A rotary kiln body waste heat recovery process monitoring and optimization method

Abstract

The present application relates to the field of industrial energy saving and waste heat recovery, and discloses a kind of rotary kiln body waste heat recovery process monitoring and optimization method, by obtaining the space-time positioning of shell surface temperature data and operating state, two-dimensional heat exchange area grid is divided and the waste heat recovery efficiency index of each grid unit is obtained;Based on the heat transfer relationship of efficiency index analysis, the heat flow direction atlas is constructed, the abnormal grid unit of efficiency attenuation is identified, and the upstream disturbance grid unit causing attenuation is located according to atlas;The heat disturbance contribution of upstream disturbance grid unit is calculated, and the independent heat exchange control parameter is generated by spatial decoupling processing of boundary heat flow;Finally, the maximum global waste heat recovery amount is taken as the target, the optimization distribution of each grid heat exchange control amount is carried out, and the control strategy is output.The method can accurately locate the source of efficiency attenuation and eliminate the thermal coupling interference between units through grid monitoring, disturbance tracing and spatial decoupling, so as to realize the efficient optimization control of waste heat recovery.

Description

Technical Field

[0001] This invention relates to the field of industrial energy conservation and waste heat recovery technology, and more specifically to a method for monitoring and optimizing the waste heat recovery process of a rotary kiln. Background Technology

[0002] Rotary kilns, as core thermal equipment in metallurgy, building materials, and chemical industries, generate significant waste heat on their surface during operation. Effectively recovering this heat has become crucial for improving energy efficiency, reducing production energy consumption, and achieving energy conservation and emission reduction goals. With the increasing demands for refined energy management in modern industry, real-time monitoring of the heat flow distribution on the kiln surface under complex and variable operating conditions, and dynamic monitoring and optimization of the waste heat recovery process, have become indispensable key aspects of intelligent management and control of current industrial energy systems. Currently, monitoring technology for waste heat recovery in rotary kilns primarily relies on infrared thermometers arranged along the kiln's axial direction to obtain overall temperature field data on the kiln surface, which serves as a basic reference for the macroscopic operation of the heat exchange equipment. At the management strategy level, existing technologies mostly rely on static adjustment of the heat exchange medium flow rate based on feedback from overall temperature measurement points, or setting fixed temperature thresholds to trigger equipment start / stop and alarm commands, thereby achieving basic waste heat acquisition and operational protection.

[0003] However, the aforementioned existing technologies have the following shortcomings in practical applications:

[0004] Firstly, existing monitoring methods mostly focus on acquiring macroscopic overall temperature field data, lacking spatiotemporal dynamic positioning based on running time and rotation state, and failing to discretize the cylinder surface into a fine-grained two-dimensional grid for feature extraction, resulting in insufficient granularity of the underlying data, which cannot support refined performance evaluation.

[0005] Secondly, existing control methods often treat each temperature measurement point or area as an isolated object when dealing with the heat exchange process, ignoring the direction of heat transfer and coupling relationship between adjacent areas. This makes it impossible to construct a global heat flow pattern and also makes it difficult to trace and locate the upstream disturbance source that causes local performance degradation.

[0006] Third, because the boundary heat flow between grids is not spatially decoupled, the existing technology cannot eliminate the cross-effects caused by thermal disturbances in adjacent areas, making it difficult to generate truly independent heat transfer control parameters.

[0007] Fourth, in terms of waste heat recovery regulation and allocation, existing technologies mostly rely on passive adjustment triggered by static thresholds or simple linear on-demand allocation, lacking an optimization solution mechanism oriented towards maximizing the total global waste heat recovery. They are prone to global suboptimal problems due to local over-regulation or coupling instability, and cannot achieve the optimal allocation of regulation amount under limited resources. Summary of the Invention

[0008] In order to overcome the above-mentioned defects of the prior art, the present invention provides a method for monitoring and optimizing the waste heat recovery process of rotary kiln body, so as to solve the problems existing in the background art.

[0009] This invention provides the following technical solution: a method for monitoring and optimizing the waste heat recovery process of a rotary kiln, comprising:

[0010] S1: Obtain the temperature data of the rotary kiln shell surface and its corresponding running time and rotation state information. Then, based on the running time and rotation state information, perform spatiotemporal positioning of the temperature data and extract temperature distribution features. Based on the temperature distribution features, discretize the shell surface into a two-dimensional heat exchange area grid to form a temperature distribution feature dataset for each grid unit.

[0011] S2: Based on the temperature distribution characteristic dataset of each grid cell, perform waste heat recovery efficiency analysis on each grid cell to obtain the waste heat recovery efficiency index of each grid cell;

[0012] S3: Based on the waste heat recovery efficiency index of each grid cell, analyze the heat transfer direction and heat transfer coupling relationship between adjacent grid cells, and construct a heat flow path map of the dynamic heat transfer path;

[0013] S4: Based on the waste heat recovery efficiency index, identify abnormal grid cells with efficiency degradation, analyze the influence and transmission relationship of abnormal grid cells on the dynamic heat transfer path according to the heat flow direction map, and locate the upstream disturbed grid cells that cause efficiency degradation.

[0014] S5: Based on the upstream disturbed grid cell, calculate its thermal disturbance contribution to the abnormal grid cell, and perform spatial decoupling processing on the boundary heat flow of each grid cell based on the thermal disturbance contribution to generate independent heat transfer control parameters for each grid cell.

[0015] S6: Based on independent heat exchange control parameters, with the goal of maximizing the total global waste heat recovery, the actual heat exchange regulation of each grid unit is optimized and allocated, and the waste heat recovery optimization control strategy is output.

[0016] Preferably, step S1, acquiring the temperature data of the rotary kiln shell surface and its corresponding running time and rotation status information, includes:

[0017] Infrared radiation energy data of the cylinder surface is collected along the axial and circumferential directions using a non-contact scanning method that tracks the cylinder rotation synchronously. The temperature data of the cylinder surface is obtained by converting the infrared radiation energy data. The acquisition time corresponding to the temperature data is recorded synchronously, and the current running time information is calculated by combining the kiln start and stop records.

[0018] The rotational angular velocity and circumferential phase reference signal of the cylinder are acquired synchronously, and the rotational state information is composed of the rotational angular velocity and circumferential phase reference signal;

[0019] The process involves spatiotemporal localization of temperature data and extraction of temperature distribution features. Based on these features, the cylinder surface is discretized into a two-dimensional heat transfer region grid, forming a temperature distribution feature dataset for each grid cell, including:

[0020] The rotational angular velocity of the cylinder is extracted based on the rotational state information, and the circumferential phase angle of the cylinder surface is calculated based on the running time and rotational angular velocity. The one-dimensional temperature observation data collected along the time series is mapped and reconstructed into a two-dimensional cylindrical unfolded surface temperature matrix, with the axial position of the cylinder from the kiln head to the kiln tail as the first coordinate and the circumferential phase angle as the second coordinate.

[0021] Set the axial and circumferential step size of the sliding window, traverse the temperature matrix of the two-dimensional cylindrical unfolded surface using the sliding window, calculate the spatial temperature gradient magnitude and gradient direction for the neighborhood of the matrix covered by the sliding window, extract the temperature data of the center position of the sliding window in the historical time series to calculate the time variation variance, and concatenate the spatial temperature gradient magnitude, gradient direction and time variation variance into vectors to generate temperature distribution features.

[0022] Set gradient threshold and variance threshold, and mark the region with spatial temperature gradient amplitude greater than the gradient threshold or time variation variance greater than the variance threshold as high heat flux characteristic region, and mark the region with spatial temperature gradient amplitude less than or equal to the gradient threshold and time variation variance less than or equal to the variance threshold as low heat flux characteristic region.

[0023] For the high heat flux characteristic region, a dense grid is generated using the first grid scale, and for the low heat flux characteristic region, a sparse grid is generated using the second grid scale, which is larger than the first grid scale. The boundary between the dense and sparse grids forms a non-uniform two-dimensional heat transfer region grid.

[0024] The temperature distribution features corresponding to each grid cell after division are extracted and aggregated to form a temperature distribution feature dataset.

[0025] Preferably, step S2 performs waste heat recovery efficiency analysis on each grid cell to obtain the waste heat recovery efficiency index of each grid cell, including:

[0026] Extract the aggregated temperature distribution features of each grid cell from the temperature distribution feature dataset, and parse the historical temperature sequence, spatial temperature gradient magnitude, spatial temperature gradient direction and time variation variance of the current grid cell based on the temperature distribution features;

[0027] The cumulative temperature integral value is calculated based on the historical temperature sequence of the current grid cell, and the slope feature of the historical temperature sequence of the current grid cell on the time axis is extracted as the dynamic temperature offset. The cumulative temperature integral value and the dynamic temperature offset are subtracted from the reference offset corresponding to the preset environmental reference temperature. The subtraction results are then weighted and fused to calculate and output the heat exchange potential energy factor of the current grid cell.

[0028] The time variation coefficient is obtained by calculating the ratio of the time variation variance to the mean of the historical temperature sequence of the current grid cell. The time variation coefficient is then reverse normalized to obtain the time stability coefficient. The heat flow stability penalty factor is determined based on the time stability coefficient. The larger the value of the time variation coefficient, the smaller the value of the heat flow stability penalty factor.

[0029] The local heat flux is calculated based on the magnitude and direction of the spatial temperature gradient. The local heat flux is then normalized to obtain the spatial diffusion coefficient. The local heat dissipation attenuation factor is determined based on the spatial diffusion coefficient. The larger the value of the local heat flux, the smaller the value of the local heat dissipation attenuation factor.

[0030] The heat exchange potential energy factor, heat flow stability penalty factor and local heat dissipation attenuation factor are multiplied and weighted and normalized to calculate and output the comprehensive quantitative value of each grid cell as the waste heat recovery efficiency index.

[0031] Preferably, step S3 analyzes the heat transfer direction and heat transfer coupling relationship between adjacent grid cells to construct a heat flow path map of dynamic heat transfer paths, including:

[0032] The axial position and circumferential phase angle of each grid cell are extracted from the temperature distribution feature dataset, and the gradient direction of each grid cell is extracted. At the same time, the waste heat recovery efficiency index of each grid cell is obtained.

[0033] The space heat flow transfer vector is obtained by projecting the gradient direction along the line connecting the axial position of the adjacent grid cell and the circumferential phase angle; the difference in the waste heat recovery efficiency index between adjacent grid cells is calculated as the efficiency potential energy difference, and the efficiency potential energy difference is multiplied with the space heat flow transfer vector as a scalar modulation coefficient with a positive or negative sign to obtain the efficiency-driven heat flow vector; the direction of heat transfer between adjacent grid cells is determined according to the direction of the efficiency-driven heat flow vector.

[0034] Historical temperature sequences of adjacent grid cells are extracted from the temperature distribution feature dataset. The time delay cross-correlation coefficient between the historical temperature sequences of adjacent grid cells is calculated as the time dynamic response similarity, and the time delay parameter corresponding to the maximum time dynamic response similarity is extracted. The time dynamic response similarity is used as the time coupling benchmark value, and the absolute value of the efficiency potential energy difference is used as the spatial driving weight. The product of the time coupling benchmark value and the spatial driving weight is calculated, and the product is positively normalized. The heat transfer coupling strength coefficient between adjacent grid cells is calculated and output.

[0035] Each grid cell is mapped to a node in a graph data structure, the heat transfer direction is mapped to a directed edge connecting the nodes, the heat transfer coupling strength coefficient is mapped to the edge weight of the directed edge, and the time delay parameter is mapped to the time span attribute of the directed edge. Edges with a heat transfer coupling strength coefficient greater than a preset coupling threshold are selected to construct a time series graph adjacency matrix. Based on the time series graph adjacency matrix and the time span attribute, dynamic connectivity paths on the time series are deduced to generate a heat flow map of dynamic heat transfer paths.

[0036] Preferably, step S4 identifies abnormal grid cells with performance degradation based on the waste heat recovery efficiency index, including:

[0037] Based on the waste heat recovery efficiency index of multiple continuous historical time windows, a time series sequence of waste heat recovery efficiency index for each grid cell is constructed. The negative rate of change of the waste heat recovery efficiency index time series sequence within the time window is calculated as the time decay rate. The difference between the current grid cell waste heat recovery efficiency index and the mean of the waste heat recovery efficiency index of grid cells sharing the boundary in a two-dimensional coordinate system formed by axial position and circumferential phase angle is calculated as the spatial dispersion. The time decay rate and spatial dispersion are weighted and fused to obtain the efficiency decay index. The efficiency decay indices of all grid cells are summarized into an efficiency decay index sample set. The mean and standard deviation of the efficiency decay index sample set are calculated. The mean is added to a set multiple of the standard deviation to obtain a dynamic adaptive threshold. Grid cells with an efficiency decay index greater than the dynamic adaptive threshold are identified as abnormal grid cells with efficiency decay.

[0038] The process of analyzing the influence and transmission relationship of abnormal grid cells on the dynamic heat transfer path based on the heat flow pattern, and locating the upstream disturbed grid cells that cause performance degradation, includes:

[0039] Based on the heat flow path map, a set of directed edges pointing to anomalous grid cells is extracted. The edge weights and time span attributes corresponding to the directed edge sets are extracted from the heat flow path map. The edge weights are used as steady-state coupling potential coefficients, and the reciprocal of the time span attribute is used as dynamic transmission rate coefficients. The product of the steady-state coupling potential coefficient and the dynamic transmission rate coefficient is calculated as the influence conductance. The influence conductance is used to characterize the coupling transmission flux of the disturbance signal per unit time. The influence conductance is used as the influence transmission relationship of anomalous grid cells on the dynamic heat transfer path.

[0040] Based on the influence propagation relationship, a reverse traversal is performed along the directed edges pointing to the abnormal grid cells. During the reverse traversal, the abrupt change time points of the performance decay of the waste heat recovery efficiency index time series of each upstream grid cell are extracted, and the actual delay difference between the abrupt change time point and the performance decay time of the abnormal grid cell is calculated. The absolute value of the difference between the actual delay difference and the time span attribute is calculated as the time series delay error. A preset time series allowable error threshold is extracted, and the difference between the preset time series allowable error threshold and the time series delay error is calculated. The ratio of the difference to the preset time series allowable error threshold is used as the time series fit.

[0041] The attenuation source probability is calculated by multiplying the influence of conductance and temporal fit. It is then determined whether there are upstream grid cells whose attenuation source probability is greater than a preset probability threshold. If so, the upstream grid cell with the highest attenuation source probability is selected and identified as the upstream disturbance grid cell that caused the performance attenuation. If the attenuation source probability of all upstream grid cells is less than or equal to the preset probability threshold, the performance attenuation of the abnormal grid cell is determined to be caused by its own internal factors, triggering a local anomaly warning for the grid cell and continuing to monitor the area in real time.

[0042] Preferably, step S5 calculates the contribution of upstream disturbed mesh cells to the thermal disturbance of anomalous mesh cells, including:

[0043] Extract the difference in waste heat recovery efficiency index of upstream perturbation grid cells before and after the abrupt change in efficiency decay time point, calculate the ratio of this difference to the waste heat recovery efficiency index before the abrupt change time point as the efficiency deviation rate, and calculate the absolute value of the efficiency deviation rate as the efficiency deviation magnitude; extract the directed transmission path connecting upstream perturbation grid cells and abnormal grid cells from the heat flow path map, calculate the geometric mean of the influence conductivity of each node on the directed transmission path as the path coupling degree, and multiply the efficiency deviation magnitude by the path coupling degree to calculate the thermal perturbation contribution.

[0044] The spatial decoupling of boundary heat flow for each grid cell based on thermal disturbance contribution is performed to generate independent heat transfer control parameters for each grid cell, including:

[0045] The surface temperature values ​​of the current abnormal grid cells within a continuous historical time window are extracted from the temperature distribution feature dataset. The absolute value of the difference between the surface temperature value and the preset ambient temperature within each time window is calculated to obtain the grid temperature difference sequence. The rate of change of the grid temperature difference sequence within the time window is calculated as the temperature difference change rate. The boundary heat flux at the current moment is calculated by multiplying the temperature difference change rate by the waste heat recovery efficiency index.

[0046] Calculate the product of thermal disturbance contribution and boundary heat flux, input the product into a preset nonlinear decay function to calculate the dynamic damping disturbance; calculate the difference between boundary heat flux and dynamic damping disturbance as its own heat transfer fluctuation.

[0047] Extract the preset ideal heat transfer benchmark, calculate the difference between its own heat transfer fluctuation and the preset ideal heat transfer benchmark, input the difference into the preset nonlinear mapping function to calculate the heat flow compensation coefficient, multiply the heat flow compensation coefficient with the preset basic heat transfer adjustment factor to generate the independent heat transfer control parameters of each grid unit.

[0048] Preferably, step S6 optimizes the allocation of actual heat exchange control quantities corresponding to each grid unit with the goal of maximizing the total global waste heat recovery, and outputs an optimized waste heat recovery control strategy, including:

[0049] An initial control demand vector is constructed based on independent heat exchange control parameters. The waste heat recovery efficiency index of each grid cell is extracted. The product of the independent heat exchange control parameters and the waste heat recovery efficiency index is calculated as the marginal recovery gain characteristic value of each grid cell.

[0050] Extract the thermal disturbance contribution of each grid cell, calculate the product of the independent heat transfer control parameter and the thermal disturbance contribution as the potential instability risk, input the potential instability risk into the preset penalty mapping function, and calculate the regulation penalty factor; calculate the difference between the marginal recovery gain characteristic value and the regulation penalty factor as the net recovery benefit evaluation value.

[0051] Based on the net recovery benefit assessment value of each grid unit, its corresponding exponential characteristic value is calculated, and the sum of the exponential characteristic values ​​of all grid units is calculated as the global response sum; the ratio of the exponential characteristic value of each grid unit to the global response sum is used as the actual control allocation weight of each grid unit.

[0052] Extract the preset global total control constraint threshold, multiply the actual control allocation weight by the global total control constraint threshold, and calculate the initial actual heat transfer control amount corresponding to each grid unit.

[0053] An objective function is constructed based on maximizing the total global waste heat recovery as the optimization criterion. The objective function is the global cumulative value of the product of the actual heat exchange control amount of each grid cell and its corresponding waste heat recovery efficiency index. The actual control allocation weights are iteratively corrected based on the objective function. Specifically, after calculating the actual heat exchange control amount in each round, the objective function value of the current round is calculated. If it is higher than the previous round, the current weight is retained. If it is not higher, it is reverted to the weight of the previous round and the sensitivity amplification coefficient of the exponential characteristic value is dynamically adjusted. The weights and actual heat exchange control amounts are recalculated until the objective function value converges.

[0054] The optimal actual heat transfer regulation obtained after convergence is mapped and associated with the corresponding grid cell identifier to generate an optimized control strategy for waste heat recovery.

[0055] The technical effects and advantages of this invention are as follows:

[0056] (1) By introducing running time and rotation state information, the temperature data is precisely located in time and space, and the surface of the cylinder is discretized into a fine-grained two-dimensional heat exchange area grid. This process can greatly improve the granularity and spatiotemporal resolution of the data from the bottom up, so that each grid unit has an independent temperature distribution feature dataset, laying a solid data foundation for subsequent high-precision performance evaluation and refined control.

[0057] (2) By deeply analyzing the heat transfer direction and coupling relationship between adjacent grids, a global dynamic heat flow path map was constructed. Based on this map, this method can analyze the influence relationship of abnormal grids on the transfer path and accurately locate the upstream disturbance grid cells that cause performance degradation. This technical effect realizes the leap from "passively discovering anomalies" to "actively tracing and locating the upstream disturbance source", providing a clear roadmap for subsequent targeted decoupling control.

[0058] (3) Based on the upstream disturbance grid cell of the location, the contribution of its thermal disturbance to the abnormal grid is quantitatively calculated, and the boundary heat flow of each grid cell is spatially decoupled based on this. This effectively removes the dynamic coupling interference between adjacent grids, so that the final generated independent heat transfer control parameters can truly reflect the decoupling control requirements of a single grid, and completely eliminate the systemic coupling instability risk of "affecting the whole body by moving one part".

[0059] (4) By iteratively optimizing the allocation of the actual heat exchange control amount of each grid with objective function verification, the optimal control amount allocation under limited resources is achieved under the premise of strictly satisfying the global total control amount constraint. At the same time, since the allocation process is based on the completely decoupled independent heat exchange control parameters generated in the previous step S5, the coupling instability caused by thermal disturbance in adjacent areas is fundamentally eliminated, which is conducive to completely avoiding the global suboptimal problem caused by local over-control or coupling instability, and truly maximizing the global waste heat recovery benefits under limited resources. Attached Figure Description

[0060] Figure 1 This is a diagram illustrating the method steps of the present invention.

[0061] Figure 2 This is a system structure block diagram of the present invention. Detailed Implementation

[0062] The technical solutions of the present invention will be clearly and completely described below with reference to the accompanying drawings. In addition, the forms of the various structures described in the following embodiments are merely illustrative. The method for monitoring and optimizing the waste heat recovery process of a rotary kiln body involved in the present invention is not limited to the structures described in the following embodiments. All other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0063] like Figure 1 The embodiment shown provides a method for monitoring and optimizing the waste heat recovery process of a rotary kiln, including:

[0064] S1: Obtain the temperature data of the rotary kiln shell surface and its corresponding running time and rotation state information. Then, based on the running time and rotation state information, perform spatiotemporal positioning of the temperature data and extract temperature distribution features. Based on the temperature distribution features, discretize the shell surface into a two-dimensional heat exchange region grid to form a temperature distribution feature dataset for each grid unit.

[0065] In this embodiment, step S1, acquiring the temperature data of the rotary kiln shell surface and its corresponding running time and rotation state information, includes:

[0066] Infrared radiation energy data of the cylinder surface is collected along the axial and circumferential directions using a non-contact scanning method that tracks the cylinder rotation synchronously. The temperature data of the cylinder surface is obtained by converting the infrared radiation energy data. The acquisition time corresponding to the temperature data is recorded synchronously, and the current running time information is calculated by combining the kiln start and stop records.

[0067] The rotational angular velocity and circumferential phase reference signal of the cylinder are acquired synchronously, and the rotational state information is composed of the rotational angular velocity and circumferential phase reference signal;

[0068] The process involves spatiotemporal localization of temperature data and extraction of temperature distribution features. Based on these features, the cylinder surface is discretized into a two-dimensional heat transfer region grid, forming a temperature distribution feature dataset for each grid cell, including:

[0069] The rotational angular velocity of the cylinder is extracted based on the rotational state information, and the circumferential phase angle of the cylinder surface is calculated based on the running time and rotational angular velocity. The one-dimensional temperature observation data collected along the time series is mapped and reconstructed into a two-dimensional cylindrical unfolded surface temperature matrix, with the axial position of the cylinder from the kiln head to the kiln tail as the first coordinate and the circumferential phase angle as the second coordinate.

[0070] Set the axial and circumferential step size of the sliding window, traverse the temperature matrix of the two-dimensional cylindrical unfolded surface using the sliding window, calculate the spatial temperature gradient magnitude and gradient direction for the neighborhood of the matrix covered by the sliding window, extract the temperature data of the center position of the sliding window in the historical time series to calculate the time variation variance, and concatenate the spatial temperature gradient magnitude, gradient direction and time variation variance into vectors to generate temperature distribution features.

[0071] Set gradient threshold and variance threshold, and mark the region with spatial temperature gradient amplitude greater than the gradient threshold or time variation variance greater than the variance threshold as high heat flux characteristic region, and mark the region with spatial temperature gradient amplitude less than or equal to the gradient threshold and time variation variance less than or equal to the variance threshold as low heat flux characteristic region.

[0072] For the high heat flux characteristic region, a dense grid is generated using the first grid scale, and for the low heat flux characteristic region, a sparse grid is generated using the second grid scale, which is larger than the first grid scale. The boundary between the dense and sparse grids forms a non-uniform two-dimensional heat transfer region grid.

[0073] The temperature distribution features corresponding to each grid cell after division are extracted and aggregated to form a temperature distribution feature dataset.

[0074] It should be specifically explained that the specific data processing flow for acquiring the temperature data of the rotary kiln shell surface and its corresponding running time and rotation status information, and performing spatiotemporal positioning and mesh generation on the temperature data, is as follows:

[0075] Along the axial and circumferential directions of the cylinder, a non-contact scanning method that tracks the cylinder's rotation synchronously (for example, in this embodiment, an infrared scanning thermometer arranged on the side of the kiln body can be used to continuously scan the rotating cylinder surface at a fixed frequency) is employed to collect infrared radiation energy data from the cylinder surface. Based on the calibration curve of Planck's blackbody radiation law relating infrared radiation energy and temperature, the infrared radiation energy data is converted into a specific temperature value matrix.

[0076] The absolute acquisition time (e.g., 10:05:30 Beijing time) for each temperature acquisition is recorded synchronously. To eliminate the accumulation of invalid time caused by equipment downtime, the kiln start-up and shutdown records (e.g., Boolean signals of motor operation status output by the PLC system) are acquired. The time periods when the kiln is in a downtime state are identified and removed from the absolute acquisition time, and the current continuous running time information after removing the downtime period is calculated. For example, if the acquisition time is 10:05:30, but the kiln is stopped between 10:00:00 and 10:03:00, the current running time information is calculated as 05 minutes and 30 seconds minus 03 minutes of downtime, that is, the actual running time is 02 minutes and 30 seconds. The rotational angular velocity and circumferential phase reference signal of the cylinder are acquired synchronously (e.g., the pulse signal generated by a specific marker point of the cylinder for each revolution is acquired through a proximity switch as the 0° phase reference), and the rotational state information is composed of the rotational angular velocity and this reference signal.

[0077] Extract the rotational angular velocity of the cylinder based on the rotational state information. Based on the current running time information t and angular velocity ω, the circumferential phase angle of the cylinder surface is calculated. The specific calculation formula is as follows: Using the axial position x of the cylinder from the kiln head to the kiln tail (e.g., 0m to 60m) as the first coordinate (i.e., the row index of the matrix), and the circumferential phase angle... The second coordinate (i.e., the column index of the matrix) maps the original one-dimensional temperature observation data collected along the time series to a two-dimensional coordinate system according to its corresponding axial position and calculated circumferential phase angle. For coordinate points that were not directly collected, bilinear interpolation can be used to complete them, and finally the time series one-dimensional data is reconstructed into a two-dimensional cylindrical unfolded surface temperature matrix with axial position and circumferential angle as coordinate axes.

[0078] The axial and circumferential step sizes of the sliding window are set. In this embodiment, the axial step size is set to 1 meter, and the circumferential step size is set to 5° (the specific step size can be adjusted according to the cylinder size and temperature measurement density). The sliding window is used to traverse the temperature matrix of the two-dimensional cylindrical unfolded surface. Spatial gradient calculation is performed on the neighborhood of the matrix covered by the sliding window: the central difference method is used to calculate the rate of temperature change at the center point of the window in the axial and circumferential directions, and the square root of the sum of squares is used to obtain the spatial temperature gradient amplitude, and the gradient direction is calculated by the arctangent function. At the same time, the temperature data of the center position of the sliding window in the historical time series is extracted (for example, the temperature values ​​of the past 10 consecutive scanning cycles), and the variance of this set of temperature data is calculated as the time variation variance. The calculated spatial temperature gradient amplitude, gradient direction and time variation variance are concatenated by connecting the first and last elements of the vector according to the dimension to generate the temperature distribution feature vector corresponding to the center position of the window.

[0079] Set gradient and variance thresholds. In this embodiment, the threshold setting rule is as follows: obtain the gradient and variance statistics of the entire cylinder under historical normal operating conditions, set the gradient threshold to 1.5 times the historical average gradient value (e.g., 30℃ / m), and set the variance threshold to 1.5 times the historical average variance value (e.g., 15℃²). Regions with a spatial temperature gradient amplitude greater than 30℃ / m or a time variation variance greater than 15℃² indicate drastic temperature changes or thermal disturbances, and are marked as high heat flux characteristic regions; regions with a spatial temperature gradient amplitude less than or equal to 30℃ / m and a time variation variance less than or equal to 15℃² indicate relatively stable temperatures, and are marked as low heat flux characteristic regions. A first grid scale (e.g., 0.5m) is used for high heat flux characteristic regions. A dense grid is used (2°) to capture fine heat flux variations; a second grid scale larger than the first grid scale (e.g., 2m) is used for low heat flux characteristic regions. A sparse mesh is generated at 0° to save computational resources. The boundary between the dense and sparse mesh regions naturally forms the boundary, creating a non-uniform two-dimensional heat transfer region mesh covering the entire surface of the cylinder.

[0080] After the non-uniform grid is generated, each grid cell contains several temperature distribution feature vectors generated earlier. All temperature distribution feature vectors within each grid cell are aggregated (e.g., the mean of the feature vectors in each dimension is calculated, or principal component features are extracted) to form a unique comprehensive temperature distribution feature for each grid cell. The comprehensive temperature distribution features of all grid cells are then combined to form the final temperature distribution feature dataset.

[0081] S2: Based on the temperature distribution characteristic dataset of each grid cell, the waste heat recovery efficiency of each grid cell is analyzed to obtain the waste heat recovery efficiency index of each grid cell.

[0082] In this embodiment, S2 performs waste heat recovery efficiency analysis on each grid cell to obtain the waste heat recovery efficiency index of each grid cell, including:

[0083] Extract the aggregated temperature distribution features of each grid cell from the temperature distribution feature dataset, and parse the historical temperature sequence, spatial temperature gradient magnitude, spatial temperature gradient direction and time variation variance of the current grid cell based on the temperature distribution features;

[0084] The cumulative temperature integral value is calculated based on the historical temperature sequence of the current grid cell, and the slope feature of the historical temperature sequence of the current grid cell on the time axis is extracted as the dynamic temperature offset. The cumulative temperature integral value and the dynamic temperature offset are subtracted from the reference offset corresponding to the preset environmental reference temperature. The subtraction results are then weighted and fused to calculate and output the heat exchange potential energy factor of the current grid cell.

[0085] The time variation coefficient is obtained by calculating the ratio of the time variation variance to the mean of the historical temperature sequence of the current grid cell. The time variation coefficient is then reverse normalized to obtain the time stability coefficient. The heat flow stability penalty factor is determined based on the time stability coefficient. The larger the value of the time variation coefficient, the smaller the value of the heat flow stability penalty factor.

[0086] The local heat flux is calculated based on the magnitude and direction of the spatial temperature gradient. The local heat flux is then normalized to obtain the spatial diffusion coefficient. The local heat dissipation attenuation factor is determined based on the spatial diffusion coefficient. The larger the value of the local heat flux, the smaller the value of the local heat dissipation attenuation factor.

[0087] The heat exchange potential energy factor, heat flow stability penalty factor and local heat dissipation attenuation factor are multiplied and weighted and normalized to calculate and output the comprehensive quantitative value of each grid cell as the waste heat recovery efficiency index.

[0088] It should be noted that the specific data processing and analysis process for analyzing the waste heat recovery efficiency of each grid cell to obtain the waste heat recovery efficiency index of each grid cell is as follows:

[0089] From the temperature distribution feature dataset finally formed in step S1, extract the aggregated temperature distribution features of each grid cell. Since the temperature distribution features generated in step S1 are a vector composed of the spatial temperature gradient magnitude, gradient direction, and time variance, this vector needs to be parsed and decomposed to reconstruct the historical temperature sequence, spatial temperature gradient magnitude, spatial temperature gradient direction, and time variance of the currently being calculated grid cell (hereinafter referred to as the current grid cell) within the historical observation window. The historical temperature sequence is the set of temperatures for that grid cell over the past N consecutive sampling periods, for example... .

[0090] Based on the historical temperature sequence of the current grid cells, the cumulative temperature integral value is calculated using the trapezoidal integral method. The calculation formula is as follows: ,in The historical temperature of the i-th sampling point in the past. This represents the historical temperature of the (i+1)th sampling point in the past. The sampling period is defined as the integral value, representing the overall heat enrichment of the grid over a historical time period. Simultaneously, the slope characteristics of the historical temperature series on the time axis are extracted. Specifically, the least squares method can be used to... A linear fit is performed, and the slope of the resulting line is denoted as k. Since this slope k reflects the deviation trend of temperature dynamics over time (i.e., the rate of temperature increase or decrease), this slope characteristic k is directly used as the dynamic temperature offset in this scheme. (i.e., order) Set the preset ambient reference temperature. (For example, set to 25℃). Since the ambient reference temperature is constant, its corresponding time axis slope is 0, therefore the reference offset... Simultaneously calculate the baseline integral. Subtract the corresponding baseline value from the cumulative temperature integral value and the dynamic temperature offset to obtain the net temperature rise integral. and net dynamic offset ,because Therefore, net dynamic offset The subtraction results are then weighted and fused to calculate and output the heat exchange potential energy factor of the current grid cell. The formula is: Among them, the weighting coefficient and The setting rule is: based on historical operating data statistics, this embodiment can... Set to 0.7 to focus on total heat, The value is set to 0.3 to account for trend changes. Calculation results may be negative, indicating that the temperature in the area is below the environmental baseline and there is no waste heat to recover.

[0091] Calculate the time variation variance Compared with the historical temperature series average of the current grid cells The ratio of the two values ​​is used to obtain the time series variation coefficient. This coefficient eliminates the dimensional misleading effect of absolute temperature variations on variance. The time-series variation coefficient is obtained by inversely normalizing it. Inverse normalization refers to a processing method where the larger the value, the smaller the output. The specific formula is: ,in This is a preset threshold for the maximum historical coefficient of variation of the entire cylinder (e.g., set to 0.5). When When it approaches 0, A value approaching 1 indicates extremely stable heat flow; when When the threshold of 0.5 is reached, A value of 0 indicates extreme fluctuations in heat flux. The heat flux stability penalty factor is determined based on the time stability coefficient. In this embodiment, the direct As Clearly, the time series variation coefficient The larger the value, the more drastic the temperature fluctuation and the greater the heat flow stability penalty factor. The smaller the value, the better.

[0092] Based on the magnitude G and direction of the space temperature gradient Calculate local heat flux divergence In the temperature matrix of the two-dimensional cylindrical unfolded surface in step S1, the formula for discretizing the divergence is: ,in and It is to determine the gradient magnitude G based on the gradient direction. The components projected onto the axial and circumferential directions. Divergence. The physical meaning of divergence is the degree to which heat at a grid point disperses outward or concentrates inward, and it carries a positive or negative sign. The spatial diffusion coefficient F is obtained by positively normalizing the local heat flux divergence. Positive normalization means that a larger value results in a larger output. Considering that divergence carries a positive or negative sign, to ensure the monotonicity of the physical meaning, this embodiment first extracts the absolute value of the local heat flux divergence. As the intensity of local heat loss, the formula is then used. Perform positive normalization, where The maximum heat loss intensity threshold for the entire cylinder is preset. The local heat dissipation attenuation factor is determined based on the spatial diffusion coefficient. In this embodiment, the formula is set as follows: Clearly, the larger the value of the local heat loss intensity (i.e., the absolute value of the divergence), the more severe the heat loss to the surrounding areas, and the greater the local heat dissipation attenuation factor. The smaller the value, the better.

[0093] Heat exchange potential factor Heat flow stability penalty factor With local heat dissipation attenuation factor Perform product operations and weighted normalization. Using product operations reflects the parallel constraints among the three factors: a grid must simultaneously possess high potential energy, high stability, and low dissipation to have high recycling value. The final calculation formula is: ,in This is the global scaling factor (for example, set to 100 so that the exponential output is within the range of 0-100 for easy and intuitive comparison). This is a preset maximum reference value for heat exchange potential energy, used for normalization. The calculated comprehensive quantized value is the waste heat recovery efficiency index of the current grid cell. .when When the value is negative (below ambient temperature), the output efficiency index is negative, indicating that the area is a cold source area; when When the value is positive, the output efficiency index is positive; the larger the value, the higher the waste heat recovery value and feasibility of the grid cell. Based on this, a traversal calculation is performed on all grid cells of the entire cylinder, generating an efficiency index matrix for subsequent optimization steps.

[0094] S3: Based on the waste heat recovery efficiency index of each grid cell, analyze the heat transfer direction and heat transfer coupling relationship between adjacent grid cells, and construct a heat flow path map of the dynamic heat transfer path.

[0095] In this embodiment, step S3 analyzes the heat transfer direction and heat transfer coupling relationship between adjacent grid cells to construct a heat flow path map of dynamic heat transfer paths, including:

[0096] The axial position and circumferential phase angle of each grid cell are extracted from the temperature distribution feature dataset, and the gradient direction of each grid cell is extracted. At the same time, the waste heat recovery efficiency index of each grid cell is obtained.

[0097] The space heat flow transfer vector is obtained by projecting the gradient direction along the line connecting the axial position of the adjacent grid cell and the circumferential phase angle; the difference in the waste heat recovery efficiency index between adjacent grid cells is calculated as the efficiency potential energy difference, and the efficiency potential energy difference is multiplied with the space heat flow transfer vector as a scalar modulation coefficient with a positive or negative sign to obtain the efficiency-driven heat flow vector; the direction of heat transfer between adjacent grid cells is determined according to the direction of the efficiency-driven heat flow vector.

[0098] Historical temperature sequences of adjacent grid cells are extracted from the temperature distribution feature dataset. The time delay cross-correlation coefficient between the historical temperature sequences of adjacent grid cells is calculated as the time dynamic response similarity, and the time delay parameter corresponding to the maximum time dynamic response similarity is extracted. The time dynamic response similarity is used as the time coupling benchmark value, and the absolute value of the efficiency potential energy difference is used as the spatial driving weight. The product of the time coupling benchmark value and the spatial driving weight is calculated, and the product is positively normalized. The heat transfer coupling strength coefficient between adjacent grid cells is calculated and output.

[0099] Each grid cell is mapped to a node in a graph data structure, the heat transfer direction is mapped to a directed edge connecting the nodes, the heat transfer coupling strength coefficient is mapped to the edge weight of the directed edge, and the time delay parameter is mapped to the time span attribute of the directed edge. Edges with a heat transfer coupling strength coefficient greater than a preset coupling threshold are selected to construct a time series graph adjacency matrix. Based on the time series graph adjacency matrix and the time span attribute, dynamic connectivity paths on the time series are deduced to generate a heat flow map of dynamic heat transfer paths.

[0100] It should be specifically noted that the detailed data processing and deduction process for analyzing the heat transfer direction and coupling relationship between adjacent grid cells to construct the heat flow path map of dynamic heat transfer is as follows:

[0101] The axial position and circumferential phase angle of each grid cell are extracted from the temperature distribution feature dataset, along with the gradient direction of each grid cell. Simultaneously, the waste heat recovery efficiency index of each grid cell is obtained. The axial position and circumferential phase angle constitute the spatial coordinates of the grid cell on the two-dimensional unfolded surface, and the gradient direction is the direction of the physical vector with the largest temperature change rate at that spatial coordinate. Using these spatial coordinates, it is possible to determine whether any two grid cells are adjacent in the data space and calculate the direction of their connection.

[0102] The gradient direction is projected along the line connecting the axial position and circumferential phase angle of adjacent grid cells to obtain the space heat flow vector. This projection operation aims to extract the directional component of pure physical heat flowing from the current grid cell to the adjacent grid cell. For example, if grid A and grid B are adjacent, the gradient direction of A is projected onto the line connecting A and B. If the projected component is positive, it indicates that there is a physical tendency for heat to be transferred from A to B. Next, the difference in waste heat recovery efficiency index between adjacent grid cells is calculated as the efficiency potential energy difference. This difference has a positive or negative sign, reflecting the difference in waste heat recovery value between the two grids. The efficiency potential energy difference is used as a scalar modulation coefficient with a positive or negative sign and multiplied by the space heat flow vector to obtain the efficiency-driven heat flow vector. Because the scalar modulation coefficient has a positive and negative sign, it plays a directional modulation role: if the efficiency potential energy difference is positive, it means that heat flows from the high-efficiency area to the low-efficiency area, and the direction of the efficiency-driven heat flow vector is in the same direction as the physical projection; if the efficiency potential energy difference is negative (for example, physically, the temperature at point A is slightly higher than that at point B, but point B is a high-value recovery area where heat accumulates while point A is a heat dissipation and air leakage area), then the scalar modulation coefficient is negative, which will reverse the direction of the spatial heat flow transfer vector, making the efficiency-driven heat flow vector point to point B. Finally, the direction of heat transfer between adjacent grid cells is determined based on the direction of the efficiency-driven heat flow vector. This process abandons the traditional single determination relying solely on temperature difference and achieves dual-driven direction determination of the physical heat flow vector and the recovery value potential energy.

[0103] Historical temperature sequences of adjacent grid cells are extracted from the temperature distribution feature dataset. The time-delay cross-correlation coefficient between these sequences is calculated as the time-dynamic response similarity, and the time-delay parameter corresponding to the maximum time-dynamic response similarity is extracted. The calculation logic for the time-delay cross-correlation coefficient is as follows: the historical temperature sequences of two grid cells are relatively translated along the time axis, and the Pearson correlation coefficient is calculated at different translation step sizes. When the translation reaches a certain step size, the correlation coefficient reaches its maximum value, which is the time-dynamic response similarity, representing the potential for causal heat transfer between the two grid cells over time. The corresponding translation step size is the time-delay parameter, representing the time period during which heat flows from the upstream grid to the downstream grid. For example, if the temperature peak of grid A occurs at minute 10 and the temperature peak of grid B occurs at minute 15, and the maximum correlation occurs at the 5-minute translation point, then the time-dynamic response similarity is taken as the correlation coefficient value at this point, and the time-delay parameter is 5 minutes. Subsequently, the time-dynamic response similarity is used as the time coupling benchmark, and the absolute value of the efficiency potential difference is used as the spatial driving weight. The product of the temporal coupling baseline value and the spatial driving weight is calculated, and the product is positively normalized (i.e., divided by a preset global maximum possible product value, which can be statistically analyzed based on the product data of all adjacent grid cells in the entire kiln under historical normal operating conditions, taking the empirical maximum value and scaling the result to between 0 and 1). The heat transfer coupling strength coefficient between adjacent grid cells is then calculated and output. This ensures that only grid pairs that have both temporal causal response and sufficient spatial efficiency difference can obtain a high coupling strength coefficient. The preset coupling threshold is based on statistical distribution fitting of multiple sets of grid coupling product data under historical normal operating conditions of the rotary kiln, taking the 75th percentile of the distribution function as the preset coupling threshold (e.g., set to 0.45). When the calculated coupling strength coefficient is greater than 0.45, significant heat coupling is considered to exist between the two.

[0104] Each grid cell is mapped to a node in a graph data structure. The directional heat transfer direction determined in the previous step is mapped to directed edges connecting nodes (e.g., a directed edge from node A to node B). The heat transfer coupling strength coefficient is mapped to the edge weight of the directed edge, and the time delay parameter is mapped to the time span attribute of the directed edge. Edges with a heat transfer coupling strength coefficient greater than a preset coupling threshold are selected to construct a time-series graph adjacency matrix. Since ordinary graph adjacency matrices only use 1 or 0 to express whether nodes are connected, the time-series graph adjacency matrix not only includes node connections and edge weights but also embeds the time span attribute value at the corresponding position in the matrix. Based on the time-series graph adjacency matrix, dynamic connectivity paths are deduced over time using the time span attribute. When deducing the transfer path, it is required that adjacent directed edges are spatially connected end-to-end, and that the time span attributes of the preceding and following edges are sequentially connected on the time axis. For example, there is a directed edge A pointing to B with a time span of 5 minutes; and a directed edge B pointing to C with a time span of 3 minutes. The algorithm verifies the process on a timeline: heat originates from point A at minute 0, arrives at point B at minute 5, then originates again at minute 5 and arrives at point C at minute 8. This monotonically increasing connectivity path deduction ultimately generates a hybrid topology and temporal map, which is the heat flow map representing the dynamic heat transfer path. This map completely solves the problem that traditional static heat flow maps cannot reflect the dynamic timing of heat transfer.

[0105] S4: Based on the waste heat recovery efficiency index, identify abnormal grid cells with efficiency degradation, analyze the influence and transmission relationship of abnormal grid cells on the dynamic heat transfer path according to the heat flow direction map, and locate the upstream disturbed grid cells that cause efficiency degradation.

[0106] In this embodiment, S4 identifies abnormal grid cells with performance degradation based on the waste heat recovery efficiency index, including:

[0107] Based on the waste heat recovery efficiency index of multiple continuous historical time windows, a time series sequence of waste heat recovery efficiency index for each grid cell is constructed. The negative rate of change of the waste heat recovery efficiency index time series sequence within the time window is calculated as the time decay rate. The difference between the current grid cell waste heat recovery efficiency index and the mean of the waste heat recovery efficiency index of grid cells sharing the boundary in a two-dimensional coordinate system formed by axial position and circumferential phase angle is calculated as the spatial dispersion. The time decay rate and spatial dispersion are weighted and fused to obtain the efficiency decay index. The efficiency decay indices of all grid cells are summarized into an efficiency decay index sample set. The mean and standard deviation of the efficiency decay index sample set are calculated. The mean is added to a set multiple of the standard deviation to obtain a dynamic adaptive threshold. Grid cells with an efficiency decay index greater than the dynamic adaptive threshold are identified as abnormal grid cells with efficiency decay.

[0108] The process of analyzing the influence and transmission relationship of abnormal grid cells on the dynamic heat transfer path based on the heat flow pattern, and locating the upstream disturbed grid cells that cause performance degradation, includes:

[0109] Based on the heat flow path map, a set of directed edges pointing to anomalous grid cells is extracted. The edge weights and time span attributes corresponding to the directed edge sets are extracted from the heat flow path map. The edge weights are used as steady-state coupling potential coefficients, and the reciprocal of the time span attribute is used as dynamic transmission rate coefficients. The product of the steady-state coupling potential coefficient and the dynamic transmission rate coefficient is calculated as the influence conductance. The influence conductance is used to characterize the coupling transmission flux of the disturbance signal per unit time. The influence conductance is used as the influence transmission relationship of anomalous grid cells on the dynamic heat transfer path.

[0110] Based on the influence propagation relationship, a reverse traversal is performed along the directed edges pointing to the abnormal grid cells. During the reverse traversal, the abrupt change time points of the performance decay of the waste heat recovery efficiency index time series of each upstream grid cell are extracted, and the actual delay difference between the abrupt change time point and the performance decay time of the abnormal grid cell is calculated. The absolute value of the difference between the actual delay difference and the time span attribute is calculated as the time series delay error. A preset time series allowable error threshold is extracted, and the difference between the preset time series allowable error threshold and the time series delay error is calculated. The ratio of the difference to the preset time series allowable error threshold is used as the time series fit.

[0111] The attenuation source probability is calculated by multiplying the influence of conductance and temporal fit. It is then determined whether there are upstream grid cells whose attenuation source probability is greater than a preset probability threshold. If so, the upstream grid cell with the highest attenuation source probability is selected and identified as the upstream disturbance grid cell that caused the performance attenuation. If the attenuation source probability of all upstream grid cells is less than or equal to the preset probability threshold, the performance attenuation of the abnormal grid cell is determined to be caused by its own internal factors, triggering a local anomaly warning for the grid cell and continuing to monitor the area in real time.

[0112] It needs to be specifically explained that, firstly, the waste heat recovery efficiency index from multiple historical time windows continuously output by S2 (which can be set based on the system's thermal inertia response time and data computation requirements, such as once every 5 minutes in the past 60 minutes) is used to construct a time series. A linear fit is then performed on this time series to obtain the absolute value of the negative slope as the time decay rate (e.g., if the fit slope is -0.08, then the time decay rate is 0.08). Simultaneously, the absolute value of the difference between the current grid efficiency index and the mean values ​​of the efficiency indices of the four adjacent grids sharing the boundary in the axial and circumferential coordinate systems is calculated as the spatial dispersion (e.g., if the current index is 0.50 and the adjacent mean is 0.80, then the dispersion is 0.30). Then... The time decay rate and spatial dispersion are weighted and summed according to the weights (e.g., 0.6 and 0.4) determined by principal component analysis based on historical operating conditions to obtain the efficiency decay index (0.6 × 0.08 + 0.4 × 0.30 = 0.168). The efficiency decay indices of all grids at the current time are aggregated into a sample set, and their mean and standard deviation are calculated (e.g., mean 0.05, standard deviation 0.02). The mean is added to the standard deviation, which is a multiple (taken as 2) set based on the normal distribution Laida criterion, to obtain the dynamic adaptive threshold (0.05 + 2 × 0.02 = 0.09). Grids with an efficiency decay index greater than 0.09 are identified as abnormal grid cells. Then, the index is extracted from the heat flow map constructed in S3. For the directed edge attributes of the anomalous grid, the edge weight (e.g., 0.8) is used as the steady-state coupling potential coefficient, and the reciprocal of the time span attribute (e.g., 4 minutes) is used as the dynamic transmission rate coefficient. Multiplying these two yields the influence transmission rate (0.8 × 1 / 4 = 0.2) to characterize the coupling transmission flux per unit time. Then, based on this influence transmission rate, the upstream grids are traversed backwards along the directed edges to extract the moment when the upstream efficiency index time series undergoes abrupt change (e.g., the 10th minute). The actual delay difference (5 minutes) between this moment and the moment when the anomalous grid decays (e.g., the 15th minute) is calculated. The absolute value of the difference between the actual delay difference and the time span attribute is taken as the time series delay error (|5 - 4| = 1 minute). Take a preset timing error threshold (e.g., 1.5 minutes) set based on the system sampling period and control accuracy requirements. Calculate the ratio of the difference between the preset timing error threshold and the timing delay error to the preset timing error threshold as the timing fit ((1.5-1) / 1.5≈0.33). Finally, multiply the influence conductance by the timing fit to calculate the output attenuation tracing probability (0.2×0.33=0.066). Then, determine whether there are upstream grid cells with an attenuation tracing probability greater than the preset probability threshold (the preset probability threshold can be set based on the system's requirements for tracing accuracy and historical operating data experience, for example, set to 0.05). If so (e.g., 0.066>0.066), the result is taken as follows:If the probability of attenuation tracing is 05, then the upstream grid cell with the highest probability of attenuation is selected and identified as the upstream disturbance grid cell causing the performance degradation. If the probability of attenuation tracing of all upstream grid cells is less than or equal to the preset probability threshold, then it is determined that the performance degradation of the abnormal grid cell is caused by its own internal factors, triggering a local anomaly warning for that grid cell, and continuing real-time monitoring of the area.

[0113] S5: Based on the upstream perturbation grid cell, calculate its thermal perturbation contribution to the abnormal grid cell, and perform spatial decoupling processing on the boundary heat flow of each grid cell based on the thermal perturbation contribution to generate independent heat transfer control parameters for each grid cell.

[0114] In this embodiment, S5 calculates the thermal perturbation contribution of the upstream perturbed mesh cell to the abnormal mesh cell based on the upstream perturbed mesh cell, including:

[0115] Extract the difference in waste heat recovery efficiency index of upstream perturbation grid cells before and after the abrupt change in efficiency decay time point, calculate the ratio of this difference to the waste heat recovery efficiency index before the abrupt change time point as the efficiency deviation rate, and calculate the absolute value of the efficiency deviation rate as the efficiency deviation magnitude; extract the directed transmission path connecting upstream perturbation grid cells and abnormal grid cells from the heat flow path map, calculate the geometric mean of the influence conductivity of each node on the directed transmission path as the path coupling degree, and multiply the efficiency deviation magnitude by the path coupling degree to calculate the thermal perturbation contribution.

[0116] The spatial decoupling of boundary heat flow for each grid cell based on thermal disturbance contribution is performed to generate independent heat transfer control parameters for each grid cell, including:

[0117] The surface temperature values ​​of the current abnormal grid cells within a continuous historical time window are extracted from the temperature distribution feature dataset. The absolute value of the difference between the surface temperature value and the preset ambient temperature within each time window is calculated to obtain the grid temperature difference sequence. The rate of change of the grid temperature difference sequence within the time window is calculated as the temperature difference change rate. The boundary heat flux at the current moment is calculated by multiplying the temperature difference change rate by the waste heat recovery efficiency index.

[0118] Calculate the product of thermal disturbance contribution and boundary heat flux, input the product into a preset nonlinear decay function to calculate the dynamic damping disturbance; calculate the difference between boundary heat flux and dynamic damping disturbance as its own heat transfer fluctuation.

[0119] Extract the preset ideal heat transfer benchmark, calculate the difference between its own heat transfer fluctuation and the preset ideal heat transfer benchmark, input the difference into the preset nonlinear mapping function to calculate the heat flow compensation coefficient, multiply the heat flow compensation coefficient with the preset basic heat transfer adjustment factor to generate the independent heat transfer control parameters of each grid unit.

[0120] It should be specifically explained that, based on the upstream perturbed grid cells, their thermal perturbation contribution to the anomalous grid cells is calculated. Based on this contribution, the boundary heat flux of each grid cell is spatially decoupled to generate independent heat transfer control parameters for each grid cell. The specific process is as follows:

[0121] First, the waste heat recovery efficiency index of the upstream perturbed grid cell before and after the efficiency decay abrupt change point is extracted, and the difference between the two is calculated. This difference is divided by the waste heat recovery efficiency index before the abrupt change point to obtain a ratio, which is taken as the efficiency deviation rate. Since the efficiency of the upstream grid may experience abnormal sudden increases (ratio greater than 0) or sudden decreases (ratio less than 0), in order to uniformly quantify the absolute severity of such fluctuations without directional interference, the absolute value of the efficiency deviation rate is taken as the efficiency deviation amplitude. Subsequently, the directed transmission path connecting the upstream perturbed grid cell and the abnormal grid cell is extracted from the heat flow path map. This path contains multiple nodes, each with an influence conductivity. To accurately reflect the conduction capacity of the entire path and avoid the problem of under-flow caused by multiplication of long paths, this embodiment calculates the geometric mean of the influence conductivity of these nodes as the path coupling degree. In this embodiment, it is assumed that there are 4 nodes on a directed transmission path, with conductivity values ​​of 0.5, 0.4, 0.5, and 0.4, respectively. Multiplying these four numbers directly yields 0.04, a very small value that would result in almost zero perturbation calculated downstream. However, using the geometric mean, the fourth root result is approximately 0.447. This more objectively reflects the overall conduction attenuation characteristics of the entire path, avoiding excessive penalty to coupling due to path length. Finally, multiplying the performance deviation by the path coupling yields the thermal perturbation contribution. A larger value indicates a more severe impact of upstream anomalous fluctuations on the downstream mesh.

[0122] The surface temperature values ​​of the current anomalous grid cells within continuous historical time windows are extracted from the temperature distribution feature dataset. The surface temperature value of each time window is subtracted from the preset ambient temperature, and the absolute value is taken to form a grid temperature difference sequence. Taking the absolute value eliminates the alternation of positive and negative values ​​caused by temperature fluctuations, and uniformly represents the magnitude of the temperature difference. Next, the rate of change of this grid temperature difference sequence within the time window is calculated as the temperature difference change rate, which reflects the dynamic trend of the grid's heat transfer state. At the system data processing level, the temperature difference change rate is multiplied by the current waste heat recovery efficiency index; the product is the boundary heat flux at the current moment. Here, the boundary heat flux is not a strictly dimensional physical quantity, but rather a data characteristic value representing the intensity of the current grid cell's overall heat transfer state, that is, the total fluctuation currently experienced by the grid.

[0123] To separate upstream-transmitted disturbances from the total boundary heat flux of the current grid, this embodiment employs a nonlinear damping decoupling mechanism instead of the traditional simple proportional subtraction. Specifically, the product of the previously obtained thermal disturbance contribution and the current boundary heat flux is first calculated. This product represents the interaction strength between the upstream disturbance and the current total fluctuation. Then, this product is input into a preset nonlinear decay function to calculate the dynamically damped disturbance. Finally, the current boundary heat flux is subtracted from this dynamically damped disturbance, and the difference is the self-heating fluctuation. In this embodiment, it is assumed that the current grid boundary heat flux (total fluctuation) is 2.0, and the upstream-transmitted thermal disturbance contribution is 0.5. The product of the two is 1.0. After inputting 1.0 into the preset nonlinear decay function, the calculated dynamically damped disturbance may be around 0.6. Finally, 2.0 is subtracted from 0.6 to obtain the self-heating fluctuation of 1.4. The advantage of this nonlinear mechanism is that if the boundary heat flux itself is extremely small, the nonlinear function will cause the damping disturbance to approach zero at a faster rate, avoiding the problem of "small total amount minus large disturbance leading to negative distortion" in traditional simple subtraction. This process cannot be mathematically simplified to a simple linear subtraction, and truly restores the actual heat transfer state of the current grid after removing upstream influences.

[0124] Extract the preset ideal heat transfer benchmark set by the system. Calculate the difference between the system's own heat transfer fluctuation and this ideal benchmark. This difference reflects the degree to which the current decoupled real state deviates from the ideal state.

[0125] To ensure smooth control and avoid hardware overshoot, the difference is input into a preset nonlinear mapping function to calculate the heat flux compensation coefficient. The mapping function outputs an appropriate compensation ratio based on the magnitude of the deviation. Finally, the heat flux compensation coefficient is multiplied by the system's underlying preset basic heat transfer adjustment factor to generate an independent heat transfer control parameter specifically for that grid cell. Since this parameter is calculated based on its own decoupled heat transfer fluctuations, it only reflects the actual adjustment needs of that grid cell and will not malfunction due to upstream disturbances.

[0126] S6: Based on independent heat exchange control parameters, with the goal of maximizing the total global waste heat recovery, the actual heat exchange regulation of each grid unit is optimized and allocated, and the waste heat recovery optimization control strategy is output.

[0127] In this embodiment, step S6 optimizes the allocation of actual heat exchange control quantities corresponding to each grid cell with the goal of maximizing the total global waste heat recovery, and outputs a waste heat recovery optimization control strategy, including:

[0128] An initial control demand vector is constructed based on independent heat exchange control parameters. The waste heat recovery efficiency index of each grid cell is extracted. The product of the independent heat exchange control parameters and the waste heat recovery efficiency index is calculated as the marginal recovery gain characteristic value of each grid cell.

[0129] Extract the thermal disturbance contribution of each grid cell, calculate the product of the independent heat transfer control parameter and the thermal disturbance contribution as the potential instability risk, input the potential instability risk into the preset penalty mapping function, and calculate the regulation penalty factor; calculate the difference between the marginal recovery gain characteristic value and the regulation penalty factor as the net recovery benefit evaluation value.

[0130] Based on the net recovery benefit assessment value of each grid unit, its corresponding exponential characteristic value is calculated, and the sum of the exponential characteristic values ​​of all grid units is calculated as the global response sum; the ratio of the exponential characteristic value of each grid unit to the global response sum is used as the actual control allocation weight of each grid unit.

[0131] Extract the preset global total control constraint threshold, multiply the actual control allocation weight by the global total control constraint threshold, and calculate the initial actual heat transfer control amount corresponding to each grid unit.

[0132] An objective function is constructed based on maximizing the total global waste heat recovery as the optimization criterion. The objective function is the global cumulative value of the product of the actual heat exchange control amount of each grid cell and its corresponding waste heat recovery efficiency index. The actual control allocation weights are iteratively corrected based on the objective function. Specifically, after calculating the actual heat exchange control amount in each round, the objective function value of the current round is calculated. If it is higher than the previous round, the current weight is retained. If it is not higher, it is reverted to the weight of the previous round and the sensitivity amplification coefficient of the exponential characteristic value is dynamically adjusted. The weights and actual heat exchange control amounts are recalculated until the objective function value converges.

[0133] The optimal actual heat transfer regulation obtained after convergence is mapped and associated with the corresponding grid cell identifier to generate an optimized control strategy for waste heat recovery.

[0134] It needs to be specifically noted that the initial regulation demand vector is constructed based on the independent heat transfer control parameters of each grid cell. This parameter only represents the independent demand of a single grid after decoupling, without considering the inherent differences in the grid's ability to convert regulation into actual recovered heat. Therefore, the waste heat recovery efficiency index of each grid cell is extracted, and the product of the independent heat transfer control parameter and the waste heat recovery efficiency index is calculated as the marginal recovery gain characteristic value of each grid cell. This characteristic value objectively characterizes the level of waste heat recovery gain that can be directly brought about by investing a unit of regulation into a specific grid. Grids with higher efficiency indices have greater marginal gains under the same demand. In resource-constrained allocation, it is necessary not only to evaluate the benefits but also to guard against the risk of systemic instability. If a grid cell has a large demand but is highly susceptible to upstream disturbances, allocating a large amount of regulation to it can easily trigger the spread of fluctuations. Therefore, the thermal disturbance contribution corresponding to each grid cell is extracted, and the product of the independent heat transfer control parameter and the thermal disturbance contribution is calculated as the potential instability risk. To quantify the degree of inhibition of resource allocation by this risk, the potential instability risk is input into a preset exponential penalty mapping function to calculate the regulatory penalty factor. The specific mathematical relationship of this exponential penalty mapping function is as follows: using the natural constant as the base, the product of the negative penalty intensity coefficient and the current grid's potential instability risk is divided by the maximum potential instability risk among all grids, and this is used as the exponent. Based on the decreasing mathematical property of the exponential function, the grid with a larger potential instability risk has a smaller calculated regulatory penalty factor. Thus, through an explicit nonlinear exponential decay algorithm, the allocation advantage of high-risk grids is automatically and significantly reduced in subsequent allocation competition, overcoming the deficiency of conventional linear penalties in distinguishing risk levels. Considering both gain and risk factors, the difference between the marginal recovery gain characteristic value and the regulatory penalty factor for each grid unit is calculated as the net recovery benefit assessment value. This value represents the actual expected comprehensive recovery benefit of the grid after eliminating instability risk. The corresponding exponential characteristic value is calculated based on the net recovery benefit assessment value of each grid unit. The specific calculation method is as follows: using the natural constant as the base, the product of a preset sensitivity amplification factor and the net recovery benefit assessment value is used as the exponent. By introducing a nonlinear mapping based on the natural exponent, grid cells with higher net benefits can gain an exponentially amplified weight advantage in allocation competition, ensuring that limited resources are absolutely prioritized for high-efficiency and highly resistant grid cells, rather than simply allocated equally according to demand. The exponential eigenvalues ​​of all grid cells are summed to obtain the global response sum. The ratio of the exponential eigenvalue of each grid cell to this global response sum is calculated as the actual control allocation weight of each grid cell. Since the sum of the weights of all grid cells is always one, this constitutes a normalized probability distribution.

[0135] A preset global total control constraint threshold is extracted. The actual control allocation weight of each grid cell is multiplied by this constraint threshold to calculate the actual heat exchange control amount corresponding to each grid cell. Since the sum of all allocation weights is one, the sum of the actual heat exchange control amounts allocated to all grids must be less than or equal to the global total control constraint threshold. The global total control constraint threshold can be set based on the maximum rated output of heat exchange equipment (such as fans or pumps) in the waste heat recovery system and the total amount of resources available under the current operating conditions, thus strictly satisfying the system's total constraint conditions. The objective function of maximizing the total global waste heat recovery is then used as the optimization criterion. This objective function is: multiply the actual heat exchange control amount allocated to all grid cells by their corresponding waste heat recovery efficiency index and then sum them globally to obtain the maximum value of the sum. Based on this objective function, the actual control allocation weights are iteratively corrected: after calculating the actual heat exchange control amount in each round, the objective function value of the current round is calculated, and it is determined whether it is an improvement over the previous round. If the weight allocation improves, it indicates that the allocation scheme is approaching the global optimum, and the weight allocation for this round is retained. If it does not improve, it indicates that the allocation has fallen into a local suboptimal state. In this case, the system reverts to the weight allocation state of the previous round and dynamically adjusts the preset sensitivity amplification coefficient to change the sensitivity of exponential amplification. Then, the exponential eigenvalue, weight, and actual heat transfer regulation are recalculated. This iterative correction process is repeated until the objective function value converges or the preset iteration limit is reached. Finally, after iterative convergence, the unique grid identifier (e.g., two-dimensional coordinate index or grid number) assigned to each grid cell during spatial discretization is extracted. Using the unique grid identifier as the lookup key and the actual heat transfer regulation that has finally converged after iterative correction as the target value, a one-to-one key-value mapping pair is constructed. All grid cells are traversed, and all generated key-value mapping pairs are aggregated and encapsulated into a global control instruction set data structure (e.g., control mapping table or data dictionary). The global control instruction set data structure serves as the output of the waste heat recovery optimization control strategy.

[0136] like Figure 2 The embodiment shown provides an implementation system corresponding to a monitoring and optimization method for the waste heat recovery process of a rotary kiln, including a spatiotemporal discretization module, an efficiency analysis module, a map construction module, a disturbance source tracing module, a spatial decoupling module, and a global optimization module. The spatiotemporal discretization module is connected to the efficiency analysis module, the efficiency analysis module is connected to the map construction module, both the efficiency analysis module and the map construction module are connected to the disturbance source tracing module, the disturbance source tracing module is connected to the spatial decoupling module, and the spatial decoupling module is connected to the global optimization module.

[0137] The spatiotemporal discretization module acquires the temperature data of the rotary kiln shell surface and its corresponding running time and rotation state information. Based on the running time and rotation state information, it performs spatiotemporal positioning of the temperature data and extracts temperature distribution features. Based on the temperature distribution features, the shell surface is discretized into a two-dimensional heat exchange region grid to form a temperature distribution feature dataset for each grid unit.

[0138] The performance analysis module performs waste heat recovery performance analysis on each grid cell based on the temperature distribution characteristic dataset of each grid cell, and obtains the waste heat recovery performance index of each grid cell.

[0139] The heat flow map construction module analyzes the heat transfer direction and heat transfer coupling relationship between adjacent grid cells based on the waste heat recovery efficiency index of each grid cell, and constructs a heat flow map of the dynamic heat transfer path.

[0140] The disturbance tracing module identifies abnormal grid cells with performance degradation based on the waste heat recovery efficiency index, analyzes the influence and transmission relationship of abnormal grid cells on the dynamic heat transfer path according to the heat flow direction map, and locates the upstream disturbed grid cell that causes performance degradation.

[0141] The spatial decoupling module calculates the thermal disturbance contribution of the upstream disturbed grid cell to the abnormal grid cell, and performs spatial decoupling processing on the boundary heat flow of each grid cell based on the thermal disturbance contribution to generate independent heat transfer control parameters for each grid cell.

[0142] The global optimization module, based on independent heat exchange control parameters, optimizes the allocation of actual heat exchange regulation for each grid unit with the goal of maximizing the total global waste heat recovery, and outputs an optimized waste heat recovery control strategy.

[0143] In conclusion, the above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the protection scope of the present invention.

[0144] The above description is merely a specific embodiment of this application, but the scope of protection of this application is not limited thereto. Any variations or substitutions that can be easily conceived by those skilled in the art within the scope of the technology disclosed in this application should be included within the scope of protection of this application. Therefore, the scope of protection of this application should be determined by the scope of the claims.

Claims

1. A method for monitoring and optimizing the waste heat recovery process of a rotary kiln, characterized in that, include: S1: Obtain the temperature data of the rotary kiln shell surface and its corresponding running time and rotation state information. Then, based on the running time and rotation state information, perform spatiotemporal positioning of the temperature data and extract temperature distribution features. Based on the temperature distribution features, discretize the shell surface into a two-dimensional heat exchange area grid to form a temperature distribution feature dataset for each grid unit. S2: Based on the temperature distribution characteristic dataset of each grid cell, perform waste heat recovery efficiency analysis on each grid cell to obtain the waste heat recovery efficiency index of each grid cell; S3: Based on the waste heat recovery efficiency index of each grid cell, analyze the heat transfer direction and heat transfer coupling relationship between adjacent grid cells, and construct a heat flow path map of the dynamic heat transfer path; S4: Based on the waste heat recovery efficiency index, identify abnormal grid cells with efficiency degradation, analyze the influence and transmission relationship of abnormal grid cells on the dynamic heat transfer path according to the heat flow direction map, and locate the upstream disturbed grid cells that cause efficiency degradation. S5: Based on the upstream disturbed grid cell, calculate its thermal disturbance contribution to the abnormal grid cell, and perform spatial decoupling processing on the boundary heat flow of each grid cell based on the thermal disturbance contribution to generate independent heat transfer control parameters for each grid cell. S6: Based on independent heat exchange control parameters, with the goal of maximizing the total global waste heat recovery, the actual heat exchange regulation of each grid unit is optimized and allocated, and the waste heat recovery optimization control strategy is output.

2. The method for monitoring and optimizing the waste heat recovery process of a rotary kiln body according to claim 1, characterized in that, The step S1, acquiring the temperature data of the rotary kiln shell surface and its corresponding running time and rotation status information, includes: Infrared radiation energy data of the cylinder surface is collected along the axial and circumferential directions using a non-contact scanning method that tracks the cylinder rotation synchronously. The temperature data of the cylinder surface is obtained by converting the infrared radiation energy data. The acquisition time corresponding to the temperature data is recorded synchronously, and the current running time information is calculated by combining the kiln start and stop records. The rotational angular velocity and circumferential phase reference signal of the cylinder are acquired synchronously, and the rotational state information is composed of the rotational angular velocity and circumferential phase reference signal; The process involves spatiotemporal localization of temperature data and extraction of temperature distribution features. Based on these features, the cylinder surface is discretized into a two-dimensional heat transfer region grid, forming a temperature distribution feature dataset for each grid cell, including: The rotational angular velocity of the cylinder is extracted based on the rotational state information, and the circumferential phase angle of the cylinder surface is calculated based on the running time and rotational angular velocity. The one-dimensional temperature observation data collected along the time series is mapped and reconstructed into a two-dimensional cylindrical unfolded surface temperature matrix, with the axial position of the cylinder from the kiln head to the kiln tail as the first coordinate and the circumferential phase angle as the second coordinate. Set the axial and circumferential step size of the sliding window, traverse the temperature matrix of the two-dimensional cylindrical unfolded surface using the sliding window, calculate the spatial temperature gradient magnitude and gradient direction for the neighborhood of the matrix covered by the sliding window, extract the temperature data of the center position of the sliding window in the historical time series to calculate the time variation variance, and concatenate the spatial temperature gradient magnitude, gradient direction and time variation variance into vectors to generate temperature distribution features. Set gradient threshold and variance threshold, and mark the region with spatial temperature gradient amplitude greater than the gradient threshold or time variation variance greater than the variance threshold as high heat flux characteristic region, and mark the region with spatial temperature gradient amplitude less than or equal to the gradient threshold and time variation variance less than or equal to the variance threshold as low heat flux characteristic region. For the high heat flux characteristic region, a dense grid is generated using the first grid scale, and for the low heat flux characteristic region, a sparse grid is generated using the second grid scale, which is larger than the first grid scale. The boundary between the dense and sparse grids forms a non-uniform two-dimensional heat transfer region grid. The temperature distribution features corresponding to each grid cell after division are extracted and aggregated to form a temperature distribution feature dataset.

3. The method for monitoring and optimizing the waste heat recovery process of a rotary kiln body according to claim 2, characterized in that, S2 performs waste heat recovery efficiency analysis on each grid cell to obtain the waste heat recovery efficiency index of each grid cell, including: Extract the aggregated temperature distribution features of each grid cell from the temperature distribution feature dataset, and parse the historical temperature sequence, spatial temperature gradient magnitude, spatial temperature gradient direction and time variation variance of the current grid cell based on the temperature distribution features; The cumulative temperature integral value is calculated based on the historical temperature sequence of the current grid cell, and the slope feature of the historical temperature sequence of the current grid cell on the time axis is extracted as the dynamic temperature offset. The cumulative temperature integral value and the dynamic temperature offset are subtracted from the reference offset corresponding to the preset environmental reference temperature. The subtraction results are then weighted and fused to calculate and output the heat exchange potential energy factor of the current grid cell. The time variation coefficient is obtained by calculating the ratio of the time variation variance to the mean of the historical temperature sequence of the current grid cell. The time variation coefficient is then reverse normalized to obtain the time stability coefficient. The heat flow stability penalty factor is determined based on the time stability coefficient. The larger the value of the time variation coefficient, the smaller the value of the heat flow stability penalty factor. The local heat flux is calculated based on the magnitude and direction of the spatial temperature gradient. The local heat flux is then normalized to obtain the spatial diffusion coefficient. The local heat dissipation attenuation factor is determined based on the spatial diffusion coefficient. The larger the value of the local heat flux, the smaller the value of the local heat dissipation attenuation factor. The heat exchange potential energy factor, heat flow stability penalty factor and local heat dissipation attenuation factor are multiplied and weighted and normalized to calculate and output the comprehensive quantitative value of each grid cell as the waste heat recovery efficiency index.

4. The method for monitoring and optimizing the waste heat recovery process of a rotary kiln body according to claim 3, characterized in that, The S3 analysis examines the heat transfer direction and coupling relationship between adjacent grid cells, constructing a heat flow map of dynamic heat transfer paths, including: The axial position and circumferential phase angle of each grid cell are extracted from the temperature distribution feature dataset, and the gradient direction of each grid cell is extracted. At the same time, the waste heat recovery efficiency index of each grid cell is obtained. The space heat flow transfer vector is obtained by projecting the gradient direction along the line connecting the axial position of the adjacent grid cell and the circumferential phase angle; the difference in the waste heat recovery efficiency index between adjacent grid cells is calculated as the efficiency potential energy difference, and the efficiency potential energy difference is multiplied with the space heat flow transfer vector as a scalar modulation coefficient with a positive or negative sign to obtain the efficiency-driven heat flow vector; the direction of heat transfer between adjacent grid cells is determined according to the direction of the efficiency-driven heat flow vector. Historical temperature sequences of adjacent grid cells are extracted from the temperature distribution feature dataset. The time delay cross-correlation coefficient between the historical temperature sequences of adjacent grid cells is calculated as the time dynamic response similarity, and the time delay parameter corresponding to the maximum time dynamic response similarity is extracted. The time dynamic response similarity is used as the time coupling benchmark value, and the absolute value of the efficiency potential energy difference is used as the spatial driving weight. The product of the time coupling benchmark value and the spatial driving weight is calculated, and the product is positively normalized. The heat transfer coupling strength coefficient between adjacent grid cells is calculated and output. Each grid cell is mapped to a node in a graph data structure, the heat transfer direction is mapped to a directed edge connecting the nodes, the heat transfer coupling strength coefficient is mapped to the edge weight of the directed edge, and the time delay parameter is mapped to the time span attribute of the directed edge. Edges with a heat transfer coupling strength coefficient greater than a preset coupling threshold are selected to construct a time series graph adjacency matrix. Based on the time series graph adjacency matrix and the time span attribute, dynamic connectivity paths on the time series are deduced to generate a heat flow map of dynamic heat transfer paths.

5. The method for monitoring and optimizing the waste heat recovery process of a rotary kiln body according to claim 4, characterized in that, S4 identifies abnormal grid cells with performance degradation based on the waste heat recovery efficiency index, including: Based on the waste heat recovery efficiency index of multiple continuous historical time windows, a time series sequence of waste heat recovery efficiency index for each grid cell is constructed. The negative rate of change of the waste heat recovery efficiency index time series sequence within the time window is calculated as the time decay rate. The difference between the current grid cell waste heat recovery efficiency index and the mean of the waste heat recovery efficiency index of grid cells sharing the boundary in a two-dimensional coordinate system formed by axial position and circumferential phase angle is calculated as the spatial dispersion. The time decay rate and spatial dispersion are weighted and fused to obtain the efficiency decay index. The efficiency decay indices of all grid cells are summarized into an efficiency decay index sample set. The mean and standard deviation of the efficiency decay index sample set are calculated. The mean is added to a set multiple of the standard deviation to obtain a dynamic adaptive threshold. Grid cells with an efficiency decay index greater than the dynamic adaptive threshold are identified as abnormal grid cells with efficiency decay. The process of analyzing the influence and transmission relationship of abnormal grid cells on the dynamic heat transfer path based on the heat flow pattern, and locating the upstream disturbed grid cells that cause performance degradation, includes: Based on the heat flow path map, a set of directed edges pointing to anomalous grid cells is extracted. The edge weights and time span attributes corresponding to the directed edge sets are extracted from the heat flow path map. The edge weights are used as steady-state coupling potential coefficients, and the reciprocal of the time span attribute is used as dynamic transmission rate coefficients. The product of the steady-state coupling potential coefficient and the dynamic transmission rate coefficient is calculated as the influence conductance. The influence conductance is used to characterize the coupling transmission flux of the disturbance signal per unit time. The influence conductance is used as the influence transmission relationship of anomalous grid cells on the dynamic heat transfer path. Based on the influence propagation relationship, a reverse traversal is performed along the directed edges pointing to the abnormal grid cells. During the reverse traversal, the abrupt change time points of the performance decay of the waste heat recovery efficiency index time series of each upstream grid cell are extracted, and the actual delay difference between the abrupt change time point and the performance decay time of the abnormal grid cell is calculated. The absolute value of the difference between the actual delay difference and the time span attribute is calculated as the time series delay error. A preset time series allowable error threshold is extracted, and the difference between the preset time series allowable error threshold and the time series delay error is calculated. The ratio of the difference to the preset time series allowable error threshold is used as the time series fit. The attenuation source probability is calculated by multiplying the influence of conductance and temporal fit. It is then determined whether there are upstream grid cells whose attenuation source probability is greater than a preset probability threshold. If so, the upstream grid cell with the highest attenuation source probability is selected and identified as the upstream disturbance grid cell that caused the performance attenuation. If the attenuation source probability of all upstream grid cells is less than or equal to the preset probability threshold, the performance attenuation of the abnormal grid cell is determined to be caused by its own internal factors, triggering a local anomaly warning for the grid cell and continuing to monitor the area in real time.

6. The method for monitoring and optimizing the waste heat recovery process of a rotary kiln body according to claim 5, characterized in that, S5 calculates the contribution of upstream perturbation grid cells to the thermal perturbation of anomalous grid cells, including: Extract the difference in waste heat recovery efficiency index of upstream perturbation grid cells before and after the abrupt change in efficiency decay time point, calculate the ratio of this difference to the waste heat recovery efficiency index before the abrupt change time point as the efficiency deviation rate, and calculate the absolute value of the efficiency deviation rate as the efficiency deviation magnitude; extract the directed transmission path connecting upstream perturbation grid cells and abnormal grid cells from the heat flow path map, calculate the geometric mean of the influence conductivity of each node on the directed transmission path as the path coupling degree, and multiply the efficiency deviation magnitude by the path coupling degree to calculate the thermal perturbation contribution. The spatial decoupling of boundary heat flow for each grid cell based on thermal disturbance contribution is performed to generate independent heat transfer control parameters for each grid cell, including: The surface temperature values ​​of the current abnormal grid cells within a continuous historical time window are extracted from the temperature distribution feature dataset. The absolute value of the difference between the surface temperature value and the preset ambient temperature within each time window is calculated to obtain the grid temperature difference sequence. The rate of change of the grid temperature difference sequence within the time window is calculated as the temperature difference change rate. The boundary heat flux at the current moment is calculated by multiplying the temperature difference change rate by the waste heat recovery efficiency index. Calculate the product of thermal disturbance contribution and boundary heat flux, input the product into a preset nonlinear decay function to calculate the dynamic damping disturbance; calculate the difference between boundary heat flux and dynamic damping disturbance as its own heat transfer fluctuation. Extract the preset ideal heat transfer benchmark, calculate the difference between its own heat transfer fluctuation and the preset ideal heat transfer benchmark, input the difference into the preset nonlinear mapping function to calculate the heat flow compensation coefficient, multiply the heat flow compensation coefficient with the preset basic heat transfer adjustment factor to generate the independent heat transfer control parameters of each grid unit.

7. The method for monitoring and optimizing the waste heat recovery process of a rotary kiln body according to claim 6, characterized in that, S6 optimizes the allocation of actual heat exchange control quantities for each grid cell by maximizing the total global waste heat recovery, and outputs an optimized waste heat recovery control strategy, including: An initial control demand vector is constructed based on independent heat exchange control parameters. The waste heat recovery efficiency index of each grid cell is extracted. The product of the independent heat exchange control parameters and the waste heat recovery efficiency index is calculated as the marginal recovery gain characteristic value of each grid cell. Extract the thermal disturbance contribution of each grid cell, calculate the product of the independent heat transfer control parameter and the thermal disturbance contribution as the potential instability risk, input the potential instability risk into the preset penalty mapping function, and calculate the regulation penalty factor; calculate the difference between the marginal recovery gain characteristic value and the regulation penalty factor as the net recovery benefit evaluation value. Based on the net recovery benefit assessment value of each grid unit, its corresponding exponential characteristic value is calculated, and the sum of the exponential characteristic values ​​of all grid units is calculated as the global response sum; the ratio of the exponential characteristic value of each grid unit to the global response sum is used as the actual control allocation weight of each grid unit. Extract the preset global total control constraint threshold, multiply the actual control allocation weight by the global total control constraint threshold, and calculate the initial actual heat transfer control amount corresponding to each grid unit. An objective function is constructed based on maximizing the total global waste heat recovery as the optimization criterion. The objective function is the global cumulative value of the product of the actual heat exchange control amount of each grid cell and its corresponding waste heat recovery efficiency index. The actual control allocation weights are iteratively corrected based on the objective function. Specifically, after calculating the actual heat exchange control amount in each round, the objective function value of the current round is calculated. If it is higher than the previous round, the current weight is retained. If it is not higher, it is reverted to the weight of the previous round and the sensitivity amplification coefficient of the exponential characteristic value is dynamically adjusted. The weights and actual heat exchange control amounts are recalculated until the objective function value converges. The optimal actual heat transfer regulation obtained after convergence is mapped and associated with the corresponding grid cell identifier to generate an optimized control strategy for waste heat recovery.

Images