Energy-saving aeration method and system for sewage treatment

Abstract

The present application relates to sewage treatment and intelligent control technical field, disclose a kind of sewage treatment energy-saving aeration method and system, the method includes obtaining the space information data of biochemical pool, space grid division and space mapping are carried out to the space information data, generate a plurality of independent grid body and the physical mapping matrix of each valve;Local biochemical state parameters of a plurality of independent grid body are obtained, and normalized operation processing is carried out, to obtain local supply-demand characteristic value;Local supply-demand characteristic value clustering analysis is carried out, to obtain the grid set of different oxygen consumption levels;Grid set is carried out oxygen consumption trend prediction, to obtain oxygen consumption demand;Oxygen consumption demand is compared with threshold value, to generate bottom valve adjustment instruction, and a plurality of independent grid bodies are controlled to valve by combining physical mapping matrix, complete global gas supply regulation.This method realizes on-demand gas supply and energy saving by the independent adjustment of bottom valve and the space grid state perception of biochemical pool.

Description

Technical Field

[0001] This invention relates to the field of wastewater treatment and intelligent control technology, and in particular to an energy-saving aeration method and system for wastewater treatment. Background Technology

[0002] The aeration stage in the biological treatment tank is the core process for maintaining microbial activity and degrading pollutants in wastewater treatment. Its energy consumption accounts for the majority of the entire treatment process, and achieving energy conservation and consumption reduction in this stage is of vital importance to water environmental protection.

[0003] In one existing technology, aeration control of the biological treatment tank mainly relies on treating the entire tank as a homogeneous whole and uniformly adjusting the global air supply based on data from a few monitoring points within the tank. While this method can maintain a basic dissolved oxygen level under ideal conditions, its implementation typically relies solely on global average values ​​for air supply decisions, lacking effective perception and response to the dynamic distribution characteristics of dissolved oxygen and sludge concentration in different areas of the tank. Essentially, it represents an empirical, extensive process control approach. Under the influence of complex flow fields and differences in biochemical reactions, the influent end of the biological treatment tank exhibits high pollutant concentrations and high oxygen demand, while at the end, pollutants have been largely degraded, resulting in a sharp decrease in oxygen demand. A simple, uniform air supply strategy struggles to accurately allocate air supply to address the differentiated oxygen demands of different areas, leading to a supply-demand mismatch where dissolved oxygen at the influent end is insufficient and oxygen at the end is severely excessive. This exposes the technical weakness of the bottom regulating valves, which lack independent control over different spatial grids.

[0004] In summary, existing technologies suffer from problems such as unbalanced gas supply distribution, substandard water quality in some areas, and high aeration energy consumption due to a lack of accurate acquisition of the spatial state of each part of the biological treatment tank and the ability to independently adjust the bottom valves. Summary of the Invention

[0005] This invention provides an energy-saving aeration method and system for wastewater treatment, which solves the problems of unbalanced air supply distribution, substandard local water quality, and high aeration energy consumption in existing technologies due to the lack of accurate acquisition of the local spatial state of the biological treatment tank and the ability to independently adjust the bottom valve.

[0006] In a first aspect, to solve the above-mentioned technical problems, the present invention provides an energy-saving aeration method for wastewater treatment, comprising: Acquire spatial information data of the biochemical pool, perform spatial grid division and spatial mapping on the spatial information data, and generate a physical mapping matrix between multiple independent grid bodies and each valve; The local biochemical state parameters of the multiple independent grid cells are obtained, and the ratio calculation and normalization of the local biochemical state parameters are performed to obtain the local supply and demand mismatch characteristic value. Based on the local supply and demand mismatch characteristic value, cluster analysis is performed on the multiple independent grids to obtain the initial grid classification result. The oxygen consumption level is then calibrated on the initial grid classification result to obtain a grid set with different oxygen consumption levels. Historical gas supply allocation data of the corresponding grid is obtained according to the grid set, and oxygen consumption trend is predicted based on the historical gas supply allocation data to obtain the predicted oxygen consumption demand of the multiple independent grid bodies. The predicted oxygen consumption demand is compared with the preset oxygen consumption threshold to obtain the grid status determination result. Based on the grid state determination result, a bottom valve opening adjustment command is generated, and the valves of the multiple independent grid bodies are controlled in conjunction with the physical mapping matrix to complete the global gas supply adjustment of the biological tank.

[0007] In a second aspect, the present invention provides an energy-saving aeration system for wastewater treatment, comprising: a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor executes the program to implement the steps of the method described above.

[0008] Thirdly, the present invention provides a computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements the steps of the method described above.

[0009] Compared with the prior art, the present invention has the following beneficial effects: (1) This invention discretizes the overall space of the biochemical pool into multiple independent grid bodies with precise three-dimensional coordinate attributes by dividing the three-dimensional spatial data into spatial grids and calculating the coordinates of the center point, thereby realizing the refined characterization and precise positioning of the internal space of the biochemical pool. This method breaks through the limitation of traditional technology that treats the biochemical pool as a uniform whole and cannot perceive local spatial differences, so that the physical positions of different areas in the biochemical pool can be accurately distinguished and independently described, thus providing a spatial positioning basis for subsequent local biochemical state perception and independent gas supply control; (2) This invention uses cluster analysis and oxygen consumption level calibration on local supply and demand mismatch characteristics to divide multiple independent grids into grid sets with different oxygen consumption levels, thereby realizing spatial partitioning and level quantification of oxygen consumption characteristics in the biochemical tank. This technology establishes a classification association between the degree of local supply and demand mismatch and the spatial location of the grid, enabling the differentiated oxygen consumption demand in different areas of the biochemical tank to be accurately identified and classified, effectively avoiding the problem of mismatch between gas supply allocation and local demand caused by the lack of local supply and demand status perception or unified gas supply strategy in the prior art; (3) This invention obtains corresponding historical gas supply allocation data based on grid sets with different oxygen consumption levels, and predicts the oxygen consumption trend of the historical gas supply allocation data, thereby realizing dynamic prediction and quantitative output of the future oxygen consumption demand of each independent grid. This method introduces historical gas supply patterns and the oxygen consumption level characteristics of each grid into the prediction process, overcoming the response lag and supply-demand mismatch problems caused by relying solely on current real-time monitoring values ​​or fixed empirical parameters for gas supply decisions in traditional methods, and improving the accuracy and foresight of gas supply pre-allocation; (4) This invention constructs a valve opening adjustment command generation mechanism based on the grid state determination results, and combines it with a physical mapping matrix to independently control the valves of multiple independent grids, thereby realizing global gas supply regulation of the biochemical tank. This technology tightly couples the determination of local oxygen consumption status with valve execution control, enabling the system to make differentiated and precise allocation of gas supply according to the real-time supply and demand status of each grid, thus overcoming the problems of local supply and demand mismatch and energy waste caused by relying on global unified regulation or static allocation rules in traditional methods, and significantly improving the accuracy and timeliness of gas supply regulation. Attached Figure Description

[0010] Figure 1 This is a schematic diagram of a wastewater treatment energy-saving aeration method provided in the first embodiment of the present invention; Figure 2 This is a two-dimensional spatial scatter plot of the grid center and the valve provided in the first embodiment of the present invention; Figure 3 This is a three-dimensional spatial scatter plot of the grid center and the valve provided in the first embodiment of the present invention; Figure 4 This is a typical grid oxygen consumption trend prediction line chart provided in the first embodiment of the present invention; Figure 5 This is a two-dimensional spatial state pseudo-color image for determining the oxygen consumption state of a grid, provided in the first embodiment of the present invention. Detailed Implementation

[0011] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0012] Reference Figure 1 The first embodiment of the present invention provides an energy-saving aeration method for wastewater treatment, comprising the following steps: S1, acquire spatial information data of the biochemical pool, perform spatial grid division and spatial mapping on the spatial information data, and generate a physical mapping matrix between multiple independent grid bodies and each valve; S2, obtain the local biochemical state parameters of the multiple independent grid cells, perform ratio calculation and normalization on the local biochemical state parameters to obtain the local supply and demand mismatch characteristic value; S3. Based on the local supply and demand mismatch characteristic value, perform cluster analysis on the multiple independent grids to obtain the initial grid classification result, and calibrate the oxygen consumption level of the initial grid classification result to obtain a grid set with different oxygen consumption levels. S4. Obtain historical gas supply allocation data for the corresponding grid according to the grid set, and predict the oxygen consumption trend of the historical gas supply allocation data to obtain the predicted oxygen consumption demand of the multiple independent grid bodies. Compare the predicted oxygen consumption demand with the preset oxygen consumption threshold to obtain the grid status determination result. S5. Generate a bottom valve opening adjustment command based on the grid state determination result, and combine it with the physical mapping matrix to control the valves of the multiple independent grid bodies to complete the global gas supply adjustment of the biochemical tank.

[0013] In step S1, spatial information data of the biochemical pool is acquired, and spatial grid division and spatial mapping are performed on the spatial information data to generate multiple independent grid bodies and physical mapping matrices of each valve, including: Acquire spatial information data of the biochemical pool, including three-dimensional spatial data of the biochemical pool and pipeline topology coordinate data of each valve; The three-dimensional spatial data is mapped to obtain the three-dimensional structure mapping data of the biochemical pool, and the three-dimensional structure mapping data is divided into spatial grids using the octree voxelization algorithm to obtain multiple independent grids of the biochemical pool. The center point coordinates of all independent meshes are traversed and calculated to obtain the three-dimensional coordinate properties of all independent meshes. The KD-tree nearest neighbor search algorithm is used to spatially match the three-dimensional coordinate attributes with the pipeline topology coordinate data to generate spatial distance association results between the biochemical tank and each valve. Based on the spatial distance correlation results, control correlation operations are performed on the multiple independent grid bodies to obtain the physical mapping matrix of each valve.

[0014] In one implementation, the specific steps for acquiring spatial information data of the biological treatment tank are as follows: First, this embodiment extracts the tank's boundary contour, internal spatial vertices, and tank depth parameters by parsing the building information model file of the biological treatment tank, and determines the extracted data as the three-dimensional spatial data of the biological treatment tank. Then, this embodiment retrieves the pipeline structure database from the sewage treatment plant's pipeline network control system, extracts the equipment number, pipeline connection relationship, and spatial installation position parameters corresponding to each aeration valve, and determines the spatial installation position parameters corresponding to each valve as pipeline topology coordinate data.

[0015] It should be noted that the building information model file is generated during the construction and design phase of the biochemical tank and stored in the engineering information management system of the wastewater treatment plant. The building information model file includes tank size parameters, spatial structure parameters, and internal equipment layout parameters.

[0016] In one implementation, the specific operation for obtaining the three-dimensional structure mapping data is as follows: Using a preset reference point at the bottom of the biochemical pool as the origin, a unified spatial coordinate transformation operation is performed on the three-dimensional spatial data. This maps spatial coordinates from different sources to a unified Cartesian coordinate system, resulting in the three-dimensional structure mapping data of the biochemical pool. The unified spatial coordinate transformation operation is calculated using an affine transformation matrix. By performing translation, rotation, and scale normalization on the spatial coordinates, coordinate offset errors between different modeling data are eliminated.

[0017] Furthermore, the preset reference point is preset according to the fixed structural boundary of the biochemical pool. The intersection of the bottom boundary of the biochemical pool is selected as the origin of the spatial coordinates, and the X-axis, Y-axis and Z-axis coordinate directions are established in the length direction, width direction and depth direction of the pool body, respectively, thereby constructing a unified spatial rectangular coordinate system.

[0018] In one implementation, this embodiment utilizes an octree voxelization algorithm to perform spatial meshing on the three-dimensional structure mapping data, obtaining multiple independent meshes for the biochemical pool. Specifically, this embodiment first reads the pool boundary range parameters from the three-dimensional structure mapping data, constructs an initial cubic region capable of completely encompassing the biochemical pool space based on the maximum spatial dimensions of the pool in the X-axis, Y-axis, and Z-axis directions, and determines the initial cubic region as the root node of the octree.

[0019] Furthermore, in this embodiment, the size of the spatial region corresponding to the root node is determined, and the side length of the current spatial region is calculated to see if it is greater than a preset grid side length threshold. If the side length of the current spatial region is greater than the preset grid side length threshold, a spatial bisection operation is simultaneously performed along the center positions of the X-axis, Y-axis, and Z-axis directions of the current spatial region, recursively dividing the current spatial region into eight sub-spatial regions of equal volume, and establishing each sub-spatial region as a corresponding child node. Subsequently, the side length determination and recursive division operation are repeated for the spatial regions corresponding to each child node. If the side length of the current spatial region is less than or equal to the preset grid side length threshold, the division of the current spatial region is stopped, and the current spatial region is output as a corresponding independent grid body, thereby obtaining multiple independent grid bodies covering the entire biochemical pool space.

[0020] It should be noted that the preset grid side length threshold is an empirical parameter pre-set based on the size of the biochemical pool and the accuracy requirements of the fluid dynamics simulation. For example, for a medium-sized biochemical pool, the preset grid side length threshold is set to 0.5 meters. This value can ensure that the spatial resolution meets the local gas supply control requirements while keeping the total number of grid cells within a reasonable range, thus avoiding excessive computational complexity.

[0021] It is worth noting that in this embodiment, the smallest spatial region after recursive partitioning is determined as an independent grid cell, thereby obtaining multiple independent grid cells for the biochemical pool.

[0022] In one implementation, this embodiment calculates the center point coordinates of all independent meshes to obtain their 3D coordinate attributes. Specifically, this embodiment iterates through the boundary vertex coordinates of each independent mesh, calculating the average coordinates of each boundary vertex along the X, Y, and Z axes to obtain the geometric center coordinates of the corresponding independent mesh. This embodiment then stores the geometric center coordinates of each independent mesh in a structured manner to obtain the 3D coordinate attributes of all independent meshes.

[0023] In one implementation, this embodiment utilizes a KD-tree nearest neighbor search algorithm to spatially match the three-dimensional coordinate attributes with the pipeline topology coordinate data, obtaining the spatial distance association results between the biochemical tank and each valve. Specifically, this embodiment constructs a KD-tree spatial index structure from the pipeline topology coordinate data. Using the three-dimensional coordinate attributes of each independent grid cell as the query point, it searches for the nearest neighbor valve node in the KD-tree spatial index structure. It calculates the three-dimensional Euclidean distance between the center point of each independent grid cell and the coordinates of each valve node, and determines the association based on a preset spatial search tolerance radius, thus obtaining the spatial distance association results.

[0024] Furthermore, this embodiment determines the association affiliation based on a preset spatial search tolerance radius. Specifically, this embodiment compares the three-dimensional Euclidean distance between each independent grid cell and the valve in the spatial distance association results with the preset spatial search tolerance radius. When the three-dimensional Euclidean distance is less than or equal to the preset spatial search tolerance radius, it is determined that there is a control association relationship between the independent grid cell and the valve, and the corresponding distance value is retained as valid association data; when the three-dimensional Euclidean distance is greater than the preset spatial search tolerance radius, it is determined that there is no direct control association between the independent grid cell and the valve, and the corresponding distance value is marked as invalid.

[0025] It is worth noting that the preset spatial search tolerance radius is set to 0.8 meters. This value is determined based on the matching relationship between the effective range of the valve and the grid side length. It is used to limit the search boundary for spatial position matching and ensure the rationality of the matching results.

[0026] In one implementation, a physical mapping matrix between the grid and the valve is determined. In this embodiment, based on the valid association data after association attribution determination in the spatial distance association results, a distance inverse weighting function is used to calculate the weight coefficient of each independent grid cell controlled by the corresponding valve. The weight coefficients and the corresponding valve numbers are then matrix-organized to construct the physical mapping matrix between the grid and the valve.

[0027] The formula for calculating the inverse distance weighting function is as follows: , In the formula, To control the weights, The distance is the three-dimensional Euclidean distance between the center point of independent mesh i and the node coordinates of valve j. To find all valve sets whose three-dimensional Euclidean distance from independent mesh i is less than or equal to a preset spatial search tolerance radius. For the valve index in the valve set; when the valve Not a set hour, =0.

[0028] It is worth noting that the dimension of the physical mapping matrix is ​​the total number of independent grid cells multiplied by the total number of valves. Each element in the matrix records the control weight of the corresponding independent grid cell to the corresponding valve. When a grid cell and a valve are determined to have no direct control relationship, the corresponding control weight is reset to zero. This matrix accurately records which specific valves control each independent grid cell and their intensity, providing a spatial mapping basis for subsequently converting upper-level fluid analysis results into control commands for lower-level equipment.

[0029] For example, by parsing the Building Information Model (BIM) file, spatial vertex coordinates are extracted to obtain three-dimensional spatial structure data with a length of 50.5 meters, a width of 20.0 meters, and a depth of 8.5 meters. Simultaneously, by connecting to the underlying pipeline network geographic information database, a depth-first search algorithm is used to accurately obtain the pipeline topology coordinate data of twenty-four aeration valves. For example, the core node coordinates of valve number 12 are extracted as (15.5, 8.0, 0.2). An affine transformation algorithm is used to map the spatial data to a standard spatial rectangular coordinate system with the southwest corner of the bottom of the biochemical tank as the origin. An octree quantization algorithm is called to set the grid side length to 0.5 meters, dividing the biochemical tank space into 68,680 independent grids. All independent grids are traversed, and the center point coordinates of each grid are calculated using the geometric center integral formula, for example, generating a center point spatial attribute with coordinates (1.25, 2.75, 4.25). The pipeline topology coordinate data is constructed into a KD-tree spatial index structure, with a spatial search tolerance radius set to 0.8 meters. The association is determined by calculating the three-dimensional Euclidean distance between the grid center point and the valve coordinates. Based on the distance calculation results, an inverse distance weighting function is used to automatically assign influence coefficients. For example, when the grid center is 0.3 meters away from the valve, a control weight of 0.85 is assigned, thereby constructing a sparse physical mapping matrix with a dimension of 68,680 x 24.

[0030] like Figure 2 As shown, this figure is a two-dimensional spatial scatter plot of the grid center and valves; it shows the relative positional relationship between the regular grid system (blue dots) established in the planar space of the biological tank and the bottom aeration valves (red triangles); the dense blue dot matrix in the figure represents high-resolution independent grids, while the scattered red triangles represent physical air supply valves, intuitively presenting the complex many-to-many mapping relationship between the grid and the valves.

[0031] like Figure 3 As shown, this figure is a three-dimensional spatial scatter plot of the grid center and valves, which shows the overall spatial mapping of the grid system and valves in the three-dimensional biochemical tank from a three-dimensional perspective. The X and Y axes in the figure constitute the tank plane, and the Z axis (height) reflects the depth information of the tank. This verifies that the mapping algorithm can accurately calculate the control coverage range of the bottom valve on the upper three-dimensional spatial grid across different heights and positions.

[0032] In step S2, obtaining the local biochemical state parameters of the multiple independent grid cells, and performing ratio calculations and normalization on the local biochemical state parameters to obtain local supply-demand mismatch characteristic values ​​includes: The local biochemical state parameters of the multiple independent grid cells are obtained, including dissolved oxygen, sludge concentration and oxygen uptake rate; The local biochemical state parameters are subjected to element-by-element ratio calculations to obtain the ratio results; The ratio results are logarithmically scaled and normalized to obtain the local supply-demand mismatch characteristic value for each independent grid cell.

[0033] In one implementation, this embodiment obtains the local biochemical state parameters of the multiple independent grid cells from sensors built into the biochemical tank. Specifically, it calls upon dissolved oxygen sensors and sludge concentration sensors deployed inside the biochemical tank to collect monitoring data in real time, obtaining biochemical state monitoring data at multiple discrete locations within the biochemical tank. Subsequently, this embodiment uses a common Kriging spatial interpolation algorithm to perform spatial interpolation calculations on the biochemical state monitoring data at the multiple discrete locations, mapping the discrete monitoring data to the spatial region corresponding to each independent grid cell, thereby obtaining the dissolved oxygen and sludge concentration corresponding to each independent grid cell.

[0034] Further, the spatial coordinates and monitoring values ​​corresponding to multiple discrete sampling points are first obtained. The data sources for these discrete sampling points include data collected by dissolved oxygen sensors, sludge concentration sensors, and mobile monitoring devices fixed within the biological treatment tank. When the number of fixed monitoring points is insufficient, a spatial sample point set is constructed by combining historical operational data. Subsequently, the semivariogram function value is calculated based on the spatial distance between each sampling point and the monitoring values, and an experimental semivariogram curve is constructed. The experimental semivariogram curve is fitted using a spherical model to obtain spatial correlation parameters. Then, using the geometric center coordinates of the independent grid cell to be estimated as the target location, a Kriging weighting equation system is established and solved based on the spatial correlation parameters to obtain the spatial weight coefficients corresponding to each sampling point. Finally, the monitoring values ​​of each sampling point are weighted and summed with the corresponding spatial weight coefficients to obtain the dissolved oxygen or sludge concentration corresponding to the target independent grid cell.

[0035] It should be noted that the number of sampling points used in the ordinary Kriging spatial interpolation algorithm is determined based on the size of the biochemical pool and the required spatial resolution. When the number of fixed online sensors is insufficient, a spatial sample point set can be constructed by combining data collected by mobile inspection and monitoring equipment, valid monitoring data from historical operation databases, or model estimation data to meet the sample quantity requirements for semi-variogram fitting.

[0036] It is worth noting that the oxygen absorption rate is obtained by calculating the dissolved oxygen concentration change data of each independent grid cell during continuous monitoring in this embodiment. Specifically, the dissolved oxygen concentration values ​​of the same independent grid cell at two adjacent sampling times are obtained, and the rate of change of dissolved oxygen concentration per unit time is calculated; the absolute value of the rate of change of dissolved oxygen concentration is determined as the oxygen absorption rate of the corresponding independent grid cell. This calculation method assumes that the change in the aeration mass transfer term is small or that aeration has been shut down in advance within adjacent sampling intervals, and is applicable to periods when the dissolved oxygen in the biological treatment tank shows a downward trend. Since the decrease in dissolved oxygen concentration mainly reflects the oxygen consumption process of microorganisms, the oxygen absorption rate can characterize the actual oxygen consumption level of the corresponding area. By performing the above calculation on each independent grid cell, the oxygen absorption rate corresponding to each independent grid cell is obtained.

[0037] In one implementation, this embodiment performs element-wise ratio calculations on the local biochemical state parameters to obtain the ratio results. Specifically, this embodiment constructs a dissolved oxygen matrix, a sludge concentration matrix, and an oxygen uptake rate matrix corresponding to each independent grid cell, and performs element-wise division operations on the data at the same positions in the three matrices to obtain the ratio results.

[0038] Further, the specific operation process of the element-by-element division operation is as follows: First, the dissolved oxygen content of each independent grid cell is constructed into a dissolved oxygen content matrix according to the spatial arrangement of the grid cells; the sludge concentration is constructed into a sludge concentration matrix; and the oxygen absorption rate is constructed into an oxygen absorption rate matrix. All three matrices have the same dimension, which is one times the total number of independent grid cells, and the same matrix position corresponds to the same independent grid cell. Then, for each matrix position, the three values ​​corresponding to that position—dissolved oxygen content, sludge concentration, and oxygen absorption rate—are extracted. The dissolved oxygen content is divided by the sludge concentration to obtain the ratio of dissolved oxygen to sludge concentration. This ratio is then divided by the oxygen absorption rate to obtain the ratio result for that independent grid cell. After performing the above division operation on all matrix positions, the ratio results for each position are arranged according to the original matrix positions to obtain the ratio result matrix.

[0039] It should be noted that the element-by-element ratio calculation is used to quantify the relative balance between oxygen supply and microbial oxygen demand within each independent grid cell. A low ratio indicates an insufficient oxygen supply in the current independent grid cell, while a high ratio indicates a local oxygen surplus in the current independent grid cell.

[0040] In one implementation, this embodiment performs logarithmic scaling and normalization on the ratio result to obtain the local supply-demand mismatch characteristic value for each independent grid cell. Specifically, this embodiment first adds a preset positive offset to the ratio result and then performs a base-10 logarithmic scaling operation to compress the numerical span between different grids and reduce the impact of local abnormal fluctuations on the overall result. The preset positive offset is used to avoid the situation where the logarithmic function is undefined when the ratio result is zero; its value can be set to 0.001. Subsequently, range normalization is performed on the logarithmically scaled data to map the data corresponding to each grid cell to a numerical range of 0 to 1, obtaining the local supply-demand mismatch characteristic value for each independent grid cell.

[0041] It should be noted that the range normalization process is performed by extracting the maximum and minimum values ​​from the data corresponding to all independent grid cells, and then calculating the range by subtracting the minimum value from the current data and dividing by the difference between the maximum and minimum values. This eliminates numerical differences between data of different dimensions. When the maximum and minimum values ​​are equal, it is determined that the current data does not exhibit numerical dispersion. In this case, the normalization results corresponding to all independent grid cells are uniformly set to a preset benchmark value of 0.5, which is the median of the normalization range, to avoid calculation anomalies caused by a denominator of zero.

[0042] For example, this embodiment acquires biochemical status monitoring data through dissolved oxygen sensors and sludge concentration sensors deployed inside the biological treatment tank, and maps the discrete monitoring data to 68,680 independent grid cells using a standard Kriging spatial interpolation algorithm. For instance, the dissolved oxygen level corresponding to independent grid cell numbered 3500 is 2.4 mg / L, and the sludge concentration is 3800.0 mg / L. This embodiment performs element-wise division on the dissolved oxygen level and sludge concentration to obtain the corresponding ratio. Subsequently, logarithmic scaling and range normalization are performed on the ratio, and the final output value for the local supply-demand mismatch characteristic corresponding to independent grid cell numbered 3500 is 0.72.

[0043] In step S3, based on the local supply-demand mismatch characteristic values, cluster analysis is performed on the multiple independent grid cells to obtain an initial grid classification result. The initial grid classification result is then calibrated for oxygen consumption levels to obtain a set of grids with different oxygen consumption levels, including: The local supply-demand mismatch feature values ​​are matrixed to obtain the grid feature matrix; The contour coefficient analysis is performed on the grid feature matrix to determine the number of clusters corresponding to the K-means clustering algorithm, and the cluster centers are initialized according to the number of clusters to obtain an initial set of cluster centers; Calculate the minimum feature distance between the multiple independent grid cells and the initial cluster center set, and cluster the multiple independent grid cells according to the minimum feature distance to obtain the initial grid classification result; The cluster centers are iteratively updated based on the initial grid classification results, and the spatial grid clustering partitioning operation is repeated to obtain stable clustering results; Based on the stable clustering results, oxygen consumption levels are calibrated to obtain grid sets with different oxygen consumption levels.

[0044] In one implementation, the determination of the grid feature matrix is ​​as follows: In this embodiment, the local supply-demand mismatch feature values ​​corresponding to all independent grid cells are arranged sequentially according to their corresponding numbering order, and the local supply-demand mismatch feature values ​​are combined and stored in row vector form to obtain the corresponding grid feature matrix.

[0045] It should be noted that each matrix element in the grid feature matrix corresponds to a local supply-demand mismatch feature value of an independent grid cell, which is used to characterize the degree of oxygen supply-demand imbalance within the corresponding independent grid cell.

[0046] In one implementation, this embodiment performs silhouette coefficient analysis on the grid feature matrix to obtain an initial set of cluster centers. Specifically, this embodiment sets different numbers of candidate clusters and performs K-means clustering on the grid feature matrix based on each candidate cluster number. Then, it calculates the average intra-cluster distance and the nearest inter-cluster distance for each independent grid cell, and calculates the corresponding silhouette coefficient based on these distances. The average silhouette coefficients corresponding to different candidate cluster numbers are compared, and the candidate cluster number with the largest average silhouette coefficient is determined as the target cluster number. Afterward, based on the target cluster number, a corresponding number of initial feature values ​​are randomly selected in the feature space as cluster centers to obtain the initial set of cluster centers.

[0047] It should be noted that the silhouette coefficient is used to characterize the tightness and separation of the clustering results. A silhouette coefficient closer to 1 indicates a higher classification and discrimination ability of the current clustering result. The number of candidate clusters is preset according to the oxygen consumption area division requirements of the biological treatment tank, with a range of 2 to 6. By calculating the average silhouette coefficient corresponding to each candidate cluster number, the candidate cluster number with the largest average silhouette coefficient is selected as the target cluster number. Since the oxygen consumption state in the biological treatment tank can usually be divided into different levels such as low oxygen consumption, medium oxygen consumption, high oxygen consumption, and extremely high oxygen consumption, limiting the range of candidate cluster numbers to 2 to 6 can balance clustering accuracy and computational efficiency.

[0048] In one implementation, this embodiment determines the initial grid classification result by calculating the minimum feature distance between the plurality of independent grid cells and the initial cluster center set. Specifically, this embodiment calculates the Euclidean distance between the local supply-demand mismatch feature value corresponding to each independent grid cell and each initial cluster center, and extracts the target cluster center with the smallest distance. Then, the number of the current independent grid cell is written into the cluster index list to which the corresponding target cluster center belongs, establishing a classification relationship between the current independent grid cell and the corresponding cluster center. Afterwards, the remaining independent grid cells are traversed, and the distance calculation and cluster assignment operations are repeated until all independent grid cells have been clustered, thus obtaining the initial grid classification result.

[0049] Furthermore, the Euclidean distance is obtained by calculating the absolute value of the difference between the feature value of the current independent grid cell and the feature value of the corresponding cluster center, and is used to characterize the feature similarity between the current independent grid cell and the corresponding cluster center.

[0050] In one implementation, this embodiment obtains a stable clustering result by iteratively updating the cluster centers and repeatedly performing spatial grid clustering operations. Specifically, this embodiment calculates the local supply-demand mismatch feature values ​​corresponding to all independent grid cells within each cluster and calculates the average value of all feature values ​​within each cluster. The average value is then used to update the new cluster centers for the corresponding clusters, and distance calculation and clustering operations are re-performed based on the updated cluster centers. Subsequently, it is determined whether the numerical offset between two consecutive cluster centers is less than a preset clustering convergence threshold; if so, the iteration stops and a stable clustering result is output; otherwise, the cluster center update and spatial grid clustering operations continue.

[0051] It should be noted that the preset clustering convergence threshold is used to determine whether the cluster centers are stable. When the system is running for the first time or when historical data is insufficient, an adaptive initialization strategy is adopted to extract the standard deviation of all local supply-demand mismatch feature values ​​in the current grid feature matrix and initialize the preset clustering convergence threshold to 0.01 times this standard deviation. The proportional coefficient 0.01 is determined based on the empirical convergence accuracy of the K-means algorithm under standard normal distribution data. After the system obtains sufficient historical clustering data, the mean of the absolute difference of the cluster centers in the last 5 iterations is statistically analyzed, and 5% of this mean is used as the new convergence threshold.

[0052] In one implementation, this embodiment calibrates the oxygen consumption level based on the stable clustering results to obtain grid sets with different oxygen consumption levels. Specifically, this embodiment sorts all clusters in ascending order according to the magnitude of the cluster center values. Then, the cluster with the smallest cluster center value is calibrated as the low oxygen consumption level grid set, the cluster with the second smallest cluster center value is calibrated as the medium oxygen consumption level grid set, the cluster with the second largest cluster center value is calibrated as the high oxygen consumption level grid set, and the cluster with the largest cluster center value is calibrated as the extremely high oxygen consumption level grid set, thus obtaining grid sets with different oxygen consumption levels.

[0053] For example, this embodiment performs matrix processing on the local supply-demand mismatch feature values ​​corresponding to 68,680 independent grid cells to obtain a grid feature matrix. This embodiment tests the silhouette coefficients corresponding to cluster numbers of 2, 3, 4, and 5, and determines that the average silhouette coefficient is highest when the number of clusters is 4. Subsequently, four cluster centers are initialized in the feature value space, with values ​​of 0.15, 0.45, 0.65, and 0.85, respectively. This embodiment further calculates the Euclidean distance between each independent grid cell and each cluster center, and performs clustering based on the minimum distance. For example, the local supply-demand mismatch feature value corresponding to independent grid cell number 3500 is 0.72. The system calculates that its distance to the cluster center (0.65) is the smallest, therefore it is assigned to the corresponding cluster. Subsequently, this embodiment continuously performs cluster center updates and clustering operations. Iteration stops when the cluster center offset is less than 0.001 for two consecutive times, ultimately obtaining the grid sets corresponding to low oxygen consumption levels, medium oxygen consumption levels, high oxygen consumption levels, and extremely high oxygen consumption levels.

[0054] In step S4, the process of obtaining historical gas supply allocation data for the corresponding grid based on the grid set, predicting oxygen consumption trends based on the historical gas supply allocation data to obtain the predicted oxygen consumption demand for the multiple independent grid cells, and comparing the predicted oxygen consumption demand with a preset oxygen consumption threshold to obtain a grid status determination result includes: Historical gas supply allocation data for the corresponding grid is obtained from the grid set, and the historical gas supply allocation data is processed into a time series to obtain historical gas supply sequence data. The historical gas supply sequence data is input into a long short-term memory network for time series feature learning to obtain gas supply time series feature data. The gas supply time series feature data is then used to predict oxygen consumption trends to obtain predicted oxygen demand data. Spatial grid mapping is performed on the predicted oxygen demand data to obtain the predicted oxygen demand corresponding to the multiple independent grid cells; The predicted oxygen demand is compared with the preset first oxygen demand threshold and second oxygen demand threshold to obtain the oxygen consumption status judgment result. If the predicted oxygen demand is greater than the first oxygen demand threshold, it is determined that the corresponding independent grid cell is in an oxygen-deficient state, and an oxygen-deficient grid cell is obtained. If the predicted oxygen demand is less than the second oxygen demand threshold, then it is determined that the corresponding independent grid cell has an end-excess state, and an excess state grid is obtained; The hypoxia-deficient state grid and the excess state grid are subjected to state labeling and aggregation processing to obtain the grid state determination result.

[0055] In one implementation, this embodiment performs time-series processing on the grid set to obtain historical gas supply sequence data. Specifically, this embodiment extracts the gas supply allocation records of the corresponding independent grid body during its historical operation, based on the oxygen consumption level grid set to which each independent grid body belongs. The gas supply allocation records include valve opening data and aeration flow rate data. This embodiment sorts the gas supply allocation records in chronological order and completes missing time nodes according to a unified sampling period to obtain historical gas supply sequence data.

[0056] It should be noted that the unified sampling period is set according to the data acquisition frequency of the aeration control system. In this embodiment, the unified sampling period is set to 5 minutes, that is, the air supply distribution status data of the corresponding independent grid cell is recorded every 5 minutes.

[0057] In one implementation, this embodiment inputs the historical gas supply sequence data into a Long Short-Term Memory (LSTM) network for temporal feature learning to obtain gas supply temporal feature data. Specifically, this embodiment constructs an LTM network model and inputs the historical gas supply sequence data into the LTM network model in chronological order. A forgetting gate is used to filter out redundant information in the historical sequence, an input gate is used to update the valid gas supply information at the current time node, and an output gate generates the corresponding hidden state vector. The hidden state vectors corresponding to each time node are combined to obtain the gas supply temporal feature data.

[0058] Further, the construction process of the Long Short-Term Memory (LSTM) network model involves acquiring historical gas supply allocation data during the continuous operation of the biochemical pool and constructing training samples in chronological order. Specifically, historical gas supply allocation data for 24 consecutive hours is used as the input sequence, and the actual oxygen demand detected in the subsequent two hours is used as the corresponding label data. The historical gas supply allocation data reflects the actual oxygen demand change trend of the grid, as the gas supply has been adjusted to follow load fluctuations during past operation. After normalizing the historical gas supply allocation data, the data is divided according to the proportions of the training set, validation set, and test set. This embodiment constructs an LTM network model comprising an input layer, a first hidden layer, a second hidden layer, and an output layer. The first hidden layer contains 128 neurons, the second hidden layer contains 64 neurons, and the output layer outputs the predicted oxygen demand value. The Adam optimizer is used to optimize the parameters of the LTM network model, with a learning rate set to 0.001, and mean squared error is used as the loss function for iterative training. During training, the model parameters are updated based on the changes in the loss function on the validation set. When the change in the loss function is less than a preset loss threshold (e.g., 0.001) during ten consecutive training rounds, training stops, and the current model is identified as the completed Long Short-Term Memory (LSTM) network model. Subsequently, real-time historical gas supply sequence data is input into the completed LTM network model, which outputs the predicted oxygen demand for the corresponding independent grid cells.

[0059] It should be noted that the preset loss threshold is set according to the model training convergence accuracy requirements. Specifically, it is determined by statistically analyzing the change value of the loss function corresponding to ten consecutive rounds of model training, and setting 10% of the average of the change values ​​as the preset loss threshold.

[0060] In one implementation, this embodiment predicts oxygen consumption trends based on the gas supply time-series characteristic data to obtain predicted oxygen demand data. Specifically, this embodiment inputs the gas supply time-series characteristic data into the output layer of a long short-term memory network model, and calculates the trend of oxygen demand changes within a preset prediction period through forward propagation. Subsequently, the predicted oxygen demand values ​​for the corresponding time nodes are output to obtain the predicted oxygen demand data.

[0061] It should be noted that the preset prediction period is set based on the adjustment response time of the aeration control system. In this embodiment, the preset prediction period is set to 2 hours to predict the changes in oxygen demand in the corresponding area within the next 2 hours.

[0062] In one implementation, this embodiment performs spatial grid mapping on the predicted oxygen demand data to obtain the predicted oxygen demand corresponding to the multiple independent grid cells. Specifically, this embodiment extracts the grid number information corresponding to the predicted oxygen demand data and establishes a one-to-one mapping relationship according to the independent grid cell numbers. Subsequently, the predicted oxygen demand values ​​are written into the data storage unit of the corresponding independent grid cell, so that each independent grid cell is associated with the corresponding predicted oxygen demand data, thereby obtaining the predicted oxygen demand corresponding to multiple independent grid cells.

[0063] In one implementation, this embodiment compares the predicted oxygen demand with preset first and second oxygen demand thresholds to obtain an oxygen consumption status judgment result. Specifically, this embodiment iterates through the predicted oxygen demand data corresponding to each independent grid cell and extracts the corresponding predicted oxygen demand value; then, the predicted oxygen demand value is compared with the preset first and second oxygen demand thresholds to obtain the corresponding oxygen consumption status judgment result.

[0064] It should be noted that the first oxygen consumption threshold and the second oxygen consumption threshold are set based on the actual oxygen demand distribution of each area during the historical operation of the biological treatment tank. Specifically, the actual oxygen demand data corresponding to each independent grid cell in the historical operation data of the biological treatment tank are obtained, and the actual oxygen demand data is statistically analyzed; the lower limit of the values ​​in the top 10% of the actual oxygen demand data distribution is determined as the first oxygen consumption threshold, which is used to characterize the criteria for judging high oxygen consumption areas; the upper limit of the values ​​in the bottom 10% of the actual oxygen demand data distribution is determined as the second oxygen consumption threshold, which is used to characterize the criteria for judging low oxygen consumption areas.

[0065] In one implementation, this embodiment performs state labeling and aggregation processing on the hypoxic state grid and the excess state grid to obtain the grid state determination result. Specifically, this embodiment extracts the grid number and spatial coordinate information corresponding to the hypoxic state grid and the excess state grid, respectively. Then, a hypoxic state label is assigned to the hypoxic state grid, and an excess state label is assigned to the excess state grid. Afterward, the corresponding grids are classified and aggregated according to the state label type, and a data table containing grid number, spatial coordinates, and state label is constructed to obtain the grid state determination result.

[0066] For example, the system first connects to an underlying distributed time-series database. For grid sets at four levels—low, medium, high, and extremely high—historical gas supply allocation data sampled every five minutes over the past thirty days is extracted, covering valve opening percentage and instantaneous aeration flow rate. Then, a Long Short-Term Memory (LSTM) network is constructed, comprising an input layer, two hidden layers with 128 and 64 neurons respectively, and an output layer. Taking grid number 3500, which belongs to the high oxygen consumption level set, as an example, its historical gas supply flow rate sequence over the past 72 hours (864 time steps) is input into the network. A forget gate filters redundant fluctuation noise, the input gate updates the effective gas supply characteristics at the current moment, and the output gate calculates the hidden state vector. After forward propagation and matrix operations, the oxygen consumption lag effect of this grid during the daily peak hours is captured. The output layer decodes and calculates the predicted oxygen demand of this individual grid cell for the next two hours to be 4.25 mg / L. After mapping the prediction results back to each independent grid cell, the predicted oxygen demand is compared with preset thresholds. The predicted value of 4.25 for this grid cell is greater than the first oxygen demand threshold of 4.0, and it is determined to be an influent anoxic state. However, the predicted oxygen demand of grid cell number 1200 is 1.12 mg / L per hour, which is less than the second oxygen demand threshold of 1.5, and it is determined to be an end-point excess state. Finally, a grid state determination result vector table containing spatial coordinates and corresponding state labels is generated to provide a decision-making basis for subsequent valve opening adjustment.

[0067] like Figure 4 As shown, this figure is a line graph of oxygen consumption trend prediction for a typical grid. It shows the trend of historical oxygen consumption demand (blue solid line) and predicted oxygen consumption demand (red dashed line) for a typical grid over time. The figure marks the high threshold (4.0 mg / L) and low threshold (1.5 mg / L). The blue line shows a clear upward trend in the later stage and approaches the high threshold, while the red line accurately fits this trend, intuitively demonstrating the system's ability to predict future oxygen consumption trends based on historical data.

[0068] like Figure 5 As shown, this figure is a two-dimensional spatial pseudo-color map of the oxygen consumption status determination of the grid. It presents the oxygen consumption status determination results of the entire biological tank in a planar space in the form of a pseudo-color map. The different colored areas in the figure, red represents hypoxia / overexposure, green represents normal, and blue represents excess, clearly delineating the supply and demand mismatch in different areas of the tank, and intuitively presenting the global status distribution map under the independent zone control strategy.

[0069] In step S5, generating a bottom valve opening adjustment command based on the grid state determination result, and combining it with the physical mapping matrix to control the valves of the multiple independent grid cells to complete the global gas supply adjustment of the biological tank, includes: The grid state determination result is input into a pre-trained support vector machine for feature mapping processing to obtain the target aeration adjustment amount corresponding to the multiple independent grid cells. Based on the target aeration adjustment amount, calculate the bottom valve opening adjustment data corresponding to the multiple independent grid cells; Based on the physical mapping matrix, the bottom valve opening adjustment data of the multiple independent grid bodies associated with the same valve are merged and encapsulated into data frames to obtain the final opening command of each valve. According to the final opening command, the bottom valves corresponding to the multiple independent grid bodies are opened to control the opening of the valves, thereby completing the global gas supply regulation of the biological tank.

[0070] In one implementation, this embodiment inputs the grid state determination result into a pre-trained support vector machine model for feature mapping processing to obtain the target aeration adjustment amount. Specifically, this embodiment extracts the state label, oxygen consumption level, and predicted oxygen demand data corresponding to each independent grid cell, and combines them to form a grid state feature vector. Subsequently, the grid state feature vector is input into the pre-trained support vector machine model for nonlinear mapping operation to obtain the corresponding target aeration adjustment amount.

[0071] It should be noted that the construction process of the support vector machine model is as follows: historical operating data corresponding to each independent grid cell during more than three consecutive months of operation of the biological treatment tank are obtained, and a training sample set is constructed in chronological order, with a total number of training samples of no less than 10,000 sets. The grid status label, oxygen consumption level, and predicted oxygen consumption demand are numerically encoded to form the input feature vector, where the influent anoxic state is encoded as 1, the normal state is encoded as 0, and the terminal excess state is encoded as -1. The low oxygen consumption level, medium oxygen consumption level, high oxygen consumption level, and extremely high oxygen consumption level are encoded as 1, 2, 3, and 4, respectively. The predicted oxygen consumption demand uses the actual value at the corresponding time. The valve opening adjustment amount recorded during the actual operation is used as the training label data to achieve regression prediction of the continuous value of aeration compensation. The valve opening adjustment amount is extracted from the historical operating log of the programmable logic controller of the bottom valve. The period when the dissolved oxygen is stably controlled within the target range in the historical data is selected. At this time, the valve opening adjustment amount can be regarded as the correct response to the current oxygen consumption demand. The recording format is the difference in the opening percentage before and after each adjustment. A radial basis function (RBF) was selected as the kernel function, and a grid search algorithm was used to optimize the penalty parameter C and the kernel parameter γ. The search range for the penalty parameter C was set to 1 to 100, and the search range for the kernel parameter γ was set to 0.001 to 1. Five-fold cross-validation was then used to determine the parameter combination with the minimum validation error. Subsequently, a support vector machine (SVM) model was trained using the optimal parameter combination, and the trained model was deployed on the biochemical pool control server. Upon receiving the real-time grid state determination result, the corresponding feature vector was input into the SVM model for online inference calculation, outputting the aeration compensation amount in continuous numerical form to obtain the target aeration adjustment amount.

[0072] In one implementation, this embodiment adjusts the bottom valve opening of the multiple independent grid cells according to the target aeration adjustment amount to obtain bottom valve opening adjustment data. Specifically, this embodiment obtains the current valve opening data corresponding to each independent grid cell and establishes a corresponding conversion relationship between the target aeration adjustment amount and the valve opening. Then, it calculates the target valve opening value based on the target aeration adjustment amount of the corresponding independent grid cell and determines the target valve opening value as the bottom valve opening adjustment data.

[0073] It should be noted that the conversion relationship between the target aeration adjustment amount and the valve opening was established through offline calibration experiments. Specifically, before the biological treatment tank was put into operation, the bottom valve was adjusted stepwise in 5% increments, gradually increasing from 0% to 100%, resulting in 21 calibration sampling points. At each calibration sampling point, the aeration system was kept running stably for more than 3 minutes, and the corresponding actual aeration flow rate data was recorded. A mapping table between valve opening and aeration flow rate was established based on all calibration sampling points. When the aeration flow rate value corresponding to the target aeration adjustment amount is between two adjacent calibration sampling points, the corresponding target valve opening is calculated using linear interpolation, and the calculated target valve opening is used as the basis for valve opening conversion.

[0074] In one implementation, this embodiment merges and encapsulates the bottom valve opening adjustment data of multiple independent grid cells associated with the same valve into data frames based on the physical mapping matrix to obtain the final opening command for each valve. Specifically, this embodiment calls the physical mapping matrix to retrieve the correspondence between each independent grid cell and the bottom valve, and identifies multiple independent grid cells controlled by the same bottom valve. Then, it extracts the bottom valve opening adjustment data corresponding to the multiple independent grid cells and performs a weighted merging calculation according to a preset weight to obtain the target opening value for the corresponding bottom valve. Afterwards, it combines the target opening value with the corresponding valve address, device number, and control timestamp, and encapsulates it into data frames according to a preset industrial communication protocol format to generate the final opening command for each valve.

[0075] It should be noted that the preset weights are determined based on the control association strength in the physical mapping matrix. Specifically, when establishing the physical mapping matrix, the control association weights are calculated based on the spatial distance between the center of each independent grid cell and the corresponding valve; the closer the spatial distance, the greater the corresponding control association weight. When merging the opening adjustment data, the weighted average of the control association weights corresponding to each independent grid cell and the opening adjustment data is used as the target opening value to improve the matching degree between the valve control results and the actual gas supply demand.

[0076] It is worth noting that the preset industrial communication protocol format is the Modbus TCP industrial communication protocol format. Specifically, the device address of the corresponding valve is used as the slave address field, the target opening value is converted into corresponding register data, and a control message is constructed in the order of transaction identifier, protocol identifier, data length field, device address field, function code field, and data field. The function code is used to characterize the valve opening write operation, and the data field is used to store the target opening value. After assembling all fields, a standard Modbus TCP data frame is generated, and this data frame is sent as the final opening command to the programmable logic controller of the corresponding bottom valve.

[0077] In one implementation, this embodiment controls the opening of the bottom valves corresponding to the multiple independent grid cells according to the final opening command, thereby completing the global gas supply regulation of the biological treatment tank. Specifically, this embodiment sends the final opening command to the programmable logic controller (PLC) of the corresponding bottom valve. The PLC parses the device address and target opening value in the data frame and locates the corresponding bottom valve actuator. Subsequently, it outputs a control signal according to the target opening value, driving the bottom valve to perform an opening adjustment action, adjusting the valve opening to the corresponding target opening state. After the valve adjustment is completed, the gas supply flow corresponding to each independent grid cell is redistributed according to the updated valve opening, resulting in an updated local gas supply distribution result. Finally, the gas supply distribution results corresponding to all independent grid cells are summarized to form the overall gas supply distribution state of the biological treatment tank, completing the global gas supply regulation of the biological treatment tank.

[0078] It should be noted that the updated local air supply allocation result is determined based on the target opening value of each bottom valve and the mapping relationship between valve opening and aeration flow rate. Specifically, based on the adjusted target opening of each bottom valve, the corresponding aeration flow rate data is queried and allocated to the corresponding independent grid cells in conjunction with the physical mapping matrix, thereby obtaining the updated local air supply allocation result for each independent grid cell.

[0079] For example, in this embodiment, the status label "influent anoxic state," oxygen consumption level "high oxygen consumption level," and predicted oxygen demand of 4.25 mg / L corresponding to the independent grid cell numbered 3500 are input into a pre-trained support vector machine model for feature mapping processing, resulting in a target aeration adjustment of 125 cubic meters of aeration flow per hour. Subsequently, based on the mapping relationship between valve opening and aeration flow, the corresponding bottom valve opening adjustment data is calculated to be 85.5%. Further, the bottom valve is simultaneously associated with independent grid cells numbered 3500, 3510, and 3520 through a physical mapping matrix, and the corresponding bottom valve opening adjustment data are merged and calculated to obtain the final target opening of the valve as 82.3%. Then, the valve device address, target opening value, and function code information are encapsulated into a final opening instruction and sent to the corresponding programmable logic controller to drive the bottom valve to complete the opening adjustment, thereby redistributing the air supply to the associated area and ultimately achieving dynamic optimization and adjustment of the overall air supply to the biological treatment tank.

[0080] In summary, this invention discloses an energy-saving aeration method for wastewater treatment. By constructing a three-dimensional spatial grid of the biological treatment tank and its physical mapping relationship with the valves, combined with local supply-demand mismatch feature analysis, oxygen consumption zone clustering, oxygen demand prediction, and grid state determination, it achieves accurate identification of the oxygen demand status of different areas. Furthermore, by generating valve opening adjustment commands through a support vector machine, it independently controls the bottom valves of each area, realizing precise allocation and dynamic adjustment of aeration volume. This effectively solves the problems of inaccurate air supply allocation, local supply-demand imbalance, and high energy consumption in traditional aeration control methods, thereby improving the operating efficiency and energy-saving effect of the wastewater treatment process.

[0081] The second embodiment of the present invention provides an energy-saving aeration system for wastewater treatment, comprising: a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor executes the program to implement the steps of the method described above.

[0082] It should be noted that the wastewater treatment energy-saving aeration system provided in this embodiment of the invention is used to execute all the process steps of the wastewater treatment energy-saving aeration method of the above embodiment. The working principles and beneficial effects of the two are one-to-one, so they will not be described again.

[0083] It should be noted that the system embodiments described above are merely illustrative. The units described as separate components may or may not be physically separate, and the components shown as units may or may not be physical units; that is, they may be located in one place or distributed across multiple network units. Some or all of the modules can be selected to achieve the purpose of this embodiment according to actual needs. Those skilled in the art can understand and implement this without any creative effort.

[0084] The specific embodiments described above further illustrate the purpose, technical solution, and beneficial effects of the present invention. It should be understood that the above descriptions are merely specific embodiments of the present invention and are not intended to limit the scope of protection of the present invention. In particular, it should be noted that any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the scope of protection of the present invention for those skilled in the art.

Claims

1. An energy-saving aeration method for wastewater treatment, characterized in that, include: Acquire spatial information data of the biochemical pool, perform spatial grid division and spatial mapping on the spatial information data, and generate a physical mapping matrix between multiple independent grid bodies and each valve; The local biochemical state parameters of the multiple independent grid cells are obtained, and the ratio calculation and normalization of the local biochemical state parameters are performed to obtain the local supply and demand mismatch characteristic value. Based on the local supply and demand mismatch characteristic value, cluster analysis is performed on the multiple independent grids to obtain the initial grid classification result. The oxygen consumption level is then calibrated on the initial grid classification result to obtain a grid set with different oxygen consumption levels. Historical gas supply allocation data of the corresponding grid is obtained according to the grid set, and oxygen consumption trend is predicted based on the historical gas supply allocation data to obtain the predicted oxygen consumption demand of the multiple independent grid bodies. The predicted oxygen consumption demand is compared with the preset oxygen consumption threshold to obtain the grid status determination result. Based on the grid state determination result, a bottom valve opening adjustment command is generated, and the valves of the multiple independent grid bodies are controlled in conjunction with the physical mapping matrix to complete the global gas supply adjustment of the biological tank. The step of obtaining the local biochemical state parameters of the multiple independent grid cells, and performing ratio calculations and normalization on the local biochemical state parameters to obtain local supply-demand mismatch characteristic values ​​includes: The local biochemical state parameters of the multiple independent grid cells are obtained, including dissolved oxygen, sludge concentration and oxygen uptake rate; The local biochemical state parameters are subjected to element-by-element ratio calculations to obtain the ratio results; The ratio results are logarithmically scaled and normalized to obtain the local supply-demand mismatch characteristic value of each independent grid cell; Specifically, the step of performing element-by-element ratio calculations on the local biochemical state parameters to obtain the ratio results involves: constructing a dissolved oxygen matrix by arranging the dissolved oxygen content of each independent grid cell according to the spatial order of the grid; constructing a sludge concentration matrix by arranging the sludge concentration; and constructing an oxygen absorption rate matrix by arranging the oxygen absorption rate. The dissolved oxygen matrix, sludge concentration matrix, and oxygen absorption rate matrix have the same dimension and are all the total number of independent grid cells multiplied by one column, with the same matrix position corresponding to the same independent grid cell. For each matrix position, the dissolved oxygen content, sludge concentration, and oxygen absorption rate corresponding to that position are extracted. The dissolved oxygen content is divided by the sludge concentration to obtain a first ratio, and then the first ratio is divided by the oxygen absorption rate to obtain the ratio result for that independent grid cell. After traversing all matrix positions, the ratio results are arranged according to their corresponding matrix positions to obtain a ratio result matrix. The process of performing logarithmic scaling and normalization on the ratio results to obtain the local supply-demand mismatch characteristic value for each independent grid cell is as follows: After adding a preset positive offset to each ratio result in the ratio result matrix, a logarithmic scaling operation with base 10 is performed. The range normalization process is then performed on the logarithmically scaled data to map each data to a numerical range of 0 to 1, thereby obtaining the local supply and demand mismatch feature value corresponding to each independent grid cell.

2. The energy-saving aeration method for wastewater treatment according to claim 1, characterized in that, The process of acquiring spatial information data of the biochemical pool, performing spatial grid division and spatial mapping on the spatial information data, and generating multiple independent grid bodies and physical mapping matrices between each valve includes: Acquire spatial information data of the biochemical pool, including three-dimensional spatial data of the biochemical pool and pipeline topology coordinate data of each valve; The three-dimensional spatial data is mapped to obtain the three-dimensional structure mapping data of the biochemical pool, and the three-dimensional structure mapping data is divided into spatial grids using the octree voxelization algorithm to obtain multiple independent grids of the biochemical pool. The center point coordinates of all independent meshes are traversed and calculated to obtain the three-dimensional coordinate properties of all independent meshes. The pipeline topology coordinate data is spatially matched and controlled by the three-dimensional coordinate attributes to obtain the physical mapping matrix of each valve.

3. The energy-saving aeration method for wastewater treatment according to claim 2, characterized in that, The step of matching and associating the pipeline topology coordinate data with the three-dimensional coordinate attributes to obtain the physical mapping matrix of each valve includes: The KD-tree nearest neighbor search algorithm is used to spatially match the three-dimensional coordinate attributes with the pipeline topology coordinate data to generate spatial distance association results between the biochemical tank and each valve. Based on the spatial distance correlation results, control correlation operations are performed on the multiple independent grid bodies to obtain the physical mapping matrix of each valve.

4. The energy-saving aeration method for wastewater treatment according to claim 1, characterized in that, The initial grid classification result is obtained by clustering the multiple independent grid cells based on the local supply-demand mismatch characteristic value. The initial grid classification result is then calibrated for oxygen consumption level to obtain a grid set with different oxygen consumption levels, including: The local supply-demand mismatch feature values ​​are matrixed to obtain the grid feature matrix; The contour coefficient analysis is performed on the grid feature matrix to determine the number of clusters corresponding to the K-means clustering algorithm, and the cluster centers are initialized according to the number of clusters to obtain an initial set of cluster centers; Calculate the minimum feature distance between the multiple independent grid cells and the initial cluster center set, and cluster the multiple independent grid cells according to the minimum feature distance to obtain the initial grid classification result; The cluster centers are iteratively updated based on the initial grid classification results, and the spatial grid clustering partitioning operation is repeated to obtain stable clustering results; Based on the stable clustering results, oxygen consumption levels are calibrated to obtain grid sets with different oxygen consumption levels.

5. The energy-saving aeration method for wastewater treatment according to claim 4, characterized in that, The process involves obtaining historical gas supply allocation data for the corresponding grid based on the grid set, predicting oxygen consumption trends from the historical gas supply allocation data to obtain the predicted oxygen consumption demand for the multiple independent grid cells, and comparing the predicted oxygen consumption demand with a preset oxygen consumption threshold to obtain a grid status determination result, including: Historical gas supply allocation data for the corresponding grid is obtained from the grid set, and the historical gas supply allocation data is processed into a time series to obtain historical gas supply sequence data. The historical gas supply sequence data is input into a long short-term memory network for time series feature learning to obtain gas supply time series feature data. The gas supply time series feature data is then used to predict oxygen consumption trends to obtain predicted oxygen demand data. Spatial grid mapping is performed on the predicted oxygen demand data to obtain the predicted oxygen demand corresponding to the multiple independent grid cells; The predicted oxygen demand is compared with the preset oxygen demand threshold to obtain the grid state determination result.

6. The energy-saving aeration method for wastewater treatment according to claim 5, characterized in that, The step of comparing the predicted oxygen demand with a preset oxygen demand threshold to obtain a grid state determination result includes: The predicted oxygen demand is compared with the preset first oxygen demand threshold and second oxygen demand threshold to obtain the oxygen consumption status judgment result. If the predicted oxygen demand is greater than the first oxygen demand threshold, it is determined that the corresponding independent grid cell is in an oxygen-deficient state, and an oxygen-deficient grid cell is obtained. If the predicted oxygen demand is less than the second oxygen demand threshold, then it is determined that the corresponding independent grid cell has an end-excess state, and an excess state grid is obtained; The hypoxia-deficient state grid and the excess state grid are subjected to state labeling and aggregation processing to obtain the grid state determination result.

7. The energy-saving aeration method for wastewater treatment according to claim 1, characterized in that, The step of generating a bottom valve opening adjustment command based on the grid state determination result, and combining it with the physical mapping matrix to control the valves of the multiple independent grid cells, thereby completing the global gas supply adjustment of the biological tank, includes: The grid state determination result is input into a pre-trained support vector machine for feature mapping processing to obtain the target aeration adjustment amount corresponding to the multiple independent grid cells. Based on the target aeration adjustment amount, calculate the bottom valve opening adjustment data corresponding to the multiple independent grid cells; Based on the physical mapping matrix, the bottom valve opening adjustment data of the multiple independent grid bodies associated with the same valve are merged and encapsulated into data frames to obtain the final opening command of each valve. According to the final opening command, the bottom valves corresponding to the multiple independent grid bodies are opened to control the opening of the valves, thereby completing the global gas supply regulation of the biological tank.

8. A wastewater treatment energy-saving aeration system, comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, characterized in that, When the processor executes the program, it implements the steps of the method according to any one of claims 1 to 7.

9. A computer-readable storage medium having a computer program stored thereon, characterized in that, When executed by a processor, the program implements the steps of the method described in any one of claims 1 to 7.

Images