An industrial waste heat recovery and distribution optimization method and system based on dynamic thermal load prediction

Abstract

The application provides an industrial waste heat recovery and distribution optimization method and system based on dynamic heat load prediction, relates to the technical field of system distribution optimization, and constructs a heat transfer influence matrix through collection of real-time heat load data and multi-point temperature distribution data of industrial production equipment, and identifies a strong coupling sub-network between heat source equipment based on the matrix, dynamically optimizes a heat recovery path and a flow distribution strategy, can comprehensively consider the heat coupling relationship between the heat source equipment in the industrial production process, improves the overall efficiency of waste heat recovery and distribution, simultaneously, utilizes a comprehensive heat load prediction value and a coupling constraint optimization algorithm, realizes accurate response to dynamic changes of the heat load, significantly improves the flexibility and intelligent level of the system, and thus maximally improves the waste heat recovery and utilization efficiency and reduces energy waste.

Description

Technical Field

[0001] This invention relates to the field of system allocation optimization technology, and more specifically, to a method and system for optimizing industrial waste heat recovery and allocation based on dynamic heat load prediction. Background Technology

[0002] With the continuous advancement of industrialization, the scale of energy consumption in industrial production is expanding daily, making energy efficiency optimization and energy conservation a global focus. In industrial production processes, waste heat is a major manifestation of a large amount of unused energy emitted as low-grade heat. The recovery and utilization of industrial waste heat can not only significantly reduce energy consumption costs but also effectively reduce environmental pollution; therefore, its research and application have gained significant traction in both academia and industry. Currently, industrial waste heat recovery technologies mainly include waste heat boiler recovery, heat pump technology, and combined heat and power (CHP), which have played a role in improving waste heat recovery rates and reducing energy loss. However, traditional waste heat recovery methods typically focus on single equipment or single production processes, lacking dynamic analysis and optimization of the thermal coupling relationships between multiple devices and processes within the entire production system. Especially in complex industrial environments with dynamically changing heat loads, existing technologies struggle to achieve efficient recovery and rational distribution of waste heat.

[0003] The shortcomings of existing industrial waste heat recovery technologies are mainly reflected in the following aspects: First, traditional methods fail to fully consider the coupling relationship between heat source equipment in the industrial production process, leading to local optimization and overall low efficiency in the waste heat distribution process. Second, existing technologies mostly use static models for heat load prediction, which are difficult to adapt to real-time changes in heat load in industrial production, resulting in insufficient responsiveness of the waste heat recovery system. Third, the optimization algorithms for waste heat recovery paths and flow distribution are relatively simplified, usually ignoring the complexity of multi-point temperature distribution and heat transfer, making it difficult to achieve global optimization in dynamic environments. These problems not only limit the further improvement of the efficiency of waste heat recovery systems but also constrain the achievement of efficient waste heat management and energy conservation and emission reduction goals. Summary of the Invention

[0004] To address the problem of low distribution efficiency caused by inaccurate identification of heat coupling relationships among multiple heat sources, this invention provides an industrial waste heat recovery and distribution optimization method and system based on dynamic heat load prediction.

[0005] According to one aspect of the present invention, an optimization method for industrial waste heat recovery and distribution based on dynamic heat load prediction is provided, comprising: Collect real-time heat load data and multi-point temperature distribution data of industrial production equipment, and construct a heat transfer influence matrix; Based on the heat transfer influence matrix, strongly coupled sub-networks between heat source devices are identified to obtain the heat source coupling topology, and the comprehensive heat load prediction value of each coupled sub-network is calculated according to the heat source coupling topology. Based on the heat source coupling topology and the comprehensive heat load prediction, the optimal flow distribution coefficient is calculated using a coupling constraint optimization algorithm. The optimal flow distribution coefficient is output to the waste heat recovery system, and the coordination degree of each recovery path is dynamically adjusted according to the real-time update of the heat transfer influence matrix. When a change in the state of the heat source in the coupled sub-network is detected, the flow distribution strategy of the entire sub-network is re-optimized based on the heat source coupling topology.

[0006] Furthermore, the heat transfer influence matrix is ​​constructed by calculating the characteristic temperature gradient of each heat source device. The Pearson correlation coefficient between them; This was achieved by filling the correlation coefficients between all heat source devices in an n×n dimensional matrix according to the device number order.

[0007] Furthermore, the heat source coupling topology is obtained by converting the heat transfer influence matrix into a weighted network graph and using the Louvain community detection algorithm to identify strongly coupled sub-networks. The Louvain community discovery algorithm identifies strongly coupled subnetworks in the heat source network through a two-stage iterative optimization process. The first stage of the Louvain community discovery algorithm is to select the move operation with the maximum positive gain by calculating the change in modularity of the node movement, and to determine the convergence state of the algorithm by using differential analysis of the modularity change sequence and relative diminishing returns judgment.

[0008] Furthermore, the change in modularity due to the movement of the computing node is calculated as follows: in, To account for the change in modularity of heat transfer characteristics, The weighted connection strength for heat transfer between node i and the target community. Let be the heat transfer delay parameter for node i. The average heat transfer delay of the target community, The heat transfer decay time constant, The total edge weight of the network. Let i be the degree of node i. The total degree of the target community, The thermal coupling strengthening coefficient is... The magnitude of temperature gradient change in the target community. This serves as a reference temperature gradient baseline value.

[0009] Furthermore, the comprehensive heat load forecast is calculated through the following steps: Weight coefficients are assigned according to the node centrality index; A weighted time series forecasting method combined with thermal coupling influence factors is used to perform inter-node mutual correction. The corrected node predictions are weighted and summed to obtain the subnetwork's overall prediction value.

[0010] Furthermore, the optimal flow allocation coefficient is calculated by establishing a multi-objective optimization function with the objectives of minimizing total system energy consumption and minimizing heat load matching error; We set up a system of coupled constraints and used the Lagrange multiplier method to introduce Lagrange multipliers to construct an unconstrained optimization problem. The optimal flow allocation coefficient is obtained by solving the KKT optimality condition equations and iteratively solving them using numerical optimization algorithms such as sequential quadratic programming or interior point method. The coupling constraint system includes balance constraints, flow coordination constraints between heat sources, and boundary constraints.

[0011] Furthermore, the multi-objective optimization function is shown in the following equation: in, To optimize the function value for multiple objectives, The weighting factor for the total system energy consumption term. This represents the total number of heat source devices. heat source Energy consumption coefficient per unit flow rate heat source Traffic allocation, heat source The flow allocation coefficient, coefficient heat source With heat source The coupling connection weights between them For reference energy consumption value, This is the weighting factor for the heat load matching error term. heat source Efficiency parameters, heat source The reference heat capacity, heat source Target heat output, This is a reference thermal power value.

[0012] According to another aspect of the present invention, an industrial waste heat recovery and distribution optimization system based on dynamic heat load prediction is provided, comprising: The data acquisition and matrix construction module is used to collect heat load data and multi-point temperature distribution data of industrial production equipment in real time, and construct a heat transfer influence matrix that reflects the mutual influence between heat sources based on the collected data. The topology identification module is used to identify strongly coupled sub-network relationships between heat source devices based on the heat transfer influence matrix and to construct a heat source coupled topology structure. The load prediction module is used to calculate the comprehensive heat load prediction value of each coupled sub-network based on the heat source coupling topology. The flow optimization calculation module is used to solve the optimal flow distribution coefficient considering the mutual influence between heat sources based on the coupled topology of the heat source and the comprehensive heat load prediction value, using a coupled constraint optimization algorithm. The flow allocation execution module is used to output the optimal flow allocation coefficient to the waste heat recovery system and execute the corresponding control operations; The coordination opening degree adjustment module is used to dynamically adjust the coordination opening degree of each recovery path based on the real-time update of the heat transfer influence matrix. The adaptive re-optimization module is used to detect changes in the state of heat sources within the coupled subnetwork and re-optimize the traffic allocation strategy of the entire subnetwork based on the coupled topology of the heat sources.

[0013] Compared with existing technologies, the industrial waste heat recovery and distribution optimization method and system based on dynamic heat load prediction provided by this invention constructs a heat transfer influence matrix by collecting real-time heat load data and multi-point temperature distribution data of industrial production equipment. Based on this matrix, it identifies strongly coupled sub-networks between heat source equipment and dynamically optimizes heat recovery paths and flow distribution strategies. This comprehensively considers the thermal coupling relationships between heat source equipment in industrial production, improving the overall efficiency of waste heat recovery and distribution. Simultaneously, by utilizing comprehensive heat load prediction values ​​and coupling constraint optimization algorithms, it achieves accurate response to dynamic changes in heat load, significantly improving the system's flexibility and intelligence, thereby maximizing waste heat recovery efficiency and reducing energy waste. Attached Figure Description

[0014] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort. In the drawings: Figure 1 This is a flowchart of an industrial waste heat recovery and distribution optimization method based on dynamic heat load prediction according to an embodiment of the present invention.

[0015] Figure 2This diagram compares the present invention with traditional methods in the industrial waste heat recovery and distribution optimization method based on dynamic heat load prediction according to an embodiment of the present invention. Detailed Implementation

[0016] Hereinafter, exemplary embodiments according to the present invention will be described in detail with reference to the accompanying drawings. Obviously, the described embodiments are merely some embodiments of the present invention, and not all embodiments of the present invention. It should be understood that the present invention is not limited to the exemplary embodiments described herein.

[0017] Figure 1 This is a flowchart of an industrial waste heat recovery and distribution optimization method based on dynamic heat load prediction according to an embodiment of the present invention. Figure 1 As shown, the optimization method for industrial waste heat recovery and distribution based on dynamic heat load prediction includes: S1: Collect real-time heat load data H(t) and multi-point temperature distribution data T_multi(t) of industrial production equipment. Construct a heat transfer influence matrix I_matrix by calculating the temperature gradient correlation coefficient between each heat source point. The heat transfer influence matrix I_matrix quantifies the heat coupling strength and transfer delay characteristics between each heat source device.

[0018] A high-precision temperature sensor array deployed at key locations on the industrial production line collects real-time heat load data H(t) from each heat source device. This heat load data H(t) includes key parameters such as device surface temperature, heat flux density, and heat output power per unit time. Simultaneously, multiple temperature monitoring nodes are set around each heat source device according to a predetermined spatial grid layout to acquire temperature distribution data in three-dimensional space. The multi-point temperature distribution data It covers information on the temperature of the heat source equipment itself, the near-field temperature distribution, and the temperature transition zone between it and adjacent equipment.

[0019] Furthermore, by setting up multiple temperature monitoring nodes around each heat source device, the temperature gradient of each monitoring point is first calculated using the finite difference method. Specifically, the temperature difference between adjacent monitoring points is divided by the spatial distance to obtain the local temperature gradient. Then, the characteristic temperature gradient of the heat source is obtained by vector averaging the temperature gradients of all monitoring points of the same heat source device. Next, for any two heat source devices i and j, calculate their characteristic temperature gradient. and The time series correlation coefficient is calculated using the Pearson correlation coefficient formula. It quantifies the correlation strength by comparing the ratio of the covariance to the standard deviation of two gradient series within a given time window. A positive correlation coefficient indicates that the temperature changes of the two heat sources follow the same trend, suggesting a positive heat transfer relationship. A negative correlation coefficient indicates a negative influence relationship. The absolute value of the correlation coefficient directly reflects the tightness of the heat coupling between the two sources. Based on this, the temperature gradient correlation coefficients calculated for each pair of heat source devices are filled into an n×n dimensional matrix according to the device number, forming a heat transfer influence matrix. , where matrix elements This is the temperature gradient correlation coefficient between heat source i and heat source j.

[0020] Preferably, key heat source nodes can be identified through the heat transfer influence matrix, and the matrix can be calculated. The degree centrality of each heat source node is calculated by summing the absolute values ​​of all elements in the i-th row to obtain the out-degree intensity of node i, and summing the absolute values ​​of all elements in the i-th column to obtain the in-degree intensity of node i. A node with a high out-degree intensity indicates that the heat source has a strong influence on other heat sources, while a node with a high in-degree intensity indicates that the heat source is easily influenced by other heat sources. Next, the eigenvector centrality of each node is calculated by applying the heat transfer influence matrix. Eigenvalue decomposition is performed to obtain the eigenvector corresponding to the largest eigenvalue. The node corresponding to the element with the largest value in this eigenvector is the most influential critical heat source in the system. When it is necessary to identify critical bridging nodes in the heat transfer path, the betweenness centrality calculation method is used to analyze the heat transfer influence matrix. The shortest paths between nodes in the constructed graph structure are calculated, and the frequency of each node's occurrence in all shortest paths is counted. The node with the highest frequency is the key heat transfer bridge point. To identify the core nodes in the heat source cluster, the influence matrix on heat transfer is then considered. Cluster analysis is performed to group heat source nodes with correlation coefficients greater than a preset threshold into the same cluster. Then, the local centrality of nodes within each cluster is calculated. The node with the highest local centrality is the core heat source node of the cluster. These core nodes are usually the key control points that control the heat distribution efficiency of the entire cluster.

[0021] S2: Based on the heat transfer influence matrix I_matrix, a graph theory clustering algorithm is used to identify strongly coupled sub-networks between heat source devices, obtain the heat source coupling topology T_topology, and calculate the comprehensive heat load prediction value H_cluster(t+Δt) of each coupled sub-network according to the heat source coupling topology T_topology.

[0022] First, the matrix Convert to an undirected weighted graph, where each heat source device is a graph node and matrix elements. As the edge weights connecting nodes i and j, if the absolute value of a matrix element is lower than a set lower limit for coupling strength, no edge connection is established between nodes, thus forming a sparse heat source coupling network graph. Next, the Louvain community detection algorithm based on maximizing modularity is used to cluster this network graph. The algorithm first initializes each node as an independent community, then iteratively calculates the modularity gain after moving each node to the community of its neighboring nodes. When the modularity gain is positive, the node movement operation is performed until all node movements can no longer further increase the modularity. Based on this, the algorithm uses each community obtained in the first stage as a new super node, reconstructs the connection relationships between communities, and repeats the above clustering process until the network's modularity no longer changes significantly. When the clustering algorithm converges, the set of nodes with high internal connection density and low external connection density is identified as a strongly coupled subnetwork, where there are close heat transfer relationships between heat source devices within each subnetwork. Finally, a heat source coupling topology is constructed based on the clustering results. The topology records the internal connection patterns of each strongly coupled subnetwork, the bridging relationships between subnetworks, and the coupling strength weight corresponding to each connection edge in the form of a graph. If a heat source node is connected to multiple subnetworks at the same time, it is marked as a key bridging node, thus forming a hierarchical topology that fully describes the heat source coupling relationship of the system.

[0023] First, each heat source node in the network is initialized as an independent community. Then, each node i is traversed in a random order, and the change in modularity ΔQ after moving node i from its current community to the communities of all its neighboring nodes j is calculated. When the calculated modularity gain ΔQ is greater than zero, a node move operation is performed, adding node i to the neighboring community that produces the maximum positive gain. If all neighboring communities produce negative gains, node i remains in its original community. After completing one round of traversal of all nodes, a continuity analysis of the modularity growth trend is used. By recording the modularity change sequence of the most recent several iterations, the first-order difference and second-order difference of the sequence are calculated to determine the trend characteristics of modularity growth. When the first-order difference remains positive for several consecutive rounds and the value is relatively stable, it indicates that the modularity is still growing. When the first-order difference starts to show negative values ​​or the second-order difference is continuously negative, it indicates that the rate of modularity growth is slowing down or has begun to decline. Meanwhile, the relative diminishing returns of modularity growth are used, which means calculating the ratio of the modularity growth in the current round to the average growth in the previous rounds. When the ratio shows a downward trend for several consecutive rounds and approaches zero, the algorithm is considered to be close to convergence, and the first stage of node movement optimization process ends.

[0024] Next, in the second stage, the algorithm shrinks the communities identified in the first stage into single supernodes. The edge weights within a community are transformed into the self-loop weights of the supernode, while the edge weights between communities are transformed into the connection weights between supernodes, forming a new compressed network graph. The node movement process of the first stage is then repeated on this compressed network graph until the network modularity can no longer be improved. The final community partitioning result is the strongly coupled subnetwork structure of the original network, in which heat source nodes within the same community have close heat transfer coupling relationships.

[0025] The specific process for calculating the change in modularity after moving node i from its current community to the community of its neighbor node j is as follows: First, the sum of edge weights between node i and other nodes in its current community is recorded as the intra-community connection weight. Simultaneously, the total degree of the current community, i.e., the sum of the degrees of all nodes within that community, is recorded. Then, the reduction in modularity caused by node i moving out of the current community is calculated. This reduction is equal to the intra-community connection weight divided by twice the total network edge weight, minus the square of the product of the current community's total degree and node i's degree divided by twice the total network edge weight. Next, the sum of edge weights between node i and all nodes in the target community (i.e., the community of neighbor node j) is recorded as the target community connection weight. Simultaneously, the total degree of the target community is obtained. Then, the increase in modularity brought about by node i joining the target community is calculated. This increase is equal to the target community connection weight divided by twice the total network edge weight, minus the square of the product of the target community's total degree and node i's degree divided by twice the total network edge weight. When moving a node, the dynamic change in community degree needs to be considered. Specifically, the total degree of the original community needs to be reduced by the degree of node i, while the total degree of the target community needs to be increased by the degree of node i. The final change in modularity equals the modularity gain from joining the target community minus the modularity loss from leaving the original community. If the result is positive, it indicates that the move operation is beneficial to improving the overall modularity of the network. The algorithm will select the move with the largest gain from all possible moves. If none of the possible moves produce a positive gain, the node remains in its current community. As shown in the following formula: in, To account for the change in modularity of heat transfer characteristics, The weighted connection strength for heat transfer between node i and the target community. Let be the heat transfer delay parameter for node i. The average heat transfer delay of the target community, The heat transfer decay time constant, The total edge weight of the network. Let i be the degree of node i. The total degree of the target community, The thermal coupling strengthening coefficient is... The magnitude of temperature gradient change in the target community. This serves as a reference temperature gradient baseline value.

[0026] On the other hand, according to the heat source coupling topology Calculate the predicted combined heat load of each coupled sub-network Based on topology The boundaries and internal connections of each strongly coupled subnetwork are identified, and then weight coefficients are assigned to the heat source nodes within each subnetwork according to their centrality index in the topology. Next, a topology-based weighted time series forecasting method is used to analyze the historical heat load data of each heat source node within the subnetwork. The subnetworks are formed by weighting and combining the data according to their respective weighting coefficients. Simultaneously, based on the heat transfer intensity reflected by the edge weights in the topology, the heat coupling influence factor between nodes is calculated. When predicting the future time t+Δt, the individual heat load of each node is first initially predicted using an autoregressive moving average model. Then, based on the connection relationships and transfer delay parameters in the topology, the mutual correction amount between the predicted values ​​of each node is calculated; that is, the predicted heat load of a node will be adjusted according to the predicted changes of its neighboring nodes and the connection strength. Finally, the predicted values ​​of each node after coupling correction are weighted and summed according to their weight coefficients in the subnetwork to obtain the comprehensive heat load prediction value of the entire coupled subnetwork. As shown in the following formula: in, This represents the predicted combined heat load of the coupled subnetwork. The total number of heat source nodes within the subnetwork. For nodes Based on the weighting coefficients of the centrality index, For nodes Individual predicted heat load values, For nodes The set of adjacent nodes, For nodes To the node Thermal coupling influencing factors, For nodes With nodes Edge weights between them Adjacent nodes Individual predicted heat load values.

[0027] In this process, heat source nodes within each subnetwork are assigned weight coefficients based on their centrality indices in the topology. First, the degree centrality of each node *i* within the subnetwork is calculated, which is determined by counting the number of edges or the sum of edge weights directly connecting that node to other nodes within the subnetwork. Next, the betweenness centrality of each node is calculated by finding all shortest paths between any two nodes within the subnetwork and counting the frequency of each node appearing as an intermediate node on these shortest paths. Then, the eigenvector centrality is calculated by performing eigenvalue decomposition on the adjacency matrix of the subnetwork to obtain the eigenvector corresponding to the largest eigenvalue. After calculating the three centrality indices, a weighted fusion method is used to combine them into a comprehensive centrality index. Specifically, the degree centrality, betweenness centrality, and eigenvector centrality are standardized to ensure uniform numerical ranges, and then linearly combined according to preset weight ratios. Typically, the degree centrality weight is set to 0.4, the betweenness centrality weight to 0.3, and the eigenvector centrality weight to 0.3. Finally, the comprehensive centrality index of each node is normalized so that the sum of the weight coefficients of all nodes in the same sub-network is equal to 1, thus obtaining the final weight coefficient of each heat source node in the comprehensive heat load prediction.

[0028] The heat coupling influence factor is calculated by extracting the direct connection strength between nodes based on the edge weight information in the topology. This connection strength reflects the heat transfer capacity between adjacent nodes. Then, a spatial attenuation coefficient is calculated by combining the physical or topological distance between nodes; nodes with greater distances have less mutual influence. Next, the time delay relationship of load changes at each node in historical heat load data is analyzed. The time delay parameter of heat transfer between nodes is determined by calculating the cross-correlation function. When the heat load of node i changes, it takes a certain amount of time to affect the heat load of node j. Then, the correlation coefficient of heat load changes between nodes is calculated to measure the synchronization degree of heat load fluctuations between two nodes; a higher correlation coefficient indicates a stronger heat coupling relationship between the two nodes. Finally, the edge weights, spatial attenuation coefficient, time delay correction factor, and correlation coefficient are normalized and multiplied to obtain the heat coupling influence factor.

[0029] It is worth noting that although the above description outlines the general process for obtaining the combined heat load prediction values ​​for each coupled subnetwork, specific implementation details may vary depending on the medical institution and its equipment configuration. For example: Taking the centralized heating system of a large industrial park as an example, this study illustrates the following: The park contains multiple heat source devices, including gas-fired boilers, combined heat and power (CHP) units, biomass boilers, solar collectors, and industrial waste heat recovery devices. These devices are interconnected through a complex pipeline system and exhibit varying degrees of heat transfer influence. When constructing the heat transfer influence matrix, it was found that the main boiler and CHP units exhibit a strong coupling relationship due to sharing main transmission pipelines and being physically adjacent. The solar collectors and biomass boilers form a moderately strong coupling relationship through auxiliary pipelines. However, some peripheral small electric heating devices, due to their greater distance and independent operation, have coupling strengths below a set lower threshold. When constructing the network graph using graph theory, high-weighted connections are established between strongly coupled devices, medium-weighted connections are established between moderately coupled devices, and no connections are established between weakly coupled devices, thus forming a sparse network topology.

[0030] When using the Louvain clustering algorithm for analysis, each heat source device initially forms an independent community. The algorithm then calculates the modularity gain of various possible node relocation schemes. When it is found that moving the main boiler to the community where the cogeneration unit is located produces a significant positive gain, the relocation operation is executed. Similarly, solar collectors and biomass boilers are grouped into the same community. After multiple rounds of iterative optimization, when the modularity growth trend changes from rapid increase to gradual increase and eventually stabilizes, the algorithm identifies several strongly coupled sub-networks. These mainly include a high-temperature heat source sub-network centered on the main boiler and cogeneration unit, a renewable energy sub-network mainly composed of solar and biomass boilers, and independently operating auxiliary heating sub-networks. In the comprehensive heat load prediction, different weight coefficients are assigned to devices within each sub-network according to their topological importance. For example, in the high-temperature heat source sub-network, the main boiler is given a higher weight due to its high connectivity and crucial role in the heat transfer path, while the auxiliary devices have relatively lower weights. Finally, the comprehensive heat load prediction result for the entire sub-network is obtained by weighted combination of the predicted values ​​of each device and correction considering the coupling effects between devices.

[0031] S3: Based on the aforementioned heat source coupling topology and comprehensive heat load forecast The optimal flow distribution coefficient considering the mutual influence between heat sources is calculated using a coupled constraint optimization algorithm. The coupling constraint optimization algorithm uses the heat transfer influence matrix I_matrix as a constraint to avoid repeated recycling and allocation conflicts.

[0032] Based on the heat source coupling topology The coupling relationships and influence strengths among the heat sources are identified, and the flow distribution coefficients of each heat source are used as optimization variables to construct a multi-objective optimization function with the goals of minimizing total system energy consumption and heat load matching error. Next, a system of coupling constraints is established, including the balance constraint that the sum of flow within each sub-network must equal the corresponding comprehensive heat load prediction value; the flow coordination constraint between heat sources determined by the connection weights in the topology; and the boundary constraints of the physical capacity limitations of each heat source device. The flow coordination constraint describes the cascading effects of a change in the flow of a heat source on the flow distribution to its adjacent heat sources. After the constrained optimization model is constructed, the constrained optimization problem is transformed into an unconstrained problem using the Lagrange multiplier method. By introducing Lagrange multipliers to correspond to various constraints, a Lagrange function containing the original objective function and constraint penalty terms is constructed. Then, numerical optimization methods such as gradient descent or Newton's method are used for iterative solution. In each iteration, the gradient information of the Lagrange function with respect to the flow distribution coefficient and the Lagrange multiplier is calculated, and the variable values ​​are updated according to the gradient direction. At the same time, the degree of satisfaction of the constraints is monitored and the values ​​of the Lagrange multipliers are adjusted accordingly. When the gradient norm is less than the convergence tolerance and the degree of violation of all constraints is within an acceptable range, the algorithm converges to obtain the optimal solution, which is the optimal flow distribution coefficient considering the mutual influence between heat sources. .

[0033] The multi-objective optimization function is established with the objective of minimizing total energy consumption. This objective includes the direct operating energy consumption of each heat source device and the additional energy consumption caused by the coupling effect between heat sources. The direct operating energy consumption is expressed as a quadratic function of the flow distribution coefficient of each heat source, i.e., the energy consumption of each heat source i is ci multiplied by . The square of, where Let be the energy consumption coefficient per unit flow rate of heat source i, while the additional energy consumption caused by the coupling effect between heat sources depends on the topology. In the calculation of connection weights, when there is a strong coupling connection between two heat sources i and j, the mismatch in their flow distribution will produce additional heat transfer losses, which are represented by the coupling weights. Multiply by the difference in flow distribution minus The absolute value of the heat load matching error is then determined. Next, a target term for minimizing the heat load matching error is established. This term measures the deviation between the actual heat load provided by each heat source and its target heat load. This is achieved by calculating the actual heat output Hi of each heat source and its target heat output. The actual heat output Hi is represented by the sum of squares of the differences between the two values, and is determined by the flow distribution coefficient. and heat source efficiency parameters Jointly determined, that is, Hi equals Multiply Multiply by the reference heat capacity After the two single-objective functions are constructed, a weighted method is used to combine them into a unified multi-objective optimization function. Specifically, the total energy consumption term of the system is multiplied by a weighting factor. Multiply the heat load matching error term by a weighting factor The weighting factor is determined based on the priority requirements of system operation; if energy saving is a greater priority, then... Take the larger value; if more emphasis is placed on precise control of heat load, then... Take the larger value. The final multi-objective optimization function expression is the weighted sum of all direct energy consumption terms from all heat sources, the weighted sum of all coupled additional energy consumption terms, and the weighted sum of all heat load matching error terms. That is, the multi-objective optimization function is as follows: This also includes constraints: in, To optimize the function value for multiple objectives, The weighting factor for the total system energy consumption term. This represents the total number of heat source devices. heat source Energy consumption coefficient per unit flow rate heat source The flow distribution coefficient, heat source The flow distribution coefficient, heat source With heat source The coupling connection weights between them For reference energy consumption value, This is the weighting factor for the heat load matching error term. heat source Efficiency parameters, heat source The reference heat capacity, heat source Target heat output, For reference thermal power value, For the first The set of heat source nodes within a subnetwork For the first The combined heat load forecast for each sub-network, The total number of sub-networks, For flow coordination tolerance parameters, heat source The set of adjacent nodes, heat source Minimum flow capacity limit, heat source Maximum flow capacity limit.

[0034] Furthermore, identify all equality and inequality constraints in the original constrained optimization problem, and express the equality constraints as follows: The inequality constraint is expressed in the form of zero. The form is less than or equal to zero, followed by each equality constraint. Introducing the corresponding Lagrange multiplier equal to zero For each inequality constraint Introducing the corresponding Lagrange multiplier for values ​​less than or equal to zero ,in The non-negativity condition must be met, i.e. The physical meaning of these Lagrange multipliers, which are greater than or equal to zero, is the degree of marginal influence of the constraints on the objective function. Then, the Lagrange function L is constructed, which is equal to the original objective function f(α) plus all the equality constraint terms. Multiply Summation plus all inequality constraint terms Multiply The summation of the sums ... As shown in the following formula: in, For the heat source coupling multi-objective optimization function, For the first Lagrange multipliers with equality constraints, For the first An equality constraint function, Let be the total number of inequality constraints. For the first Lagrange multipliers constrained by inequality, For the first Inequality constraint functions.

[0035] After constructing the Lagrangian function, a system of equations is established based on the necessary condition for optimality, namely the KKT conditions. This system of equations includes the partial derivative of the Lagrangian function with respect to the decision variable α being equal to zero, and all equality constraints. Equal to zero, all inequality constraints Lagrange multipliers less than or equal to zero and corresponding to all inequality constraints Greater than or equal to zero, and complementary relaxation conditions, i.e. Multiply The value equals zero. This occurs if the inequality constraint has no effect at the optimal solution. If the value is less than zero, then the corresponding Lagrange multiplier μj is equal to zero, if the inequality constraint takes effect at the optimal solution. If the value is zero, then the corresponding Lagrange multiplier is... The values ​​may be greater than zero. By solving this KKT equation system, the optimal decision variable α and Lagrange multiplier values ​​can be obtained. In the numerical solution process, iterative algorithms such as sequential quadratic programming or interior-point methods are used to continuously update the values ​​of the decision variables and Lagrange multipliers, gradually satisfying the KKT conditions. When the degree of violation of all KKT conditions is less than the preset tolerance, the algorithm converges to obtain the optimal flow allocation coefficient α_coupled, as shown in the following equation: in, This is the optimal flow allocation coefficient vector. For the heat source coupling multi-objective optimization function, This represents the total number of equality constraints. For the first Lagrange multipliers with equality constraints, For the first An equality constraint function, Let be the total number of inequality constraints. For the first Lagrange multipliers constrained by inequality, For the first Inequality constraint functions.

[0036] S4: The optimal flow allocation coefficient The heat is output to the waste heat recovery system and determined according to the heat transfer influence matrix. Real-time updates dynamically adjust the coordination and activation level of each recycling path. When a change in the state of a heat source is detected within the coupled subnetwork, based on the heat source coupling topology... The traffic allocation strategy for the entire subnetwork was re-optimized.

[0037] Optimal flow allocation coefficient The complete process of outputting to the waste heat recovery system and achieving dynamic coordinated control comprises three core stages: first, the system-level deployment and execution of the allocation coefficients; second, real-time coordinated adjustment based on the heat transfer influence matrix; and finally, an adaptive re-optimization mechanism based on state changes. In the allocation coefficient deployment stage, the system will calculate the optimal flow allocation coefficients... The data is transmitted via a digital control interface to various execution units of the waste heat recovery system, including key equipment such as heat exchanger groups, circulating pump stations, valve controllers, and pipeline regulating devices. The flow distribution coefficient corresponding to each heat source is converted into specific operating parameters for the corresponding equipment, such as the speed adjustment signal of the circulating pump, the opening control command of the regulating valve, and the heat exchange area configuration command of the heat exchanger. This ensures that the theoretical optimization results can be accurately translated into actual physical control actions. During the dynamic adjustment of the opening degree, the system continuously monitors the heat transfer influence matrix. The numerical changes of each element in the matrix reflect the intensity of the mutual influence and time delay characteristics of heat transfer between different recovery paths. When some key elements in the matrix change significantly, it indicates that the coupling relationship or transfer efficiency between heat sources has been dynamically adjusted. In this case, it is necessary to adjust the coordination degree of each recovery path accordingly. .

[0038] Furthermore, when the system detects the influence matrix medium elements When a change occurs, first calculate the rate of change of that element relative to the previous time step. If the absolute value of the rate of change exceeds a preset sensitivity threshold, the coordination opening adjustment mechanism of the corresponding recovery path j is triggered. During the adjustment process, a proportional-integral adjustment strategy is adopted, setting the adjustment amount of the coordination opening as a function of the rate of change, i.e. Equal to the proportionality coefficient Multiply by the current rate of change Add integral coefficient Multiplying by the cumulative sum of historical change rates, the proportional term is used to respond to current changes in influence intensity, and the integral term is used to eliminate long-term steady-state biases. When multiple elements in the influence matrix change simultaneously, the weighted average of the changes in all relevant matrix elements corresponding to each recovery path is calculated. The weights are determined by the importance of each heat source in the topology. Then, the coordinated opening adjustment amount for each recovery path is calculated based on the weighted average change rate. To ensure system stability, the coordinated opening adjustment adopts a gradual update method, i.e., the new coordinated opening... Equal to current value Add adjustment amount Multiply by smoothing factor The smoothing factor ranges from 0 to 1; a smaller value results in a smoother adjustment. Simultaneously, upper and lower limits for the coordination opening degree are set to ensure it remains within a reasonable operating range. The preset sensitivity threshold is set based on the calculation of each element of the heat transfer influence matrix using historical system operating data. Standard deviation under steady state Then, the sensitivity threshold was set to 2 to 3 times the standard deviation.

[0039] When the system detects changes in the state of heat sources within the coupled subnetwork, including the start-up and shutdown of heat source devices, significant changes in load levels, degradation of device efficiency, or the addition or deletion of heat source devices, it triggers a global re-optimization mechanism based on the heat source coupled topology T_topology: The updated topology undergoes a state change assessment. By comparing the current topology with the previous one, the type and scope of change are identified. If only node parameter changes, such as adjustments to heat source power or efficiency, are involved, a local optimization strategy is employed to recalculate flow allocation only for the affected nodes and their direct neighbors. If topology changes, such as node additions, deletions, or alterations to connectivity, a global re-optimization process is initiated. During global re-optimization, the system re-executes a community detection algorithm based on maximizing modularity to partition the updated topology into subnetworks, identifying new strongly coupled subnetwork boundaries. Then, based on the new topology and current heat load forecast requirements, a new coupling constraint optimization model is reconstructed, and the aforementioned Lagrange multiplier method is used to solve for the new optimal flow allocation coefficients. .

[0040] It's worth noting that the process of re-optimizing the flow allocation strategy of the entire sub-network based on the coupled topology of heat sources can be illustrated by the following example: Suppose that the waste heat recovery system of an industrial park originally contained three main heat sources: a boiler room, a chemical workshop, and a refrigeration station. These formed a relatively stable coupled sub-network through a pipe network, and the system had already calculated the corresponding flow allocation strategies for these three heat sources. When a data center is added to the park as a fourth heat source, the first step is to assess the impact of this change on the original topology. Because the data center has a strong coupling relationship with the refrigeration station, it can directly utilize the waste heat of the data center for cooling. At the same time, the data center can also provide low-quality heat energy to the boiler room. This new connection relationship changes the original three-node network structure. The system then re-executes the community detection algorithm and finds that the addition of the data center has made the previously relatively independent refrigeration station more closely connected with other heat sources. It may be necessary to re-divide the sub-network boundaries or treat the entire system as a larger strongly coupled sub-network. Next, the system needs to rebuild the constraint optimization model. Originally, it only needed to consider the flow balance and coupling effects between the boiler room, chemical workshop, and refrigeration station. Now, it must also include the heat load demand of the data center, the direct heat exchange constraints between the data center and the refrigeration station, and the coordination constraints of the data center providing auxiliary heat sources to the boiler room. When solving for the new optimal flow allocation, the system will find that the flow allocation coefficient of the cooling station may increase significantly because it can simultaneously utilize the waste heat from the data center and the chemical plant, while the flow allocation coefficient of the boiler room may decrease accordingly because the data center provides additional heat source supplementation. The entire re-optimization process ensures that the four heat sources can achieve coordinated operation and maximize overall efficiency under the new topology.

[0041] Preferred, such as Figure 2 As shown in the figure, this invention compares the performance of the present invention with that of traditional methods across multiple dimensions. The comparison results show that the method of the present invention significantly outperforms the traditional method in six key performance indicators: heat recovery efficiency, response speed, system stability, energy consumption reduction rate, coordination optimization degree, and adaptability. Among these, the improvement in coordination optimization degree and response speed is the most significant, while heat recovery efficiency and system stability have also been greatly improved. This indicates that the optimization control method based on the heat source coupling topology not only improves on single performance indicators but also achieves comprehensive performance enhancements in multiple dimensions such as overall system coordination, dynamic response capability, and long-term stable operation. This proves that the technical solution can effectively solve the shortcomings of traditional methods in multi-heat source coordinated control and has significant comprehensive advantages and practical value.

[0042] In summary, the industrial waste heat recovery and distribution optimization method based on dynamic heat load prediction, as described in this invention, is explained. It constructs a heat transfer influence matrix by collecting real-time heat load data and multi-point temperature distribution data from industrial production equipment. Based on this matrix, it identifies strongly coupled sub-networks among heat source equipment and dynamically optimizes heat recovery paths and flow distribution strategies. This comprehensively considers the thermal coupling relationships between heat source equipment during industrial production, improving the overall efficiency of waste heat recovery and distribution. Simultaneously, by utilizing comprehensive heat load prediction values ​​and coupling constraint optimization algorithms, it achieves accurate response to dynamic changes in heat load, significantly improving the system's flexibility and intelligence, thereby maximizing waste heat recovery efficiency and reducing energy waste.

[0043] Here, those skilled in the art will understand that the specific operations of each step in the above-described industrial waste heat recovery and distribution optimization system based on dynamic heat load prediction have been referenced above. Figure 1 and Figure 2 The method for optimizing industrial waste heat recovery and distribution based on dynamic heat load prediction has been described in detail, and therefore, its repeated description will be omitted.

[0044] According to another aspect of the present invention, an industrial waste heat recovery and distribution optimization system based on dynamic heat load prediction is provided, comprising: The data acquisition and matrix construction module is used to collect heat load data and multi-point temperature distribution data of industrial production equipment in real time, and construct a heat transfer influence matrix that reflects the mutual influence between heat sources based on the collected data. The topology identification module is used to identify strongly coupled sub-network relationships between heat source devices based on the heat transfer influence matrix and to construct a heat source coupled topology structure. The load prediction module is used to calculate the comprehensive heat load prediction value of each coupled sub-network based on the heat source coupling topology. The flow optimization calculation module is used to solve the optimal flow distribution coefficient considering the mutual influence between heat sources based on the coupled topology of the heat source and the comprehensive heat load prediction value, using a coupled constraint optimization algorithm. The flow allocation execution module is used to output the optimal flow allocation coefficient to the waste heat recovery system and execute the corresponding control operations; The coordination opening degree adjustment module is used to dynamically adjust the coordination opening degree of each recovery path based on the real-time update of the heat transfer influence matrix. The adaptive re-optimization module is used to detect changes in the state of heat sources within the coupled subnetwork and re-optimize the traffic allocation strategy of the entire subnetwork based on the coupled topology of the heat sources.

[0045] In summary, the industrial waste heat recovery and distribution optimization system based on dynamic heat load prediction, as described in this invention, is explained. It constructs a heat transfer influence matrix by collecting real-time heat load data and multi-point temperature distribution data from industrial production equipment. Based on this matrix, it identifies strongly coupled sub-networks among heat source equipment and dynamically optimizes heat recovery paths and flow distribution strategies. This comprehensively considers the thermal coupling relationships between heat source equipment during industrial production, improving the overall efficiency of waste heat recovery and distribution. Simultaneously, by utilizing comprehensive heat load prediction values ​​and coupling constraint optimization algorithms, it achieves accurate responses to dynamic changes in heat load, significantly enhancing the system's flexibility and intelligence, thereby maximizing waste heat recovery efficiency and reducing energy waste.

Claims

1. A method for optimizing industrial waste heat recovery and distribution based on dynamic heat load prediction, characterized in that, include: Collect real-time heat load data and multi-point temperature distribution data of industrial production equipment, and construct a heat transfer influence matrix; Based on the heat transfer influence matrix, strongly coupled sub-networks between heat source devices are identified to obtain the heat source coupling topology, and the comprehensive heat load prediction value of each coupled sub-network is calculated according to the heat source coupling topology. Based on the heat source coupling topology and the comprehensive heat load prediction, the optimal flow distribution coefficient is calculated using a coupling constraint optimization algorithm. The optimal flow distribution coefficient is output to the waste heat recovery system, and the coordination degree of each recovery path is dynamically adjusted according to the real-time update of the heat transfer influence matrix. When a change in the state of the heat source in the coupled sub-network is detected, the flow distribution strategy of the entire sub-network is re-optimized based on the heat source coupling topology.

2. The industrial waste heat recovery and distribution optimization method based on dynamic heat load prediction according to claim 1, characterized in that, The heat transfer influence matrix is ​​constructed by calculating the characteristic temperature gradient of each heat source device. The Pearson correlation coefficient between them; This was achieved by filling the correlation coefficients between all heat source devices in an n×n dimensional matrix according to the device number order.

3. The industrial waste heat recovery and distribution optimization method based on dynamic heat load prediction according to claim 2, characterized in that, The heat source coupling topology is obtained by converting the heat transfer influence matrix into a weighted network graph and using the Louvain community detection algorithm to identify strongly coupled sub-networks. The Louvain community discovery algorithm identifies strongly coupled subnetworks in the heat source network through a two-stage iterative optimization process. The first stage of the Louvain community discovery algorithm selects the move operation with the maximum positive gain by calculating the change in modularity of the node movement, and uses differential analysis of the modularity change sequence and relative diminishing returns to determine the convergence state of the algorithm.

4. The industrial waste heat recovery and distribution optimization method based on dynamic heat load prediction according to claim 3, characterized in that, The change in modularity during the movement of the computing node is calculated as follows: in, To account for the change in modularity of heat transfer characteristics, The weighted connection strength for heat transfer between node i and the target community. Let be the heat transfer delay parameter for node i. The average heat transfer delay of the target community, The heat transfer decay time constant, The total edge weight of the network. Let be the degree of node i. The total degree of the target community, The thermal coupling strengthening coefficient is... The magnitude of temperature gradient change in the target community. This serves as a reference temperature gradient baseline value.

5. The industrial waste heat recovery and distribution optimization method based on dynamic heat load prediction according to claim 4, characterized in that, The comprehensive heat load forecast is calculated through the following steps: Weight coefficients are assigned according to the node centrality index; A weighted time series forecasting method combined with thermal coupling influence factors is used to perform inter-node mutual correction. The corrected node predictions are weighted and summed to obtain the subnetwork's overall prediction value.

6. The industrial waste heat recovery and distribution optimization method based on dynamic heat load prediction according to claim 5, characterized in that, The optimal flow allocation coefficient is calculated by establishing a multi-objective optimization function with the objectives of minimizing total system energy consumption and minimizing heat load matching error; We set up a system of coupled constraints and used the Lagrange multiplier method to introduce Lagrange multipliers to construct an unconstrained optimization problem. The optimal flow allocation coefficient is obtained by solving the KKT optimality condition equations and iteratively solving them using numerical optimization algorithms such as sequential quadratic programming or interior point method. The coupling constraint system includes balance constraints, flow coordination constraints between heat sources, and boundary constraints.

7. The industrial waste heat recovery and distribution optimization method based on dynamic heat load prediction according to claim 6, characterized in that, The multi-objective optimization function is shown in the following equation: in, To optimize the function value for multiple objectives, The weighting factor for the total system energy consumption term. This represents the total number of heat source devices. heat source Energy consumption coefficient per unit flow rate heat source The flow distribution coefficient, heat source The flow distribution coefficient, heat source With heat source The coupling connection weights between them For reference energy consumption value, This is the weighting factor for the heat load matching error term. heat source Efficiency parameters, heat source The reference heat capacity, heat source Target heat output, This is a reference thermal power value.

8. An industrial waste heat recovery and distribution optimization system based on dynamic heat load prediction, characterized in that, include: The data acquisition and matrix construction module is used to collect heat load data and multi-point temperature distribution data of industrial production equipment in real time, and construct a heat transfer influence matrix that reflects the mutual influence between heat sources based on the collected data. The topology identification module is used to identify strongly coupled sub-network relationships between heat source devices based on the heat transfer influence matrix and to construct a heat source coupled topology structure. The load prediction module is used to calculate the comprehensive heat load prediction value of each coupled sub-network based on the heat source coupling topology. The flow optimization calculation module is used to solve the optimal flow distribution coefficient considering the mutual influence between heat sources based on the coupled topology of the heat source and the comprehensive heat load prediction value, using a coupled constraint optimization algorithm. The flow allocation execution module is used to output the optimal flow allocation coefficient to the waste heat recovery system and execute the corresponding control operations; The coordination opening degree adjustment module is used to dynamically adjust the coordination opening degree of each recovery path based on the real-time update of the heat transfer influence matrix. The adaptive re-optimization module is used to detect changes in the state of heat sources within the coupled subnetwork and re-optimize the traffic allocation strategy of the entire subnetwork based on the coupled topology of the heat sources.

9. A computer device comprising a memory and a processor, wherein the memory stores a computer program, characterized in that, When the processor executes the computer program, it implements the steps of the industrial waste heat recovery and distribution optimization method based on dynamic heat load prediction as described in any one of claims 1 to 7.

10. A computer-readable storage medium having a computer program stored thereon, characterized in that, When the computer program is executed by the processor, it implements the steps of the industrial waste heat recovery and distribution optimization method based on dynamic heat load prediction as described in any one of claims 1 to 7.

Images