A flue gas pollution treatment cost optimization management method

Abstract

This invention discloses a method for optimizing the cost management of flue gas pollution control, comprising: Step 1: generating a set of operating units for control equipment; Step 2: generating a sequence of combined energy consumption paths; Step 3: inputting the combined energy consumption path sequence into an improved Koopa model; Step 4: generating initial path features through a path embedding module and generating steady-state evolution features through a steady-state evolution coding module; Step 5: generating rearranged path features through a disturbance track rearrangement module using a smoke load tidal migration mechanism and generating a path cost evolution sequence through a cost trend output module; Step 6: performing weighted splitting of path affiliation to generate single-path control cost units; Step 7: performing emission contribution mapping to generate a path contribution sequence and using the MCTS algorithm for tree search to generate a control path adjustment set; Step 8: outputting the cost optimization results. This invention combines the improved Koopa model and the MCTS algorithm to achieve dynamic optimization of flue gas control paths.

Description

Technical Field

[0001] This invention relates to the field of industrial big data analysis and resource optimization management technology, and in particular to a method for optimizing the management of flue gas pollution control costs. Background Technology

[0002] With the increasing demand for flue gas pollution control in industries such as thermal power, steel, chemicals, and waste incineration, cost optimization and control technologies for the coordinated operation of multiple treatment devices, including desulfurization, denitrification, dust removal, and induced draft ventilation, have received widespread attention. Existing flue gas treatment systems mainly rely on fixed operating parameter adjustments or experience-based control methods based on the operating status of single devices for cost management. However, the following problems commonly exist in actual operation:

[0003] There are obvious coupling relationships among desulfurization units, denitrification units, dust removal units, and induced draft equipment in flue gas treatment systems. The operating parameters of different treatment devices will generate dynamic propagation effects with changes in flue gas load. Existing technologies usually only adjust individual devices independently, lacking continuous correlation analysis of the coupling paths of multiple devices. This results in weak synergistic adjustment capabilities among different treatment devices, and is prone to the problem of local device energy saving but overall treatment cost increase. Operating parameters such as flue gas flow rate, sulfur dioxide concentration, nitrogen oxide concentration, and reagent consumption have obvious time-series fluctuation characteristics. Existing time-series analysis methods cannot simultaneously take into account the linkage relationship between devices and the propagation relationship of operational disturbances. This leads to insufficient ability to identify cost evolution trends under high flue gas load disturbance conditions, which can easily cause deviations in path cost assessment results. For the multi-path cost optimization problem in complex flue gas treatment systems, traditional scheduling methods based on fixed rules or single optimization are difficult to dynamically search and jointly optimize treatment paths in continuous operating time windows. They cannot achieve the coordinated screening of low power consumption, low reagent consumption, and low load fluctuation paths, resulting in high power consumption, low reagent utilization, and large equipment load fluctuations in the long-term operation of flue gas treatment systems.

[0004] Therefore, how to provide a method for optimizing the cost management of flue gas pollution control is a problem that urgently needs to be solved by those skilled in the art. Summary of the Invention

[0005] One objective of this invention is to propose a cost optimization management method for flue gas pollution control. This invention combines an improved Koopa model and the MCTS algorithm to perform dynamic evolution analysis and tree search optimization of the joint consumption paths in the flue gas control system, thereby achieving coordinated adjustment and time-sharing control among different control devices. It has the advantages of low control cost, high path identification accuracy, strong joint consumption coordination, and high operational stability.

[0006] A method for optimizing the cost management of flue gas pollution control according to an embodiment of the present invention includes the following steps:

[0007] Step 1: Collect flue gas treatment operation data of the target flue gas treatment system during its operating cycle, and perform equipment segment mapping processing to generate a set of treatment equipment operation units;

[0008] Step 2: Perform time-series operational correlation analysis on the set of treatment equipment operating units, construct an equipment joint consumption correlation diagram, and generate a joint consumption path sequence;

[0009] Step 3: Input the joint cost path sequence into the improved Koopa model, which includes a path embedding module, a steady-state evolution coding module, a disturbance track rearrangement module, and a cost trend output module;

[0010] Step 4: Map the path state of the joint-consumption path sequence through the path embedding module to generate initial path features, and perform continuous time-period evolution coding on the initial path features through the steady-state evolution coding module to generate steady-state evolution features;

[0011] Step 5: The disturbance track rearrangement module rearranges the path evolution order of the steady-state evolution characteristics through the smoke load tidal migration mechanism, generates rearranged path characteristics, and performs time-period trend analysis on the rearranged path characteristics through the cost trend output module to generate a path cost evolution sequence.

[0012] Step 6: Based on the path cost evolution sequence, perform path-attribution weighted decomposition of reagent consumption, power consumption, and equipment load changes to generate single-path governance cost units;

[0013] Step 7: Perform emission contribution mapping on the single-path governance cost unit to generate a path contribution sequence, and use the MCTS algorithm to perform tree search based on the path cost evolution sequence and the path contribution sequence to generate a governance path adjustment set;

[0014] Step 8: Generate a time-sharing governance schedule table based on the governance path adjustment set, and output the cost optimization results based on the time-sharing governance schedule table.

[0015] Optionally, step one specifically includes:

[0016] Collect flue gas treatment operation data of the target flue gas treatment system during the operating cycle. The flue gas treatment operation data includes desulfurization tower slurry circulation volume, denitrification ammonia injection flow rate, dust collector pressure difference, induced draft fan load power, flue gas flow rate, sulfur dioxide concentration, nitrogen oxide concentration, power consumption, and reagent consumption.

[0017] The flue gas treatment operation data is time-synchronized according to a uniform time interval to generate a time-series operation data group.

[0018] Based on the process connection sequence of the desulfurization unit, denitrification unit, dust removal unit, and induced draft equipment in the flue gas treatment system, the time-series operation data group is divided into equipment segments to generate equipment segment data groups.

[0019] Extract the device identifier, runtime index, and corresponding runtime parameters from each device segment data group, and establish the device segment correspondence;

[0020] Based on the corresponding relationship between the equipment segments, the data groups of each equipment segment are mapped and associated to generate a set of governance equipment operation units.

[0021] Optionally, step two specifically includes:

[0022] The operating units of each treatment equipment in the treatment equipment operating unit set are sorted by time according to the operating time index, and the operating parameters in the corresponding equipment segment are extracted according to the equipment identifier to generate a time-series operating unit sequence.

[0023] Extract the changes in desulfurization tower slurry circulation volume, denitrification ammonia injection flow rate, dust collector differential pressure, and induced draft fan load power within adjacent operating time windows in the time-series operating unit sequence;

[0024] Based on the direction of change of the corresponding operating parameters of each treatment equipment operating unit, the synchronous change relationship between different treatment equipment operating units is matched to generate equipment association node pairs;

[0025] The synchronization ratio of the direction of change of operating parameters of each associated node within a continuous operating time window is statistically analyzed.

[0026] The device-associated node pairs whose synchronization ratio reaches the set association ratio are identified as joint consumption associated nodes;

[0027] Establish node connection relationships based on the equipment identifiers and runtime indexes corresponding to each interconnected node, and construct an equipment interconnection graph.

[0028] According to the node connection direction in the equipment co-consumption association diagram, the path of each co-consumption association node is traversed, and the equipment identifier, running time index, flue gas flow rate, sulfur dioxide concentration, nitrogen oxide concentration, power consumption and reagent consumption corresponding to each co-consumption association node are recorded to generate a co-consumption path sequence.

[0029] Optionally, the step of mapping the path state of the joint-consumption path sequence through the path embedding module to generate initial path features specifically involves:

[0030] According to the node connection order in the joint consumption path sequence, the equipment identifier, running time index, flue gas flow, sulfur dioxide concentration, nitrogen oxide concentration, power consumption and reagent consumption corresponding to each joint consumption associated node are read sequentially to generate a path status data group.

[0031] Based on the device identifier, the different device segments in the path status data group are numbered and mapped to generate a device segment number sequence;

[0032] Based on the running time index, the running order of each node in the path status data group is mapped to a time position sequence to generate a time position sequence.

[0033] The flue gas flow rate, sulfur dioxide concentration, and nitrogen oxide concentration constitute the flue gas condition operation parameter group, and the power consumption and reagent consumption constitute the cost condition operation parameter group.

[0034] The operating parameters in the flue gas state operating parameter group are subjected to maximum and minimum normalization to generate the flue gas state vector; the operating parameters in the cost state operating parameter group are subjected to maximum and minimum normalization to generate the cost state vector.

[0035] The equipment segment number sequence, time location sequence, flue gas state vector, and cost state vector are concatenated according to the node connection order to generate a path state matrix.

[0036] The path state matrix is ​​vector-mapped using a fully connected mapping layer in the path embedding module to generate initial path features.

[0037] Optionally, the step of performing continuous time-period evolution coding on the initial path features through the steady-state evolution coding module to generate steady-state evolution features specifically involves:

[0038] The initial path features are divided into multiple consecutive time period feature segments according to the running time index, and arranged according to the node connection order in the joint consumption path sequence to generate a consecutive time period feature sequence;

[0039] The continuous time period feature sequence is input into the one-dimensional temporal convolutional layer in the steady-state evolution coding module, and local evolution features between adjacent time periods are extracted along the running time index direction to generate a local evolution feature sequence.

[0040] The local evolution feature sequence is input into the depth-separable convolutional layer in the steady-state evolution coding module. The channel features corresponding to each device segment are independently convolved, and the convolutional features corresponding to different device segments are combined and mapped across device segments by point-by-point convolution to generate the device segment evolution feature sequence.

[0041] The device segment evolution feature sequence is input into the layer normalization layer in the steady-state evolution coding module. The device segment evolution features under the same running time index are subjected to layer normalization processing to generate a normalized evolution feature sequence.

[0042] The normalized evolution feature sequence is input into the feedforward mapping layer in the steady-state evolution coding module. The normalized evolution features under each running time index are nonlinearly mapped by the GELU activation function to generate candidate evolution feature sequences.

[0043] The candidate evolution feature sequence and the local evolution feature sequence are added together by residual addition to generate a fused evolution feature sequence; the fused evolution features corresponding to adjacent running time indices in the fused evolution feature sequence are differentially analyzed in one dimension to generate a feature difference sequence;

[0044] Based on the positive and negative directions of the difference values ​​of each dimension in the feature difference sequence, feature dimensions that maintain the same direction of change under continuous running time index are extracted to generate a steady-state dimension label sequence.

[0045] Based on the steady-state dimension labeling sequence, the corresponding fusion evolution feature dimension is retained from the fusion evolution feature sequence, and the fusion evolution feature dimensions that are not labeled as steady-state dimensions are set to zero to generate a steady-state preservation feature sequence.

[0046] The steady-state evolution features are generated by linearly mapping the steady-state feature sequence through a linear output layer.

[0047] Optionally, the disturbance track rearrangement module rearranges the path evolution order of the steady-state evolution features through the smoke load tidal migration mechanism to generate rearranged path features, specifically:

[0048] The steady-state evolution features are input into the disturbance track rearrangement module, which is equipped with a smoke load tidal migration mechanism. Based on the running time index and equipment segment number corresponding to each feature dimension in the steady-state evolution features, the feature dimensions belonging to the same equipment segment under the same running time window are vector-concatenated according to the feature dimension order to restore the evolution features of the corresponding equipment segment.

[0049] The device segment evolution characteristics in different runtime windows are continuously arranged according to the order of runtime index to generate a time period arrangement sequence;

[0050] Extract the flue gas flow rate, power consumption, and reagent consumption corresponding to the evolution characteristics of each equipment segment in the time period sequence, and calculate the flue gas flow rate change direction, power consumption change direction, and reagent consumption change direction between adjacent operating time windows respectively.

[0051] The evolution characteristics of adjacent equipment sections whose flue gas flow rate changes in the same direction as the power consumption change are identified as the first tidal migration segment.

[0052] The evolution characteristics of adjacent equipment sections whose flue gas flow rate changes in the same direction as the reagent consumption change are identified as the second tidal migration segment.

[0053] The evolution characteristics of adjacent equipment sections whose flue gas flow rate changes are inconsistent with the direction of power consumption changes or the direction of reagent consumption changes are identified as the third tidal migration segment.

[0054] The magnitude of the change in power consumption in each first tidal migration segment is statistically analyzed, and the segments are arranged in ascending order of the magnitude of the change in power consumption.

[0055] The variation amplitude of drug consumption in each second tidal migration segment was statistically analyzed, and the segments were arranged in ascending order of the variation amplitude of drug consumption.

[0056] The amplitude of flue gas flow rate change corresponding to each third tidal migration segment is statistically analyzed, and the segments are arranged in descending order of flue gas flow rate change amplitude.

[0057] The results of the first, middle and last sections are sequentially concatenated, and the running time index corresponding to the evolution characteristics of each device segment is retained to generate rearranged path features.

[0058] Optionally, the step of performing time-period trend analysis on the rearranged path features through the cost trend output module to generate a path cost evolution sequence specifically involves:

[0059] Input the rearranged path features into the cost trend output module, and extract the equipment segment evolution features, running time index, and corresponding arrangement position marker from the rearranged path features;

[0060] Based on the arrangement position markers, corresponding path priority features are generated, where the first arrangement result corresponds to the first path priority feature, the middle arrangement result corresponds to the second path priority feature, and the last arrangement result corresponds to the third path priority feature.

[0061] Extract the power consumption, reagent consumption and flue gas flow corresponding to the evolution characteristics of each equipment segment, calculate the difference in power consumption between adjacent operating time windows to obtain the direction of power consumption change, and calculate the difference in reagent consumption between adjacent operating time windows to obtain the direction of reagent consumption change.

[0062] When both the direction of change in power consumption and the direction of change in reagent consumption are increasing, the direction of change in path cost between the corresponding path operating states is determined to be upward.

[0063] When both the direction of change in power consumption and the direction of change in reagent consumption are decreasing, the direction of change in path cost between the corresponding path operating states is determined to be decreasing.

[0064] When the direction of change in power consumption is inconsistent with the direction of change in reagent consumption, the direction of change in path cost between the corresponding path operation states is determined as the differentiation direction.

[0065] The continuous path operation status that maintains a consistent direction of path cost change is defined as the same trend evolution segment;

[0066] Record the direction of change in power consumption, the direction of change in reagent consumption, the direction of change in flue gas flow, and the corresponding equipment segment number in each trend evolution segment to generate segment cost characteristics;

[0067] The segment cost features and the corresponding path priority features are concatenated according to the feature dimension order to generate spatiotemporal correlation features;

[0068] The spatiotemporal correlation features are arranged continuously according to the running time index to generate a path cost evolution sequence.

[0069] Optionally, step six specifically includes:

[0070] Extract the spatiotemporal correlation features, running time index and corresponding equipment segment number from the path cost evolution sequence. Divide the spatiotemporal correlation features into continuous path evolution time periods based on the running time index, and determine the spatiotemporal correlation features in the same path evolution time period as the same path time period feature group.

[0071] Input the time period feature groups of each path into the fully connected mapping layer, perform feature dimension mapping on the spatiotemporal correlation features in each time period feature group of the path, and generate the corresponding path attribution value;

[0072] Softmax normalization is applied to the path attribution values ​​of each path during the same path evolution period to generate corresponding path attribution weights;

[0073] Based on the running time index within the same path evolution period, calculate the sequence of changes in power consumption, reagent consumption, and equipment load between adjacent running time indices within the path evolution period;

[0074] The path energy consumption characteristics are generated by multiplying the corresponding path attribution weight with the sequence of changes in energy consumption; the path reagent consumption characteristics are generated by multiplying the corresponding path attribution weight with the sequence of changes in reagent consumption; and the path load change characteristics are generated by multiplying the corresponding path attribution weight with the sequence of changes in equipment load.

[0075] By combining the path's power consumption characteristics, path's reagent consumption characteristics, and path's load change characteristics, a single-path governance cost unit is generated.

[0076] Optionally, step seven specifically includes:

[0077] The path energy consumption characteristics, path reagent consumption characteristics, and path load change characteristics in the single path governance cost unit are respectively subjected to maximum and minimum normalization processing to generate normalized energy consumption characteristics, normalized reagent consumption characteristics, and normalized load change characteristics.

[0078] The path contribution characteristics are generated by combining the normalized energy consumption characteristics, normalized reagent consumption characteristics, and normalized load change characteristics.

[0079] The contribution characteristics of each path are arranged consecutively according to the running time index to generate a path contribution sequence;

[0080] The path cost evolution sequence and path contribution sequence are input into the MCTS algorithm for tree search. The path contribution feature corresponding to the current running time index is used as the root node, and the path contribution features corresponding to multiple subsequent running time indices are used as candidate child nodes to generate a path search tree.

[0081] The normalized energy consumption feature, normalized reagent consumption feature, and normalized load change feature in the path contribution features of each candidate sub-node are squared and then summed. The summation result is then squared to generate the corresponding path contribution value.

[0082] When the contribution value of the path corresponding to the candidate child node is less than the contribution value of the path corresponding to the current root node, and the change in sulfur dioxide concentration and the change in nitrogen oxide concentration of the corresponding path are both less than the change in sulfur dioxide concentration and the change in nitrogen oxide concentration of the current root node, the corresponding candidate child node is determined as a low-cost candidate node.

[0083] The candidate child node with the smallest path contribution value is selected from the low-cost candidate nodes as the expansion node, and the path contribution feature corresponding to the subsequent running time index of the expansion node is extracted as the next layer candidate child node until the path search tree reaches the set search layer.

[0084] The sum of the path contribution values ​​in each path search branch is calculated, and the path search branch with the smallest sum of path contribution values ​​is taken as the governance path adjustment set.

[0085] Optionally, step eight specifically includes:

[0086] Extract the governance device paths, corresponding runtime indices, and corresponding path contribution values ​​from the governance path adjustment set, and sort the different governance device paths by time according to the runtime index to generate the governance adjustment order for the corresponding time period.

[0087] When the path contribution value of the corresponding treatment equipment path is less than the path contribution value of the previous running time index, reduce the desulfurization tower slurry circulation volume and denitrification ammonia injection flow rate in the corresponding time period, and simultaneously reduce the induced draft fan load power.

[0088] When the path contribution value of the corresponding treatment equipment path is greater than the path contribution value of the previous running time index, the circulation volume of desulfurization tower slurry and the flow rate of denitrification ammonia injection in the corresponding time period are increased, and the load power of the induced draft fan is increased simultaneously.

[0089] When the path contribution value of the corresponding treatment equipment path is the same as the path contribution value of the previous running time index, the desulfurization tower slurry circulation volume, denitrification ammonia injection flow rate and induced draft fan load power in the corresponding time period remain unchanged.

[0090] A time-sharing management and scheduling table is generated based on the desulfurization tower slurry circulation volume, denitrification ammonia injection flow rate, and induced draft fan load power for each time period.

[0091] Based on the time-sharing management and scheduling table, the changes in electricity consumption, reagent consumption, and equipment load during the corresponding time periods are statistically analyzed to generate corresponding cost statistics.

[0092] The cost statistics for each time period are summed up to output the cost optimization results.

[0093] The beneficial effects of this invention are:

[0094] This invention addresses the challenges of complex multi-device cost propagation relationships, strong operational disturbance coupling, and difficulty in continuously optimizing dynamic changes in treatment costs in flue gas treatment systems by improving the joint construction of the Koopa model and the MCTS algorithm. It employs a path embedding module, a steady-state evolution coding module, a disturbance track rearrangement module, and a cost trend output module to perform continuous time-period evolution analysis on cost propagation path sequences. In the steady-state evolution coding process, a one-dimensional temporal convolutional layer, a depthwise separable convolutional layer, and a steady-state dimension marker sequence are combined to dynamically extract the cost propagation relationships and stable operating states between different treatment devices. The disturbance track rearrangement module introduces a smoke load tidal migration mechanism, effectively reducing the interference of high smoke load disturbance paths on the path cost evolution analysis results and improving the continuity of stable cost propagation paths in the rearranged path features by differentially rearranging the first, second, and third tidal migration segments.

[0095] Meanwhile, this invention employs the MCTS algorithm for tree-like search based on the path cost evolution sequence and path contribution sequence. By continuously expanding and dynamically filtering different treatment paths through path contribution values, it achieves joint search optimization of low power consumption, low chemical consumption, and low load fluctuation paths, thereby improving the dynamic optimization capability of the treatment path adjustment set. In addition, this invention performs dynamic attribution analysis on the changes in power consumption, chemical consumption, and equipment load in different joint consumption paths through path affiliation weighted decomposition, and generates a time-sharing treatment scheduling table based on the treatment path adjustment set, realizing dynamic coordinated adjustment and cost optimization control among different treatment equipment in the flue gas treatment system. Attached Figure Description

[0096] The accompanying drawings are provided to further illustrate the invention and form part of the specification. They are used in conjunction with embodiments of the invention to explain the invention and do not constitute a limitation thereof. In the drawings:

[0097] Figure 1 This is an overall flowchart of a method for optimizing the cost management of flue gas pollution control proposed in this invention;

[0098] Figure 2 This is a schematic diagram of the improved Koopa model structure for a cost optimization management method for flue gas pollution control proposed in this invention. Detailed Implementation

[0099] The present invention will now be described in further detail with reference to the accompanying drawings. These drawings are simplified schematic diagrams, illustrating only the basic structure of the invention, and therefore only show the components relevant to the invention.

[0100] refer to Figure 1 and Figure 2 A method for optimizing the cost management of flue gas pollution control includes the following steps:

[0101] Step 1: Collect flue gas treatment operation data of the target flue gas treatment system during its operating cycle, and perform equipment segment mapping processing to generate a set of treatment equipment operation units;

[0102] Step 2: Perform time-series operational correlation analysis on the set of treatment equipment operating units, construct an equipment joint consumption correlation diagram, and generate a joint consumption path sequence;

[0103] Step 3: Input the combined cost path sequence into the improved Koopa model. The improved Koopa model includes a path embedding module, a steady-state evolution coding module, a disturbance track rearrangement module, and a cost trend output module.

[0104] Step 4: Map the path state of the joint-consumption path sequence through the path embedding module to generate initial path features, and perform continuous time-period evolution coding on the initial path features through the steady-state evolution coding module to generate steady-state evolution features;

[0105] Step 5: The disturbance track rearrangement module rearranges the path evolution order of the steady-state evolution characteristics through the smoke load tidal migration mechanism, generates rearranged path characteristics, and performs time-period trend analysis on the rearranged path characteristics through the cost trend output module to generate a path cost evolution sequence.

[0106] Step 6: Based on the path cost evolution sequence, perform path-attribution weighted decomposition of reagent consumption, power consumption, and equipment load changes to generate single-path governance cost units;

[0107] Step 7: Perform emission contribution mapping on the single-path governance cost unit to generate a path contribution sequence, and use the MCTS algorithm to perform tree search based on the path cost evolution sequence and the path contribution sequence to generate a governance path adjustment set;

[0108] Step 8: Generate a time-sharing governance schedule based on the governance path adjustment set, and output the cost optimization results based on the time-sharing governance schedule.

[0109] In this embodiment, step one specifically includes:

[0110] Collect flue gas treatment operation data of the target flue gas treatment system during the operating cycle. The flue gas treatment operation data includes the desulfurization tower slurry circulation volume, denitrification ammonia injection flow rate, dust collector pressure difference, induced draft fan load power, flue gas flow rate, sulfur dioxide concentration, nitrogen oxide concentration, power consumption, and reagent consumption.

[0111] The flue gas treatment operation data is synchronized at a uniform time interval to generate a time-series operation data group.

[0112] Based on the process connection sequence of the desulfurization unit, denitrification unit, dust removal unit and induced draft equipment in the flue gas treatment system, the time-series operation data group is divided into equipment segments to generate equipment segment data groups;

[0113] Extract the device identifier, runtime index, and corresponding runtime parameters from each device segment data group, and establish the device segment correspondence;

[0114] Based on the correspondence between equipment segments, the data groups of each equipment segment are mapped and associated to generate a set of management equipment operation units;

[0115] In the specific implementation process, flue gas treatment operation data is collected in real time using industrial sensors deployed in the denitrification reaction zone, at the front end of the dust collector, at the inlet of the desulfurization tower, and at the output end of the induced draft fan. The collection cycle for the desulfurization tower slurry circulation volume is set to 5 seconds, the collection cycle for the denitrification ammonia injection flow rate is set to 3 seconds, the collection cycle for the dust collector differential pressure is set to 2 seconds, the collection cycle for the induced draft fan load power is set to 1 second, and the collection cycles for flue gas flow rate, sulfur dioxide concentration, and nitrogen oxide concentration are uniformly set to 5 seconds. The statistical cycle for power consumption and reagent consumption is set to 60 seconds. To ensure time consistency between the operating data of different equipment, a 1-minute time window is used as a unified time window. The timestamps of each operating parameter within the same time window are aligned, and the average value of multiple operating parameter values ​​collected within the same time window is taken to generate the average operating parameter data for the corresponding operating time window. During equipment segmentation, the equipment segments are physically divided according to the flow direction of flue gas in the flue gas treatment system. Data between the ammonia injection inlet of the denitrification unit and the outlet of the denitrification reaction tower is defined as the denitrification equipment segment; data between the dust collector inlet and the dust hopper outlet is defined as the dust removal equipment segment; data between the desulfurization unit inlet and the slurry circulation pump outlet is defined as the desulfurization equipment segment; and data between the induced draft fan inlet and the chimney inlet is defined as the induced draft equipment segment. Equipment identification uses a fixed coding method: the denitrification equipment segment is coded as TXQ, the dust removal equipment segment as CCQ, the desulfurization equipment segment as DSQ, and the induced draft equipment segment as YFQ. A corresponding operating time index is appended to each operating parameter, forming a unified format for equipment segment data groups. This processing method enables a unified data association structure for the operating status of different treatment equipment, improving the timing accuracy of the joint consumption path construction process.

[0116] In this embodiment, step two specifically involves:

[0117] The operating units of each treatment equipment in the treatment equipment operating unit set are sorted by time according to the operating time index, and the operating parameters in the corresponding equipment segment are extracted according to the equipment identifier to generate a time-series operating unit sequence.

[0118] Extract the changes in desulfurization tower slurry circulation volume, denitrification ammonia injection flow rate, dust collector differential pressure, and induced draft fan load power within adjacent operating time windows in the time-series operating unit sequence;

[0119] Based on the direction of change of the corresponding operating parameters of each treatment equipment operating unit, the synchronous change relationship between different treatment equipment operating units is matched to generate equipment association node pairs;

[0120] The synchronization ratio of the direction of change of operating parameters of each associated node within a continuous operating time window is statistically analyzed.

[0121] Devices whose synchronization ratio reaches the set association ratio are identified as joint consumption associated nodes;

[0122] Establish node connection relationships based on the equipment identifiers and runtime indexes corresponding to each interconnected node, and construct an equipment interconnection graph.

[0123] According to the node connection direction in the equipment co-consumption association diagram, the path of each co-consumption association node is traversed, and the equipment identifier, running time index, flue gas flow, sulfur dioxide concentration, nitrogen oxide concentration, power consumption and reagent consumption corresponding to each co-consumption association node are recorded to generate a co-consumption path sequence.

[0124] In the specific implementation process, a 1-minute operating time window is used. The operating units of the treatment equipment in the treatment equipment operating unit set are arranged in ascending order according to their operating time index. Based on the equipment identifier, the corresponding operating parameters of the desulfurization equipment section, denitrification equipment section, dust removal equipment section, and induced draft fan section are extracted to form a time-series operating unit sequence. The change in the slurry circulation volume of the desulfurization tower between adjacent operating time windows is obtained by subtracting the slurry circulation volume of the previous time window from the current time window's slurry circulation volume. The changes in the ammonia injection flow rate for denitrification, the differential pressure of the dust collector, and the load power of the induced draft fan are calculated in the same way. When the corresponding operating parameter shows an upward trend in two consecutive operating time windows, the direction of change is determined to be the same; when it shows a downward trend in two consecutive operating time windows, the direction of change is determined to be opposite. The direction of change in operating parameters between different treatment equipment operating units is synchronized window by window. When the number of synchronizations of the same-direction or opposite-direction changes reaches 8 times in 10 consecutive operating time windows, the synchronization ratio of the direction of change in operating parameters is calculated to be 0.8. Equipment-related node pairs with a synchronization ratio of 0.8 are identified as joint consumption-related nodes. This method can extract the long-term stable linkage of equipment operation relationships from discrete operating data, reducing the interference of occasional fluctuations on the joint consumption relationship identification results.

[0125] In constructing the equipment co-consumption correlation graph, the equipment identifiers corresponding to the co-consumption correlation nodes are used as graph nodes, and the running time index is used as the node connection order. Node connection edges are established based on the order of running time. The connection direction from the desulfurization equipment section to the induced draft fan section is defined as the forward co-consumption path, and the connection direction from the denitrification equipment section to the dust removal equipment section is defined as the collaborative co-consumption path. During path traversal, a single path traversal length of 5 consecutive nodes is used. The flue gas flow rate, sulfur dioxide concentration, nitrogen oxide concentration, power consumption, and reagent consumption corresponding to each co-consumption correlation node are recorded synchronously and formed into a co-consumption path sequence according to the node connection order. By recording the operating parameters in the co-consumption path in a unified order, the temporal correspondence accuracy between different equipment sections during path cost evolution analysis can be improved.

[0126] In this embodiment, the path embedding module performs path state mapping on the joint-consumption path sequence to generate initial path features, specifically:

[0127] According to the node connection order in the joint consumption path sequence, the equipment identifier, running time index, flue gas flow, sulfur dioxide concentration, nitrogen oxide concentration, power consumption and reagent consumption corresponding to each joint consumption associated node are read sequentially to generate a path status data group.

[0128] Based on the device identifier, the different device segments in the path status data group are numbered and mapped to generate a device segment number sequence;

[0129] Based on the running time index, the running order of each node in the path status data group is mapped to a time position sequence to generate a time position sequence.

[0130] The flue gas flow rate, sulfur dioxide concentration, and nitrogen oxide concentration constitute the flue gas condition operation parameter group, and the power consumption and reagent consumption constitute the cost condition operation parameter group.

[0131] The operating parameters in the flue gas state operating parameter group are subjected to maximum and minimum normalization to generate the flue gas state vector; the operating parameters in the cost state operating parameter group are subjected to maximum and minimum normalization to generate the cost state vector.

[0132] The equipment segment number sequence, time location sequence, flue gas state vector, and cost state vector are concatenated according to the node connection order to generate a path state matrix.

[0133] The path state matrix is ​​vector-mapped through the fully connected mapping layer in the path embedding module to generate initial path features;

[0134] In the specific implementation process, each associated node of the joint consumption path is read sequentially according to the node connection direction in the joint consumption path sequence. The equipment identifier, running time index, flue gas flow rate, sulfur dioxide concentration, nitrogen oxide concentration, power consumption, and reagent consumption are written into the path status data group in chronological order. The running time index is recorded in a continuously increasing manner, and the time interval between adjacent nodes is set to 1 minute. Different equipment segments are distinguished by a fixed character encoding method, which can maintain a stable correspondence between equipment nodes in the joint consumption path in the path status matrix and improve the consistency of node identification during the path evolution process.

[0135] In the process of generating the flue gas state vector and cost state vector, the flue gas flow rate, sulfur dioxide concentration, and nitrogen oxide concentration in the flue gas state operating parameter group are subjected to maximum and minimum normalization, respectively, and the normalized parameter values ​​are restricted to between 0 and 1. The power consumption and reagent consumption in the cost state operating parameter group are subjected to maximum and minimum normalization in the same way. After the normalization process is completed, the equipment segment number sequence, time position sequence, flue gas state vector, and cost state vector are concatenated according to the node connection order to form a path state matrix. The input dimension of the fully connected mapping layer in the path embedding module is set to 7, and the output dimension is set to 64, generating 64-dimensional initial path features through matrix mapping. By uniformly mapping different operating parameters, the numerical fluctuation differences between operating parameters of different dimensions can be reduced, and the stability of path state representation in the steady-state evolution coding process can be improved.

[0136] In this embodiment, the initial path features are continuously time-period evolution encoded by the steady-state evolution coding module to generate steady-state evolution features, specifically:

[0137] The initial path features are divided into multiple consecutive time period feature segments according to the running time index, and arranged according to the node connection order in the joint consumption path sequence to generate a consecutive time period feature sequence;

[0138] The continuous time-period feature sequence is input into the one-dimensional temporal convolutional layer in the steady-state evolution coding module, and the local evolution features between adjacent time periods are extracted along the running time index direction to generate a local evolution feature sequence.

[0139] The local evolution feature sequence is input into the depth-separable convolutional layer in the steady-state evolution coding module. The channel features corresponding to each device segment are independently convolved, and the convolutional features corresponding to different device segments are combined and mapped across device segments by point-by-point convolution to generate the device segment evolution feature sequence.

[0140] The equipment segment evolution feature sequence is input into the layer normalization layer in the steady-state evolution coding module. The equipment segment evolution features under the same running time index are processed by layer normalization to generate a normalized evolution feature sequence.

[0141] The normalized evolution feature sequence is input into the feedforward mapping layer in the steady-state evolution coding module. The normalized evolution features under each running time index are nonlinearly mapped by the GELU activation function to generate candidate evolution feature sequences.

[0142] The candidate evolution feature sequence and the local evolution feature sequence are added together by residual addition to generate the fused evolution feature sequence; the fused evolution features corresponding to adjacent running time indices in the fused evolution feature sequence are differentially analyzed in one dimension to generate the feature difference sequence.

[0143] Based on the positive and negative directions of the difference values ​​of each dimension in the feature difference sequence, feature dimensions that maintain the same direction of change under continuous running time index are extracted to generate a steady-state dimension label sequence.

[0144] Based on the steady-state dimension labeling sequence, the corresponding fusion evolution feature dimensions are retained from the fusion evolution feature sequence, and the fusion evolution feature dimensions that are not labeled as steady-state dimensions are set to zero to generate a steady-state preservation feature sequence.

[0145] The steady-state evolution features are generated by linearly mapping the steady-state preservation feature sequence through a linear output layer.

[0146] In the specific implementation process, the initial path features are divided into a continuous time period feature segment according to eight consecutive runtime indices. The sliding stride between adjacent continuous time period feature segments is set to four runtime indices, and a continuous time period feature sequence is formed according to the node connection order in the continuous consumption path sequence. The kernel length of the one-dimensional temporal convolutional layer in the steady-state evolution coding module is set to 3, the convolution stride is set to 1, and the output channel dimension is set to 64, which is used to extract the local evolution relationship between adjacent runtime windows. The depthwise separable convolutional layer has a depthwise convolutional kernel length of 5 and a pointwise convolution output channel dimension of 128. By performing cross-device segment combination mapping on the convolutional features corresponding to TXQ, CCQ, DSQ, and YFQ, the continuous consumption propagation relationship between different device segments is kept continuously associated in the same convolutional space. The normalization dimension in the layer normalization layer is set to 128. The evolution features of device segments under the same runtime index are uniformly normalized, which can reduce the numerical fluctuation difference between convolutional features of different device segments and improve the feature stability in the continuous time period evolution process.

[0147] In the feedforward mapping layer, the input dimension is set to 128, the hidden layer dimension is set to 256, and the output dimension is restored to 128. The GELU activation function is used to complete the nonlinear mapping. After adding the residuals, the fused evolution features corresponding to adjacent runtime indices in the fused evolution feature sequence are differentially analyzed dimension by dimension. When the difference values ​​in three consecutive runtime indices maintain the same positive or negative direction, the corresponding feature dimension is marked as the steady-state dimension. Based on the steady-state dimension marking sequence, the fused evolution feature dimensions in the fused evolution feature sequence that are not marked as steady-state dimensions are set to zero, retaining only the fused evolution features corresponding to the steady-state dimensions. The feature dimension is then compressed to 64 dimensions through a linear output layer to generate the steady-state evolution features. This processing method reduces the interference of short-term fluctuating operating parameters on the path evolution process and improves the ability to identify stable states during path evolution sequence rearrangement.

[0148] In this embodiment, the disturbance track rearrangement module rearranges the path evolution order of the steady-state evolution characteristics through the smoke load tidal migration mechanism to generate rearranged path characteristics, specifically:

[0149] The steady-state evolution features are input into the disturbance track rearrangement module. The disturbance track rearrangement module is equipped with a smoke load tidal migration mechanism. Based on the running time index and equipment segment number corresponding to each feature dimension in the steady-state evolution features, the feature dimensions belonging to the same equipment segment under the same running time window are vectorized according to the feature dimension order to restore the evolution features of the corresponding equipment segment.

[0150] The device segment evolution characteristics in different runtime windows are continuously arranged according to the order of runtime index to generate a time period arrangement sequence;

[0151] Extract the flue gas flow rate, power consumption, and reagent consumption corresponding to the evolution characteristics of each equipment segment in the time period sequence, and calculate the direction of flue gas flow rate change, power consumption change, and reagent consumption change between adjacent operating time windows respectively.

[0152] The evolution characteristics of adjacent equipment sections whose flue gas flow rate changes in the same direction as the power consumption change are identified as the first tidal migration segment.

[0153] The evolution characteristics of adjacent equipment sections whose flue gas flow rate changes in the same direction as the reagent consumption change are identified as the second tidal migration segment.

[0154] The evolution characteristics of adjacent equipment sections whose flue gas flow rate changes are inconsistent with the direction of power consumption changes or the direction of reagent consumption changes are identified as the third tidal migration segment.

[0155] The magnitude of the change in power consumption in each first tidal migration segment is statistically analyzed, and the segments are arranged in ascending order of the magnitude of the change in power consumption.

[0156] The variation amplitude of drug consumption in each second tidal migration segment was statistically analyzed, and the segments were arranged in ascending order of the variation amplitude of drug consumption.

[0157] The amplitude of flue gas flow rate change corresponding to each third tidal migration segment is statistically analyzed, and the segments are arranged in descending order of flue gas flow rate change amplitude.

[0158] The results of the first, middle and last sections are sequentially concatenated, and the running time index corresponding to the evolution characteristics of each device section is retained to generate rearranged path features.

[0159] In the specific implementation process, after inputting the steady-state evolution features into the disturbance track rearrangement module, based on the running time index and equipment segment number corresponding to each feature dimension, the feature dimensions corresponding to TXQ, CCQ, DSQ, and YFQ in the same running time window are vectorized according to the original feature dimension order. The feature vector corresponding to each equipment segment constitutes the evolution feature of the corresponding equipment segment. Subsequently, the evolution features of equipment segments in different running time windows are continuously arranged according to the order of the running time index to form a time period arrangement sequence. In the flue gas tidal migration mechanism, by dividing the synchronous relationship between the flue gas flow change direction and the power consumption change direction and the reagent consumption change direction, the joint consumption evolution state in different equipment segments can be divided into different tidal migration segments.

[0160] In the initial arrangement, the energy consumption variation amplitudes of the first tidal migration segment are sorted in ascending order, prioritizing the first tidal migration segment with smaller energy consumption variation amplitudes to the initial region, thereby increasing the priority of low energy consumption and combined consumption paths during path evolution. In the middle arrangement, the reagent consumption variation amplitudes of the second tidal migration segment are sorted in ascending order, prioritizing the second tidal migration segment with smaller reagent consumption variation amplitudes to the middle region, thereby maintaining the continuous and stable state of low reagent consumption and combined consumption paths during path evolution. In the final arrangement, the flue gas flow variation amplitudes of the third tidal migration segment are sorted in descending order, prioritizing the third tidal migration segment with larger flue gas flow variation amplitudes to the final region. Since the third tidal migration segment corresponds to a joint consumption path where the direction of flue gas flow change is inconsistent with the direction of power consumption change or the direction of reagent consumption change, such joint consumption paths may experience increased smoke load disturbance and unstable cost coordination relationship. Therefore, prioritizing the shift of the third tidal migration segment with a large flue gas flow change amplitude can reduce the proportion of high smoke load disturbance paths in the path cost evolution analysis and improve the continuity of stable joint consumption paths in the rearranged path characteristics.

[0161] In this embodiment, the cost trend output module performs time-period trend analysis on the rearranged path features to generate a path cost evolution sequence, specifically:

[0162] Input the rearranged path features into the cost trend output module, and extract the equipment segment evolution features, running time index, and corresponding arrangement position marker from the rearranged path features;

[0163] Based on the arrangement position markers, corresponding path priority features are generated, where the first arrangement result corresponds to the first path priority feature, the middle arrangement result corresponds to the second path priority feature, and the last arrangement result corresponds to the third path priority feature.

[0164] Extract the power consumption, reagent consumption and flue gas flow corresponding to the evolution characteristics of each equipment segment, calculate the difference in power consumption between adjacent operating time windows to obtain the direction of power consumption change, and calculate the difference in reagent consumption between adjacent operating time windows to obtain the direction of reagent consumption change.

[0165] When both the direction of change in power consumption and the direction of change in reagent consumption are increasing, the direction of change in path cost between the corresponding path operating states is determined to be upward.

[0166] When both the direction of change in power consumption and the direction of change in reagent consumption are decreasing, the direction of change in path cost between the corresponding path operating states is determined to be decreasing.

[0167] When the direction of change in power consumption is inconsistent with the direction of change in reagent consumption, the direction of change in path cost between the corresponding path operation states is determined as the differentiation direction.

[0168] The continuous path operation status that maintains a consistent direction of path cost change is defined as the same trend evolution segment;

[0169] Record the direction of change in power consumption, the direction of change in reagent consumption, the direction of change in flue gas flow, and the corresponding equipment segment number in each trend evolution segment to generate segment cost characteristics;

[0170] The segment cost features and the corresponding path priority features are concatenated according to the feature dimension order to generate spatiotemporal correlation features;

[0171] The spatiotemporal correlation features are arranged continuously according to the running time index to generate a path cost evolution sequence;

[0172] In the specific implementation process, after inputting the rearranged path features into the cost trend output module, the equipment segment evolution features, running time index, and arrangement position markers are first extracted, and corresponding path priority features are generated based on the arrangement position markers. Specifically, the first path priority feature corresponding to the first segment arrangement result is set to [1,0,0], the second path priority feature corresponding to the middle segment arrangement result is set to [0,1,0], and the third path priority feature corresponding to the last segment arrangement result is set to [0,0,1]. Different path priority features correspond to low power consumption priority paths, medium chemical consumption stable paths, and high smoke load disturbance paths, respectively. Subsequently, the power consumption, chemical consumption, and flue gas flow rate corresponding to each equipment segment evolution feature are extracted, and the difference is calculated according to adjacent running time windows. When the power consumption in the current running time window is greater than the power consumption in the previous running time window, by continuously tracking the direction of change in different running time windows, the cost change relationship between the corresponding path running states can be formed.

[0173] In the process of generating segment cost features, the continuous path operation status with consistent path cost change direction is divided into segments, and the change direction of power consumption, reagent consumption, flue gas flow, and equipment segment number in the corresponding segments are recorded. The change directions of power consumption, reagent consumption, and flue gas flow are all represented using directional encoding. When the value of the corresponding operating parameter in the current operating time window is greater than that in the previous operating time window, the change direction is encoded as 1; when the value of the corresponding operating parameter in the current operating time window is less than that in the previous operating time window, the change direction is encoded as -1; when the value of the corresponding operating parameter in the current operating time window is the same as that in the previous operating time window, the change direction is encoded as 0. After extracting the segment cost features, the segment cost features and the corresponding path priority features are vector-concatenated according to the feature dimension order to form spatiotemporal correlation features. The segment cost features are used to characterize the cost change status in different operating time windows, and the path priority features are used to characterize the priority position of different paths in the smoke load tidal migration mechanism. By combining the cost change status with the path priority position, the path cost evolution sequence can simultaneously possess temporal continuity and path priority differences, thereby improving the ability to identify path status during the governance path adjustment process.

[0174] In this invention, the improved Koopa model is a network for analyzing the joint consumption path evolution, formed by structurally extending the existing Koopa time series prediction model. It retains the basic ability of the Koopa model to decompose and model the dynamic evolution process of time series. In order to address the problems of multi-device joint consumption propagation, operational disturbance coupling, and dynamic changes in path costs in flue gas treatment systems, the time series evolution structure in the original Koopa model is reconstructed in a path-based manner. The modules of the improved Koopa model are connected sequentially according to the data flow direction of the joint consumption path sequence. The 64-dimensional initial path features output by the path embedding module are used as the input of the steady-state evolution coding module. The 64-dimensional steady-state evolution features output by the steady-state evolution coding module are passed to the disturbance track rearrangement module. The rearranged path features output by the disturbance track rearrangement module are further input into the cost trend output module to generate the path cost evolution sequence.

[0175] The model training data comes from the historical operation database of the flue gas treatment system. The training samples use the sequence of joint consumption paths within a continuous operating time window as input, and the changes in power consumption, reagent consumption, and equipment load within the corresponding operating time window as supervision labels. Each label is annotated using the difference between operating time windows. During model training, the mean squared error loss function is used to calculate the error between the predicted results and the true labels, and the Adam optimizer is used to update the parameters. The initial learning rate is set to 0.001, the batch size is set to 32, and the number of training epochs is set to 200. Compared to the existing Koopa model, this invention introduces a smoke load tidal migration mechanism to rearrange the propagation order of joint consumption disturbances between different equipment sections. It also utilizes path priority features to achieve differentiated evolutionary expressions between low power consumption paths, low chemical consumption paths, and high smoke load disturbance paths, thereby enhancing the model's ability to dynamically identify cost propagation relationships in complex joint consumption paths. Simultaneously, through path affiliation weighted splitting and MCTS tree search process, it achieves dynamic allocation and search optimization of cost contribution relationships under different treatment paths, improving the cost optimization accuracy and multi-equipment collaborative adjustment capability of the flue gas treatment system.

[0176] In this embodiment, step six specifically includes:

[0177] Extract the spatiotemporal correlation features, running time index and corresponding equipment segment number from the path cost evolution sequence. Divide the spatiotemporal correlation features into continuous path evolution time periods based on the running time index, and determine the spatiotemporal correlation features in the same path evolution time period as the same path time period feature group.

[0178] Input the time period feature groups of each path into the fully connected mapping layer, perform feature dimension mapping on the spatiotemporal correlation features in each time period feature group of the path, and generate the corresponding path attribution value;

[0179] Softmax normalization is applied to the path attribution values ​​of each path during the same path evolution period to generate corresponding path attribution weights;

[0180] Based on the running time index within the same path evolution period, calculate the sequence of changes in power consumption, reagent consumption, and equipment load between adjacent running time indices within the path evolution period;

[0181] The path energy consumption characteristics are generated by multiplying the corresponding path attribution weight with the sequence of changes in energy consumption; the path reagent consumption characteristics are generated by multiplying the corresponding path attribution weight with the sequence of changes in reagent consumption; and the path load change characteristics are generated by multiplying the corresponding path attribution weight with the sequence of changes in equipment load.

[0182] By combining the path energy consumption characteristics, path reagent consumption characteristics, and path load change characteristics, a single path governance cost unit is generated.

[0183] In the specific implementation process, the spatiotemporal correlation features, runtime indices, and equipment segment numbers are first extracted from the path cost evolution sequence. Continuous path evolution periods are then divided based on runtime indices, with each five consecutive runtime indices constituting a path evolution period. For spatiotemporal correlation features within the same path evolution period, they are grouped according to equipment segment numbers to form corresponding path period feature groups. Subsequently, each path period feature group is input into a fully connected mapping layer. The input feature dimension of the fully connected mapping layer is set to 96 dimensions, and the output dimension is set to 1 dimension. Through matrix mapping, the spatiotemporal correlation features corresponding to each path period feature group are mapped to individual path attribution values. After generating the path attribution values, all path attribution values ​​within the same path evolution period are subjected to Softmax normalization to generate corresponding path attribution weights. This ensures that the sum of all path attribution weights within the same path evolution period remains 1, thereby representing the contribution ratio of different paths to the overall governance cost.

[0184] Next, the path energy consumption characteristics are generated by multiplying the corresponding path attribution weights with the energy consumption change sequence; the path reagent consumption characteristics are generated by multiplying the corresponding path attribution weights with the reagent consumption change sequence; and the path load change characteristics are generated by multiplying the corresponding path attribution weights with the equipment load change sequence. Finally, the path energy consumption characteristics, path reagent consumption characteristics, and path load change characteristics are combined to generate a single-path governance cost unit, thereby realizing the independent decomposition and dynamic attribution analysis of the governance cost change process under different joint consumption paths.

[0185] In this embodiment, step seven specifically includes:

[0186] The path energy consumption characteristics, path reagent consumption characteristics, and path load change characteristics in the single path governance cost unit are respectively subjected to maximum and minimum normalization processing to generate normalized energy consumption characteristics, normalized reagent consumption characteristics, and normalized load change characteristics.

[0187] The path contribution characteristics are generated by combining the normalized energy consumption characteristics, normalized reagent consumption characteristics, and normalized load change characteristics.

[0188] The contribution characteristics of each path are arranged consecutively according to the running time index to generate a path contribution sequence;

[0189] The path cost evolution sequence and path contribution sequence are input into the MCTS algorithm for tree search. The path contribution feature corresponding to the current running time index is used as the root node, and the path contribution features corresponding to multiple subsequent running time indices are used as candidate child nodes to generate a path search tree.

[0190] The normalized energy consumption feature, normalized reagent consumption feature, and normalized load change feature in the path contribution features of each candidate sub-node are squared and then summed. The summation result is then squared to generate the corresponding path contribution value.

[0191] When the contribution value of the path corresponding to the candidate child node is less than the contribution value of the path corresponding to the current root node, and the change in sulfur dioxide concentration and the change in nitrogen oxide concentration of the corresponding path are both less than the change in sulfur dioxide concentration and the change in nitrogen oxide concentration of the current root node, the corresponding candidate child node is determined as a low-cost candidate node.

[0192] The candidate child node with the smallest path contribution value is selected from the low-cost candidate nodes as the expansion node, and the path contribution feature corresponding to the subsequent running time index of the expansion node is extracted as the next layer candidate child node until the path search tree reaches the set search layer.

[0193] The sum of the path contribution values ​​in each path search branch is calculated, and the path search branch with the smallest sum of path contribution values ​​is taken as the governance path adjustment set;

[0194] In the specific implementation process, the path energy consumption characteristics, path reagent consumption characteristics, and path load change characteristics in the single-path governance cost unit are first subjected to maximum and minimum normalization processing, where the normalized feature values ​​are all limited to between 0 and 1, thereby eliminating the dimensional differences between different features. Then, the normalized energy consumption characteristics, normalized reagent consumption characteristics, and normalized load change characteristics are combined according to the feature dimension order to generate path contribution characteristics, and the path contribution characteristics are continuously arranged according to the running time index to form a path contribution sequence. In the MCTS tree search process, the path contribution characteristic corresponding to the current running time index is used as the root node, and the path contribution characteristics corresponding to the next three running time indices are used as candidate child nodes to generate a path search tree. For each candidate child node, the corresponding path contribution value is calculated, thereby using a unified numerical space to represent the comprehensive governance cost status corresponding to different paths.

[0195] In the process of screening low-cost candidate nodes, when the path contribution value of a candidate child node is less than that of the current root node, and the changes in sulfur dioxide concentration and nitrogen oxide concentration in the corresponding path are both less than those of the current root node, the corresponding candidate child node is identified as a low-cost candidate node. Subsequently, the candidate child node with the smallest path contribution value is selected from the low-cost candidate nodes as an extended node, and the path contribution features corresponding to the subsequent running time index of the extended node are extracted as the next layer of candidate child nodes, until the path search tree reaches a search depth of 5 layers. After completing the tree search, the path contribution values ​​in each path search branch are accumulated and calculated, and the path search branch with the smallest sum of path contribution values ​​is used as the treatment path adjustment set, thereby realizing the joint screening of low-emission, low-energy-consumption, and low-load-fluctuation paths, improving the cost optimization capability and operational stability in the flue gas treatment process.

[0196] In this embodiment, step eight specifically includes:

[0197] Extract the governance device paths, corresponding runtime indices, and corresponding path contribution values ​​from the governance path adjustment set, and sort the different governance device paths by time according to the runtime index to generate the governance adjustment order for the corresponding time period.

[0198] When the path contribution value of the corresponding treatment equipment path is less than the path contribution value of the previous running time index, reduce the desulfurization tower slurry circulation volume and denitrification ammonia injection flow rate in the corresponding time period, and simultaneously reduce the induced draft fan load power.

[0199] When the path contribution value of the corresponding treatment equipment path is greater than the path contribution value of the previous running time index, the circulation volume of desulfurization tower slurry and the flow rate of denitrification ammonia injection in the corresponding time period are increased, and the load power of the induced draft fan is increased simultaneously.

[0200] When the path contribution value of the corresponding treatment equipment path is the same as the path contribution value of the previous running time index, the desulfurization tower slurry circulation volume, denitrification ammonia injection flow rate and induced draft fan load power in the corresponding time period remain unchanged.

[0201] A time-sharing management and scheduling table is generated based on the desulfurization tower slurry circulation volume, denitrification ammonia injection flow rate, and induced draft fan load power for each time period.

[0202] Based on the time-sharing management and scheduling table, the changes in electricity consumption, reagent consumption, and equipment load during the corresponding time periods are statistically analyzed to generate corresponding cost statistics.

[0203] The cost statistics for each time period are summed up to output the cost optimization results;

[0204] In the specific implementation process, firstly, the treatment equipment paths, corresponding operating time indices, and corresponding path contribution values ​​of the treatment path adjustment set are extracted. Then, different treatment equipment paths are sorted by time according to the operating time index to generate the treatment adjustment order for the corresponding time period. For the change in path contribution values ​​between adjacent operating time indices, the difference between the path contribution value corresponding to the current operating time index and the path contribution value corresponding to the previous operating time index is calculated. This difference is then divided by the path contribution value corresponding to the previous operating time index to generate the path contribution change ratio. When the path contribution value corresponding to the current operating time index is less than that corresponding to the previous operating time index, the desulfurization tower slurry circulation volume is reduced proportionally, the denitrification ammonia injection flow rate is reduced proportionally, and the induced draft fan load power is reduced proportionally, according to the path contribution change ratio. When the path contribution value corresponding to the current operating time index is greater than that corresponding to the previous operating time index, the path contribution change ratio is adjusted accordingly. Based on the path contribution change ratio, the desulfurization tower slurry circulation volume, the denitrification ammonia injection flow rate, and the induced draft fan load power are increased proportionally. When the path contribution value corresponding to the current operating time index is the same as that corresponding to the previous operating time index, the desulfurization tower slurry circulation volume, denitrification ammonia injection flow rate, and induced draft fan load power remain unchanged. After completing the corresponding parameter adjustment, the desulfurization tower slurry circulation volume, denitrification ammonia injection flow rate, and induced draft fan load power corresponding to each operating time index are written into the corresponding time period to generate a time-sharing treatment scheduling table. Subsequently, based on the time-sharing treatment scheduling table, the changes in power consumption, reagent consumption, and equipment load between each operating time index are statistically analyzed, and corresponding cost statistics are generated. Finally, the cost statistics results corresponding to each operating time index are accumulated to output the cost optimization results, thereby realizing dynamic coordinated adjustment and optimized control of operating costs among different treatment equipment in the flue gas treatment process.

[0205] Example 1: To verify the feasibility of this invention in practice, it was applied to a flue gas pollution control system at a waste-to-energy incineration plant. This system processes approximately 1200 tons of waste per day, and the flue gas treatment process includes a denitrification section, a dust removal section, a desulfurization section, and an induced draft fan section. The problem encountered on-site was that the flue gas load fluctuated significantly during the morning and evening peak feeding periods. The ammonia injection flow rate in the denitrification system, the slurry circulation volume in the desulfurization tower, and the load power of the induced draft fan were often adjusted manually based on experience. This easily led to problems such as excessive ammonia injection, prolonged high-load operation of the slurry circulation pump, and frequent load adjustments of the induced draft fan, resulting in emissions meeting standards but with high treatment costs.

[0206] In this embodiment, the system uses a 1-minute time window to collect data on the desulfurization tower slurry circulation rate, denitrification ammonia injection flow rate, dust collector pressure differential, induced draft fan load power, flue gas flow rate, sulfur dioxide concentration, nitrogen oxide concentration, power consumption, and reagent consumption. The collected data is mapped to equipment segments to form a set of treatment equipment operating units. Then, a time-series operational correlation analysis is used to construct an equipment co-consumption correlation graph, generating a co-consumption path sequence. This co-consumption path sequence is then input into an improved Koopa model, where an initial path feature is generated through a path embedding module, and then encoded using steady-state evolution. The module identifies stable power consumption states, and then the smoke load tidal migration mechanism in the disturbance track rearrangement module distinguishes low power consumption paths, low chemical consumption paths, and high smoke load disturbance paths, generating rearranged path features. After the cost trend output module outputs the path cost evolution sequence, the system performs path-attribution weighted decomposition on the changes in power consumption, chemical consumption, and equipment load to obtain single-path governance cost units. Then, the MCTS algorithm searches for the path search branch with the minimum sum of path contribution values, forming a governance path adjustment set, generating a time-sharing governance scheduling table, and outputting cost optimization results.

[0207] To verify the actual effectiveness of the present invention, three comparative schemes were set up. Comparative scheme one is the manual experience adjustment scheme, in which the operators manually adjust the ammonia injection rate, slurry circulation rate, and induced draft fan load according to the emission curve; comparative scheme two is the single-equipment energy-saving control scheme, in which the desulfurization, denitrification, and induced draft equipment are independently adjusted for energy saving; comparative scheme three is the conventional time-series predictive scheduling scheme, in which the power consumption and chemical consumption are predicted using an LSTM model and then scheduled according to rules. The comparison results are shown in Table 1.

[0208] Table 1. Comparison of the operational effects of different flue gas treatment cost optimization schemes

[0209]

[0210] As shown in Table 1 above, there are significant differences in energy consumption control, reagent utilization rate, and emission stability among different flue gas treatment schemes. Scheme 1, which uses manual experience-based adjustment, suffers from frequent fluctuations in the induced draft fan due to the lack of continuous analysis of the interconnected energy consumption relationships between the desulfurization, denitrification, and induced draft fan sections. This results in an energy consumption of 1648 kWh / h and a reagent consumption of 91.6 kg / h, while the average NOx emission concentration reaches 46.5 mg / m³. This indicates that traditional manual adjustment methods are prone to problems such as delayed ammonia injection and insufficient equipment coordination. Scheme 2, using single-equipment energy-saving control, reduces energy consumption to 1512 kWh / h and the induced draft fan load fluctuation rate to 15.2%. However, due to the lack of dynamic optimization of interconnected energy consumption paths between different treatment devices, the average SO2 emission concentration actually increases to 22.4 mg / m³, indicating that localized energy-saving control can easily affect overall emission stability. Scheme 3, using traditional LSTM... After the time-series prediction scheme, the consumption of reagents and the load fluctuation rate were further reduced, indicating that time-series prediction can improve the rationality of scheduling to some extent. However, due to the lack of tree-structured search optimization of path contribution relationships, the average NOx emission concentration still reached 44.3 mg / m³. In contrast, the method of this invention uses an improved Koopa model to perform dynamic evolution analysis of the combined consumption path and combines the MCTS algorithm to continuously search for low-cost paths, reducing the power consumption to 1284 kWh / h, the reagent consumption to 72.5 kg / h, and the induced draft fan load fluctuation rate to 8.6%. At the same time, the average SO2 emission concentration and the average NOx emission concentration were reduced to 18.7 mg / m³ and 39.6 mg / m³, respectively. This shows that the present invention can achieve synergistic optimization of treatment costs and equipment operation stability while maintaining stable compliance with pollutant standards.

[0211] This embodiment applies the present invention to the flue gas treatment system of a waste incineration power plant. By dynamically analyzing the interconnected consumption relationships between the desulfurization equipment section, the denitrification equipment section, the dust removal equipment section, and the induced draft equipment section, and combining the improved Koopa model to complete the path evolution modeling, the MCTS algorithm is used to optimize the treatment path through tree search. This achieves coordinated adjustment and time-sharing treatment control among different equipment in the flue gas treatment process, effectively reducing energy and chemical consumption in the flue gas treatment process, improving equipment operation stability, and ensuring that pollutant emissions continuously and stably meet standards.

[0212] The above description is only a preferred embodiment of the present invention, but the scope of protection of the present invention is not limited thereto. Any equivalent substitutions or modifications made by those skilled in the art within the scope of the technology disclosed in the present invention, based on the technical solution and inventive concept of the present invention, should be covered within the scope of protection of the present invention.

Claims

1. A method for optimizing the cost management of flue gas pollution control, characterized in that, Includes the following steps: Step 1: Collect flue gas treatment operation data of the target flue gas treatment system during its operating cycle, and perform equipment segment mapping processing to generate a set of treatment equipment operation units; Step 2: Perform time-series operational correlation analysis on the set of treatment equipment operating units, construct an equipment joint consumption correlation diagram, and generate a joint consumption path sequence; Step 3: Input the joint cost path sequence into the improved Koopa model, which includes a path embedding module, a steady-state evolution coding module, a disturbance track rearrangement module, and a cost trend output module; Step 4: Map the path state of the joint-consumption path sequence through the path embedding module to generate initial path features, and perform continuous time-period evolution coding on the initial path features through the steady-state evolution coding module to generate steady-state evolution features; Step 5: The disturbance track rearrangement module rearranges the path evolution order of the steady-state evolution characteristics through the smoke load tidal migration mechanism, generates rearranged path characteristics, and performs time-period trend analysis on the rearranged path characteristics through the cost trend output module to generate a path cost evolution sequence. Step 6: Based on the path cost evolution sequence, perform path-attribution weighted decomposition of reagent consumption, power consumption, and equipment load changes to generate single-path governance cost units; Step 7: Perform emission contribution mapping on the single-path governance cost unit to generate a path contribution sequence, and use the MCTS algorithm to perform tree search based on the path cost evolution sequence and the path contribution sequence to generate a governance path adjustment set; Step 8: Generate a time-sharing governance schedule table based on the governance path adjustment set, and output the cost optimization results based on the time-sharing governance schedule table.

2. The method for optimizing the cost management of flue gas pollution control according to claim 1, characterized in that, Step one specifically involves: Collect flue gas treatment operation data of the target flue gas treatment system during the operating cycle. The flue gas treatment operation data includes desulfurization tower slurry circulation volume, denitrification ammonia injection flow rate, dust collector pressure difference, induced draft fan load power, flue gas flow rate, sulfur dioxide concentration, nitrogen oxide concentration, power consumption, and reagent consumption. The flue gas treatment operation data is time-synchronized according to a uniform time interval to generate a time-series operation data group. Based on the process connection sequence of the desulfurization unit, denitrification unit, dust removal unit, and induced draft equipment in the flue gas treatment system, the time-series operation data group is divided into equipment segments to generate equipment segment data groups. Extract the device identifier, runtime index, and corresponding runtime parameters from each device segment data group, and establish the device segment correspondence; Based on the corresponding relationship between the equipment segments, the data groups of each equipment segment are mapped and associated to generate a set of governance equipment operation units.

3. The method for optimizing the cost management of flue gas pollution control according to claim 1, characterized in that, Step two specifically involves: The operating units of each treatment equipment in the treatment equipment operating unit set are sorted by time according to the operating time index, and the operating parameters in the corresponding equipment segment are extracted according to the equipment identifier to generate a time-series operating unit sequence. Extract the changes in desulfurization tower slurry circulation volume, denitrification ammonia injection flow rate, dust collector differential pressure, and induced draft fan load power within adjacent operating time windows in the time-series operating unit sequence; Based on the direction of change of the corresponding operating parameters of each treatment equipment operating unit, the synchronous change relationship between different treatment equipment operating units is matched to generate equipment association node pairs; The synchronization ratio of the direction of change of operating parameters of each associated node within a continuous operating time window is statistically analyzed. The device-associated node pairs whose synchronization ratio reaches the set association ratio are identified as joint consumption associated nodes; Establish node connection relationships based on the equipment identifiers and runtime indexes corresponding to each interconnected node, and construct an equipment interconnection graph. According to the node connection direction in the equipment co-consumption association diagram, the path of each co-consumption association node is traversed, and the equipment identifier, running time index, flue gas flow rate, sulfur dioxide concentration, nitrogen oxide concentration, power consumption and reagent consumption corresponding to each co-consumption association node are recorded to generate a co-consumption path sequence.

4. The method for optimizing the cost management of flue gas pollution control according to claim 1, characterized in that, The step of mapping the path state of the joint-consumption path sequence through the path embedding module to generate initial path features is as follows: According to the node connection order in the joint consumption path sequence, the equipment identifier, running time index, flue gas flow, sulfur dioxide concentration, nitrogen oxide concentration, power consumption and reagent consumption corresponding to each joint consumption associated node are read sequentially to generate a path status data group. Based on the device identifier, the different device segments in the path status data group are numbered and mapped to generate a device segment number sequence; Based on the running time index, the running order of each node in the path status data group is mapped to a time position sequence to generate a time position sequence. The flue gas flow rate, sulfur dioxide concentration, and nitrogen oxide concentration constitute the flue gas condition operation parameter group, and the power consumption and reagent consumption constitute the cost condition operation parameter group. The operating parameters in the flue gas state operating parameter group are subjected to maximum and minimum normalization to generate the flue gas state vector; the operating parameters in the cost state operating parameter group are subjected to maximum and minimum normalization to generate the cost state vector. The equipment segment number sequence, time location sequence, flue gas state vector, and cost state vector are concatenated according to the node connection order to generate a path state matrix. The path state matrix is ​​vector-mapped using a fully connected mapping layer in the path embedding module to generate initial path features.

5. The method for optimizing the cost management of flue gas pollution control according to claim 1, characterized in that, The step of generating steady-state evolution features by performing continuous time-period evolution coding on the initial path features through a steady-state evolution coding module is as follows: The initial path features are divided into multiple consecutive time period feature segments according to the running time index, and arranged according to the node connection order in the joint consumption path sequence to generate a consecutive time period feature sequence; The continuous time period feature sequence is input into the one-dimensional temporal convolutional layer in the steady-state evolution coding module, and local evolution features between adjacent time periods are extracted along the running time index direction to generate a local evolution feature sequence. The local evolution feature sequence is input into the depth-separable convolutional layer in the steady-state evolution coding module. The channel features corresponding to each device segment are independently convolved, and the convolutional features corresponding to different device segments are combined and mapped across device segments by point-by-point convolution to generate the device segment evolution feature sequence. The device segment evolution feature sequence is input into the layer normalization layer in the steady-state evolution coding module. The device segment evolution features under the same running time index are subjected to layer normalization processing to generate a normalized evolution feature sequence. The normalized evolution feature sequence is input into the feedforward mapping layer in the steady-state evolution coding module. The normalized evolution features under each running time index are nonlinearly mapped by the GELU activation function to generate candidate evolution feature sequences. The candidate evolution feature sequence and the local evolution feature sequence are added together by residual addition to generate a fused evolution feature sequence; the fused evolution features corresponding to adjacent running time indices in the fused evolution feature sequence are differentially analyzed in one dimension to generate a feature difference sequence; Based on the positive and negative directions of the difference values ​​of each dimension in the feature difference sequence, feature dimensions that maintain the same direction of change under continuous running time index are extracted to generate a steady-state dimension label sequence. Based on the steady-state dimension labeling sequence, the corresponding fusion evolution feature dimension is retained from the fusion evolution feature sequence, and the fusion evolution feature dimensions that are not labeled as steady-state dimensions are set to zero to generate a steady-state preservation feature sequence. The steady-state evolution features are generated by linearly mapping the steady-state feature sequence through a linear output layer.

6. The method for optimizing the cost management of flue gas pollution control according to claim 1, characterized in that, The disturbance track rearrangement module rearranges the path evolution order of steady-state evolution features through a smoke load tidal migration mechanism, generating rearranged path features, specifically: The steady-state evolution features are input into the disturbance track rearrangement module, which is equipped with a smoke load tidal migration mechanism. Based on the running time index and equipment segment number corresponding to each feature dimension in the steady-state evolution features, the feature dimensions belonging to the same equipment segment under the same running time window are vector-concatenated according to the feature dimension order to restore the evolution features of the corresponding equipment segment. The device segment evolution characteristics in different runtime windows are continuously arranged according to the order of runtime index to generate a time period arrangement sequence; Extract the flue gas flow rate, power consumption, and reagent consumption corresponding to the evolution characteristics of each equipment segment in the time period sequence, and calculate the flue gas flow rate change direction, power consumption change direction, and reagent consumption change direction between adjacent operating time windows respectively. The evolution characteristics of adjacent equipment sections whose flue gas flow rate changes in the same direction as the power consumption change are identified as the first tidal migration segment. The evolution characteristics of adjacent equipment sections whose flue gas flow rate changes in the same direction as the reagent consumption change are identified as the second tidal migration segment. The evolution characteristics of adjacent equipment sections whose flue gas flow rate changes are inconsistent with the direction of power consumption changes or the direction of reagent consumption changes are identified as the third tidal migration segment. The magnitude of the change in power consumption in each first tidal migration segment is statistically analyzed, and the segments are arranged in ascending order of the magnitude of the change in power consumption. The variation amplitude of drug consumption in each second tidal migration segment was statistically analyzed, and the segments were arranged in ascending order of the variation amplitude of drug consumption. The amplitude of flue gas flow rate change corresponding to each third tidal migration segment is statistically analyzed, and the segments are arranged in descending order of flue gas flow rate change amplitude. The results of the first, middle and last sections are sequentially concatenated, and the running time index corresponding to the evolution characteristics of each device segment is retained to generate rearranged path features.

7. The method for optimizing the cost management of flue gas pollution control according to claim 1, characterized in that, The step of performing time-period trend analysis on the rearranged path features through the cost trend output module to generate a path cost evolution sequence is as follows: Input the rearranged path features into the cost trend output module, and extract the equipment segment evolution features, running time index, and corresponding arrangement position marker from the rearranged path features; Based on the arrangement position markers, corresponding path priority features are generated, where the first arrangement result corresponds to the first path priority feature, the middle arrangement result corresponds to the second path priority feature, and the last arrangement result corresponds to the third path priority feature. Extract the power consumption, reagent consumption and flue gas flow corresponding to the evolution characteristics of each equipment segment, calculate the difference in power consumption between adjacent operating time windows to obtain the direction of power consumption change, and calculate the difference in reagent consumption between adjacent operating time windows to obtain the direction of reagent consumption change. When both the direction of change in power consumption and the direction of change in reagent consumption are increasing, the direction of change in path cost between the corresponding path operating states is determined to be upward. When both the direction of change in power consumption and the direction of change in reagent consumption are decreasing, the direction of change in path cost between the corresponding path operating states is determined to be decreasing. When the direction of change in power consumption is inconsistent with the direction of change in reagent consumption, the direction of change in path cost between the corresponding path operation states is determined as the differentiation direction. The continuous path operation status that maintains a consistent direction of path cost change is defined as the same trend evolution segment; Record the direction of change in power consumption, the direction of change in reagent consumption, the direction of change in flue gas flow, and the corresponding equipment segment number in each trend evolution segment to generate segment cost characteristics; The segment cost features and the corresponding path priority features are concatenated according to the feature dimension order to generate spatiotemporal correlation features; The spatiotemporal correlation features are arranged continuously according to the running time index to generate a path cost evolution sequence.

8. The method for optimizing the cost management of flue gas pollution control according to claim 1, characterized in that, Step six specifically involves: Extract the spatiotemporal correlation features, running time index and corresponding equipment segment number from the path cost evolution sequence. Divide the spatiotemporal correlation features into continuous path evolution time periods based on the running time index, and determine the spatiotemporal correlation features in the same path evolution time period as the same path time period feature group. Input the time period feature groups of each path into the fully connected mapping layer, perform feature dimension mapping on the spatiotemporal correlation features in each time period feature group of the path, and generate the corresponding path attribution value; Softmax normalization is applied to the path attribution values ​​of each path during the same path evolution period to generate corresponding path attribution weights; Based on the running time index within the same path evolution period, calculate the sequence of changes in power consumption, reagent consumption, and equipment load between adjacent running time indices within the path evolution period; The path energy consumption characteristics are generated by multiplying the corresponding path affiliation weight with the energy consumption change sequence. The path-specific drug consumption characteristics are generated by multiplying the corresponding path-attribution weight with the drug consumption change sequence. The path load change characteristics are generated by multiplying the corresponding path affiliation weight with the equipment load change sequence. By combining the path's power consumption characteristics, path's reagent consumption characteristics, and path's load change characteristics, a single-path governance cost unit is generated.

9. The method for optimizing the cost management of flue gas pollution control according to claim 1, characterized in that, Step seven specifically involves: The path energy consumption characteristics, path reagent consumption characteristics, and path load change characteristics in the single path governance cost unit are respectively subjected to maximum and minimum normalization processing to generate normalized energy consumption characteristics, normalized reagent consumption characteristics, and normalized load change characteristics. The path contribution characteristics are generated by combining the normalized energy consumption characteristics, normalized reagent consumption characteristics, and normalized load change characteristics. The contribution characteristics of each path are arranged consecutively according to the running time index to generate a path contribution sequence; The path cost evolution sequence and path contribution sequence are input into the MCTS algorithm for tree search. The path contribution feature corresponding to the current running time index is used as the root node, and the path contribution features corresponding to multiple subsequent running time indices are used as candidate child nodes to generate a path search tree. The normalized energy consumption feature, normalized reagent consumption feature, and normalized load change feature in the path contribution features of each candidate sub-node are squared and then summed. The summation result is then squared to generate the corresponding path contribution value. When the contribution value of the path corresponding to the candidate child node is less than the contribution value of the path corresponding to the current root node, and the change in sulfur dioxide concentration and the change in nitrogen oxide concentration of the corresponding path are both less than the change in sulfur dioxide concentration and the change in nitrogen oxide concentration of the current root node, the corresponding candidate child node is determined as a low-cost candidate node. The candidate child node with the smallest path contribution value is selected from the low-cost candidate nodes as the expansion node, and the path contribution feature corresponding to the subsequent running time index of the expansion node is extracted as the next layer candidate child node until the path search tree reaches the set search layer. The sum of the path contribution values ​​in each path search branch is calculated, and the path search branch with the smallest sum of path contribution values ​​is taken as the governance path adjustment set.

10. The method for optimizing the cost management of flue gas pollution control according to claim 1, characterized in that, Step eight specifically involves: Extract the governance device paths, corresponding runtime indices, and corresponding path contribution values ​​from the governance path adjustment set, and sort the different governance device paths by time according to the runtime index to generate the governance adjustment order for the corresponding time period. When the path contribution value of the corresponding treatment equipment path is less than the path contribution value of the previous running time index, reduce the desulfurization tower slurry circulation volume and denitrification ammonia injection flow rate in the corresponding time period, and simultaneously reduce the induced draft fan load power. When the path contribution value of the corresponding treatment equipment path is greater than the path contribution value of the previous running time index, the circulation volume of desulfurization tower slurry and the flow rate of denitrification ammonia injection in the corresponding time period are increased, and the load power of the induced draft fan is increased simultaneously. When the path contribution value of the corresponding treatment equipment path is the same as the path contribution value of the previous running time index, the desulfurization tower slurry circulation volume, denitrification ammonia injection flow rate and induced draft fan load power in the corresponding time period remain unchanged. A time-sharing management and scheduling table is generated based on the desulfurization tower slurry circulation volume, denitrification ammonia injection flow rate, and induced draft fan load power for each time period. Based on the time-sharing management and scheduling table, the changes in electricity consumption, reagent consumption, and equipment load during the corresponding time periods are statistically analyzed to generate corresponding cost statistics. The cost statistics for each time period are summed up to output the cost optimization results.

Images