the region with a confidence score higher than the threshold value is considered to have a state judgment reliable enough to qualify as a reference for subsequent state propagation.

[0083] S4. According to the pseudo-resistance values of the control regions in the reference region set, the alumina concentration reference states of the control regions in the reference region set are determined through a preset concentration state mapping relationship;

[0084] Each reference region selected in step S3 is given an explicit discretized alumina concentration process state label, thereby providing a reliable known state source for the state propagation process in step S5. After S3 screening, the regions in the reference region set have met the conditions of state explicitness and data stability, and the pseudo-resistance values thereof can reliably reflect the alumina concentration level of the region. These reliable pseudo-resistance measurement values are mapped to state categories that can be understood and executed at the process operation level.

[0085] The alumina concentration state division threshold value refers to the critical value for dividing the continuous pseudo-resistance value range into several discrete state intervals, which mainly divides the alumina concentration state into three categories, corresponding to different process conditions and control requirements:

[0086] Underdose state: the alumina concentration is lower than the ideal range, and the feeding needs to be increased.

[0087] Normal state: the alumina concentration is within the ideal control range, and the current feeding should be maintained.

[0088] Overdose state: the alumina concentration is higher than the ideal range, and the feeding needs to be reduced or stopped.

[0089] The pseudo-resistance value threshold is the specific pseudo-resistance value corresponding to the boundaries of the above three state intervals. For example, two key thresholds can be defined, where the high threshold is the pseudo-resistance value higher than this threshold, which is determined as an underdose state; the low threshold is the pseudo-resistance value lower than this threshold, which is determined as an overdose state; and between and , it is determined as a normal state.

[0090] The specific process of determining the reference state is as follows:

[0091] 1. Continuously compare the pseudo-resistance value of each control region in the reference area with the alumina concentration state division threshold value;

[0092] Continuous comparison means that for a given reference area, its real-time pseudo-resistance value is compared with the preset low threshold and high threshold in turn.

[0093] 2. According to the specific interval in which the pseudo-resistance value falls in the pseudo-resistance value interval defined by the division threshold value, determine the alumina concentration reference state of the control region as one of underdose state, normal state or overdose state.

[0094] According to the comparison result, the deterministic mapping rule is executed:

[0095] If the pseudo-resistance value , the region falls into the underdose state interval, and the alumina concentration reference state is determined as underdose state;

[0096] If the pseudo-resistance value , the region falls into the normal state interval, and the alumina concentration reference state is determined as normal state;

[0097] If the pseudo-resistance value , the region falls into the overdose state interval, and the alumina concentration reference state is determined as overdose state.

[0098] By assigning a unique process state label to each reference region, these reference regions with under-, normal or over-quantity labels become reliable initial information sources for the subsequent state propagation algorithm.

[0099] S5. Starting from the reference region set, the alumina concentration state is propagated to all control regions through iterative matching according to the spatial proximity and similarity of pseudo-resistances between the control regions with uncertain state and the reference region set, until the alumina concentration state of each control region is determined;

[0100] Using the reference regions with determined states in step S4 as reliable information sources, the alumina concentration states of all other control regions are intelligently inferred by simulating the diffusion process of state information in the physical space. The basic principle is that the physical and chemical fields in the aluminum reduction cell have spatial continuity, so adjacent regions in the physical space often have similar states; at the same time, the similarity of pseudo-resistances as a key representation of the state also indicates the similarity of the state. Please refer to Figure 3 , by constructing an adjacency network, calculating similarity and confidence, and iterative propagation algorithms, the physical intuition is converted into a calculable and executable automated decision-making process.

[0101] 501. Construct an adjacency network of control regions, take each control region as a node in the adjacency network of control regions, and if two control regions are adjacent in space, establish a connection edge between the corresponding two nodes;

[0102] 502. Mark the nodes corresponding to each control region in the reference region set as determined state nodes, and take the alumina concentration reference state of each control region in the reference region set as the known state of the corresponding node;

[0103] Specifically, each node uniquely corresponds to a control region, and the establishment rule of the connection edge is: if two control regions are directly adjacent in the physical space, for example, sharing the boundary or the distance is less than the preset proximity distance threshold, then a edge is established between the two nodes corresponding to them. The network node corresponding to the reference region in step S4 which has been assigned a clear state is marked as a determined state node, and the corresponding alumina concentration reference state is taken as the known state of the node. These nodes constitute the initial information source set.

[0104] 503. Repeat the following steps until all nodes in the adjacency network of control regions are marked as determined state nodes:

[0105] The goal of each iteration below is to infer the most likely state of the directly adjacent node with unknown state based on the known state of the current determined state node:

[0106] 503a. From the control region contiguous network, select all nodes directly adjacent to the determined state nodes and not yet determined state, to form a set of nodes to be matched;

[0107] Traverse all determined state nodes, collect all neighbor nodes directly connected to each such node through the connection edge and not yet marked as determined state nodes. The set of these neighbor nodes is the set of nodes to be matched in this iteration. This ensures that the state propagation always advances from the known region to its nearest unknown neighborhood.

[0108] 503b. For each node to be matched in the set of nodes to be matched, calculate the pseudo-resistance value similarity between the node to be matched and each adjacent determined state node;

[0109] For each node in the set of nodes to be matched, identify all adjacent determined state nodes in the network, i.e. neighbor nodes whose state is known through edges. Then, calculate the similarity between the pseudo-resistance value of the current node to be matched and the pseudo-resistance value of each such neighbor node. The pseudo-resistance value similarity is an indicator that quantifies the closeness of the pseudo-resistance signals of the two regions, usually in the range [0, 1], the larger the value, the more similar.

[0110] 503c. For each node to be matched in the set of nodes to be matched, based on the pseudo-resistance value similarity between the node to be matched and each adjacent determined state node and the alumina concentration state of each adjacent determined state node, calculate the state confidence of the node to be matched for each alumina concentration state;

[0111] 1. Based on the alumina concentration state of each adjacent determined state node of the node to be matched and the pseudo-resistance value similarity between the node to be matched and each adjacent determined state node, calculate the comprehensive support of the node to be matched for each alumina concentration state;

[0112] 2. Normalize the comprehensive support to obtain the state confidence of the node to be matched for each alumina concentration state.

[0113] For each alumina concentration state , traverse all adjacent determined state nodes of the current node to be matched. If the known state of a neighbor node is exactly , then the neighbor node is regarded as a supporter of state S. The contribution of the neighbor node to the current node to be matched belonging to state is defined as: the state determination confidence of the neighbor node itself multiplied by the pseudo-resistance value similarity between the current node to be matched and the neighbor node. Sum the contributions of all neighbor nodes supporting state , i.e. the current node to be matched belongs to state Overall support The calculation formula is: ,in For all known states And the set of already determined state nodes adjacent to the current node to be matched, For adjacent nodes with already determined states The degree of certainty about one's own state For the current node to be matched and its adjacent nodes whose states have been determined The similarity of pseudo-resistance values ​​between states is considered. Since the absolute value of the overall support obtained from different states may be affected by factors such as the number of neighbors, normalization is required to convert it into comparable state confidence. Specifically, the overall support of the current node to be matched belonging to all three states is summed to obtain the total support, and then the overall support belonging to each state is divided by this total support. After normalization, the sum of the state confidences corresponding to the three states is 1. Each confidence value intuitively reflects the relative probability that the node belongs to the corresponding state in the current iteration round.

[0114] 503d. For each node to be matched in the set of nodes to be matched, if the maximum state confidence of the node to be matched is greater than the preset propagation confidence threshold, then the node to be matched is marked as a node with a determined state, and the alumina concentration state corresponding to the maximum state confidence is taken as the alumina concentration state of the node.

[0115] 503e. If the maximum state confidence of all nodes in the set of nodes to be matched is not greater than the preset propagation confidence threshold, then stop the iteration.

[0116] Specifically, for each node to be matched, the maximum value among its three state confidence scores is found. If this maximum value is greater than a preset propagation confidence threshold, the inference is considered sufficiently reliable. Subsequently, the node to be matched is officially marked as a node with a determined state, and its state is determined as the alumina concentration state corresponding to this maximum value. Simultaneously, the state determination confidence score of this node is recorded as this maximum value, serving as the confidence score for its support in subsequent iterations. The preset propagation confidence threshold here controls the strictness of state propagation; a higher threshold requires stronger certainty in the inference, resulting in a more conservative but potentially more accurate propagation process; a lower threshold leads to faster propagation but may introduce more uncertainty.

[0117] If, in a certain iteration, the maximum state confidence of all nodes to be matched does not exceed the propagation confidence threshold, it means that all unknown regions of the current propagation front cannot make sufficiently reliable inferences based on existing information. In this case, continuing the iteration may not generate new definite nodes, so the main iteration loop is stopped to avoid an infinite loop.

[0118] 503f.When the preset maximum iteration number is reached, if there are still uncertain state nodes in the control region adjacent network, the following steps are performed:

[0119] 1. For each uncertain state node, obtain the alumina concentration state of all adjacent nodes of the uncertain state node in the control region adjacent network;

[0120] 2. The alumina concentration state that appears most frequently in the adjacent nodes is taken as the alumina concentration state of the uncertain state node;

[0121] 3. The uncertain state node is marked as a determined state node.

[0122] Here, the preset maximum iteration number is to prevent the algorithm from being trapped in a long loop due to abnormal conditions. If the algorithm has not marked all nodes as determined state nodes when the maximum iteration number is reached, the above steps are used as a backup process. Specifically, for each remaining uncertain state node, collect the alumina concentration states of all its neighbor nodes. Count the frequency of occurrence of various states in the neighbor nodes, and assign the state with the highest frequency of occurrence to the uncertain state node; if there is a tie, a random choice can be made.

[0123] Through the iterative state propagation algorithm of the above S5 step, the alumina concentration states of all control regions in the electrolytic cell are gradually and robustly inferred from a small number of reliable reference region states, providing comprehensive state-aware input data for generating accurate region-independent discharge instructions.

[0124] S6. According to the alumina concentration states of all control regions, combined with the material diffusion effect between control regions and the action constraints of the discharge equipment, independent discharge control instructions for each control region are generated.

[0125] Directly controlling the simple rule of adding understock and stopping overstock according to the state may lead to control oscillation or low efficiency due to the neglect of material diffusion between regions and physical limitations of the discharge equipment. Therefore, this step realizes the balance between full-cell concentration uniformity and equipment action economy under the premise of meeting process requirements.

[0126] 1. Based on the alumina concentration state of each control region, determine the target discharge amount required by each control region;

[0127] The target feed rate is the amount of alumina to be replenished or reduced within a control cycle, and it is preset based on the zone's condition. Specifically: for zones in a deficit state, the target feed rate is a positive value, which can be set according to the severity of the deficit or determined based on the average feed rate required to eliminate similar deficit states in historical data. For zones in a normal state, the target feed rate is zero or a small, sustaining amount. For zones in an oversupply state, the target feed rate is zero, and it may be necessary to consider reducing the feed rate in subsequent cycles.

[0128] 2. Construct a feeding instruction optimization model. The optimization model takes minimizing the uneven distribution of alumina concentration in the whole tank and the total number of feeding equipment actions as the objective function, and the target feeding amount, the material diffusion effect model and the feeding equipment action constraints as the constraints.

[0129] This step formalizes the control decision as a constrained optimization problem. The objective functions include minimizing the unevenness of the alumina concentration distribution across the entire tank and minimizing the total number of feeder actions. Minimizing the unevenness of the alumina concentration distribution measures the uniformity of the expected concentration distribution across the entire tank after feeder action. This can be quantified by calculating the variance or Gini coefficient of the expected concentration in all control regions. Minimizing this value aims to make the concentration across the entire tank more uniform. Minimizing the total number of feeder actions means that one feeder action refers to one feeder performing one feed. This objective aims to reduce equipment wear and energy consumption. The expression is:

[0130]

[0131]

[0132]

[0133] In the formula: The uneven distribution of alumina concentration throughout the tank. This represents the total number of actions performed by the feeding equipment. For control area The predicted concentration, This is the average concentration. For the total number of regions, For decision variables, 1 indicates that the feeder will operate in the next cycle, and 0 indicates that it will not operate; , They are the weighting coefficients, and satisfy... .

[0134] The constraint conditions of the optimization problem include target material discharge amount constraint, material diffusion effect model and material discharge equipment action constraint. The target material discharge amount constraint is that the actual total material discharge amount of each region should be as close as possible to the target material discharge amount. The material diffusion effect model describes that the alumina discharged to a region will diffuse to adjacent regions over time, and a simplified diffusion equation is used to describe the material interaction between regions, and the model is used to predict the net effect of material discharge action on the whole tank concentration distribution. The material discharge equipment action constraint includes single maximum / minimum material discharge amount, minimum action interval time, interlocking relationship of material discharge actions between regions and maximum action times of each material discharger in a control period.

[0135] 3. Solving the material discharge instruction optimization model to obtain the material discharge time and amount of each control region in the next control period, and forming independent material discharge control instructions.

[0136] is to find the combination of decision variables that optimizes the objective function under all constraints. The decision variables include: whether each material discharger will act in the next period, when it will act and how much it will discharge. The solution output is the precise operation instruction of each material discharger in the next control period, for example: No. 1 material discharger, act at the 120th second after the start of the period, and the discharge amount is 5.0 kg. This series of instructions constitutes the final set of independent region material discharge control instructions, which are sent to the local controllers of each material discharger through the network for execution.

[0137] This step makes the whole tank concentration distribution quickly tend to be uniform through system-level optimization, while the equipment action cost is maximally saved, thereby realizing the synergistic optimization in improving current efficiency and reducing comprehensive energy consumption.

[0138] It can be understood that those skilled in the art can combine various embodiments in each of the above embodiments under the guidance of the above embodiments to obtain various technical solutions of the embodiments.

[0139] The above only describes the preferred embodiments of the present application and should not be used to limit the present application. Any modification, equivalent replacement and improvement made within the spirit and principle of the present application should be included in the protection scope of the present application.

Claims

1. A method for regionalized independent tapping control of an aluminum electrolysis cell power distribution system based on pseudo-resistance monitoring, characterized in that, The application relates to a method for determining the alumina concentration state of each control region in an aluminum electrolysis cell. The method comprises the following steps: Real-time acquisition of anode current data of each measuring point by deploying optical fiber current sensors at multiple measuring points on the anode bus of an aluminum electrolysis cell; meanwhile, measuring the cell voltage of the aluminum electrolysis cell; According to the spatial topological relationship of each measuring point, the multiple measuring points are divided into several control regions; for each control region obtained by division, the pseudo-resistance value of the control region is calculated according to the sum of the anode current data of all measuring points in the control region and the cell voltage; According to the pseudo-resistance values of the control regions and the historical change trend of the pseudo-resistance values of the control regions, a control region that meets the reference judgment condition is selected from all the control regions to form a reference region set; The reference judgment condition comprises that the pseudo-resistance value of a control region is in a preset clear state interval for a continuous preset period of time, and the fluctuation coefficient of the pseudo-resistance value is less than a preset fluctuation threshold; According to the pseudo-resistance values of the control regions in the reference region set, the alumina concentration reference state of each control region in the reference region set is determined through a preset concentration state mapping relationship; Taking the reference region set as a starting point, the alumina concentration state is propagated to all control regions through iterative matching according to the spatial proximity relationship and the similarity of the pseudo-resistance values between the control regions with an undetermined state and the reference region set, until the alumina concentration state of each control region is determined; 2. The method of claim 1, wherein the method further comprises: According to the alumina concentration state of all control regions, combined with the material diffusion effect between control regions and the action constraint of the discharging equipment, independent discharging control instructions of each control region are generated. The method comprises the following steps: A multi-dimensional feature vector is constructed for each measuring point, and the multi-dimensional feature vector comprises the spatial position coordinates of the measuring point, the identifier of the discharger to which the measuring point belongs and the current response intensity of the measuring point to the adjacent discharging action based on historical data statistics; A two-stage clustering division is performed, which comprises the following steps: Based on the spatial position coordinates and the discharger identifier, the multiple measuring points are initially coarsely divided to form multiple initial candidate regions; 3. The method of claim 2, wherein the method further comprises: For each initial candidate region, the initial candidate region is internally merged or split optimized based on the current response intensity of each measuring point in the region, and several control regions are obtained. The internal merging or splitting optimization of the initial candidate region comprises the following steps: The variance of the current response intensity feature between all measuring points in each initial candidate region is calculated as a consistency index of the initial candidate region; The difference degree of the current response intensity feature between adjacent measuring points at the boundary of any two adjacent initial candidate regions is calculated as a boundary coupling degree index between the two adjacent initial candidate regions; When the consistency index of an initial candidate region is lower than a first preset threshold, the initial candidate region is kept intact; When the consistency index of an initial candidate region is higher than the first preset threshold, the initial candidate region is split into at least two new control regions. When the coupling degree index of the boundary between two adjacent initial candidate regions is higher than a second preset threshold, the two adjacent initial candidate regions are merged into one control region.

4. The pseudo-resistance monitoring based regionalized independent tapping control method for power distribution system of aluminum reduction cells according to any one of claims 1 to 3, characterized in that, The filtering of the control regions satisfying the benchmark determination condition from all the control regions comprises: For each control region, a state determination confidence score is calculated according to the depth of the pseudo-resistance value of the control region falling into the explicit state interval within the continuous preset number of periods and the matching degree of the pseudo-resistance variation trend of the control region with the typical variation mode of the state interval; All the control regions with the state determination confidence score higher than a preset confidence threshold are determined as the benchmark region set.

5. The method of claim 1, wherein, The preset concentration state mapping relationship is defined by preset alumina concentration state division thresholds; the alumina concentration state division thresholds comprise pseudo-resistance value thresholds corresponding to the under-amount state, the normal state and the over-amount state of the alumina concentration state.

6. The method of claim 5, wherein the method further comprises: The determination of the alumina concentration benchmark state of each control region in the benchmark region set according to the pseudo-resistance value of the control region comprises: The pseudo-resistance value of each control region in the benchmark region set is compared with the alumina concentration state division thresholds continuously; The alumina concentration benchmark state of the control region is determined as one of the under-amount state, the normal state and the over-amount state according to the specific interval into which the pseudo-resistance value falls in the pseudo-resistance value interval defined by the division thresholds.

7. The method of claim 1, wherein, The propagation of the alumina concentration state to all the control regions by iterative matching according to the spatial proximity relationship and the similarity of the pseudo-resistance values between the control regions with undetermined states and the benchmark region set, starting from the benchmark region set, until the alumina concentration state of each control region is determined, comprises: A control region adjacency network is constructed, each control region is taken as a node in the control region adjacency network, and a connection edge is established between the corresponding two nodes if two control regions are adjacent in space; The nodes corresponding to each control region in the benchmark region set are marked as determined state nodes, and the alumina concentration benchmark state of each control region in the benchmark region set is taken as the known state of the corresponding node; The following steps are repeatedly executed until all the nodes in the control region adjacency network are marked as determined state nodes or a preset maximum number of iterations is reached: All the nodes adjacent to the determined state nodes and with undetermined states are selected from the control region adjacency network to form a set of nodes to be matched; For each node to be matched in the set of nodes to be matched, the pseudo-resistance value similarity between the node to be matched and each adjacent determined state node is calculated; For each node to be matched in the set of nodes to be matched, the state confidence of the node to be matched belonging to each alumina concentration state is calculated based on the pseudo-resistance value similarity between the node to be matched and each adjacent determined state node and the alumina concentration state of each adjacent determined state node. For each of the to-be-matched nodes in the to-be-matched node set, if the maximum state confidence of the to-be-matched node is greater than a preset propagation confidence threshold, the to-be-matched node is marked as a determined state node, and the alumina concentration state corresponding to the maximum state confidence is taken as the alumina concentration state of the node; If the maximum state confidence of all to-be-matched nodes in the to-be-matched node set is not greater than the preset propagation confidence threshold, the iteration is stopped.

8. The method of claim 7, wherein the method further comprises: The calculation of the state confidence of the to-be-matched node belonging to each alumina concentration state comprises: According to the alumina concentration states of each adjacent determined state node of the to-be-matched node and the pseudo-resistance value similarity between the to-be-matched node and each adjacent determined state node, the comprehensive support degree of the to-be-matched node belonging to each alumina concentration state is calculated; The comprehensive support degree is normalized to obtain the state confidence of the to-be-matched node belonging to each alumina concentration state.

9. The method of claim 7, wherein the method further comprises: When the repeated execution step is terminated due to reaching the preset maximum iteration number, if there are still undetermined state nodes in the control region adjacent network, the following steps are performed: For each undetermined state node, the alumina concentration states of all adjacent nodes of the undetermined state node in the control region adjacent network are obtained; The alumina concentration state with the highest occurrence frequency among the adjacent nodes is taken as the alumina concentration state of the undetermined state node; The undetermined state node is marked as a determined state node.

10. The method of claim 1, wherein, The generation of the independent unloading control instruction of each control region according to the alumina concentration states of all control regions, in combination with the material diffusion effect between control regions and the action constraint of the unloading equipment, comprises: Based on the alumina concentration states of each control region, the target unloading amount required by each control region is determined; An unloading instruction optimization model is constructed, the optimization model taking the minimization of the unevenness of the overall tank alumina concentration distribution and the total action number of the unloading equipment as an objective function, and taking the target unloading amount, the material diffusion effect model and the unloading equipment action constraint as constraint conditions; The unloading instruction optimization model is solved to obtain the unloading time and unloading amount of each control region in the next control period, forming an independent unloading control instruction.

Images