Heat supply system optimization scheduling control method based on operation benchmark library and working condition matching

Abstract

The invention discloses a heat supply system optimization scheduling control method based on an operation benchmark library and working condition matching. The method comprises the following steps: constructing an offline operation benchmark library comprising a scheduling control business demand type, a business working condition type, business working condition relevance and a business operation strategy; business demand similarity and business working condition type similarity are calculated through the scheduling control agent; if the business demand similarity and the business working condition type similarity both exceed threshold values, business operation strategies under the business demand and the business working condition are obtained through an offline operation benchmark library, and operation strategies of other types of scheduling control businesses are adaptively adjusted according to business working condition relevance; taking the offline operation benchmark library as a training sample set through a scheduling control agent, carrying out training learning of each business scheduling control model, constructing each business scheduling control model, and obtaining a corresponding business operation strategy; and fusing and analyzing the benchmark strategy and the model strategy to obtain a final operation strategy of each scheduling control service.

Description

Technical Field

[0001] This invention belongs to the field of heating system scheduling and control technology, specifically relating to an optimized scheduling and control method for heating systems based on an operational benchmark library and operating condition matching. Background Technology

[0002] During actual operation, the heat source, heating station, and heat users of the heating system generate a large amount of operational data. By monitoring and analyzing the data, the operation strategy of the heating system can be optimized, and the energy-saving, environmental protection, economy, and ability to meet heat load requirements during operation can be improved.

[0003] The scheduling and control of heating systems involves the output scheduling of source-side units, the regulation of flow and temperature at heating stations, and the pump and valve control of heat user buildings. Furthermore, there are business-related couplings between the source side, heating stations, and heat users, further complicating the scheduling and control of heating systems. Therefore, the scheduling and control of heating systems are correlated with business types and operating conditions. Different businesses and operating conditions have corresponding operating strategies. How to consider the correlations between scheduling and control businesses and system operating conditions, reduce the real-time computation load during scheduling and control, and quickly, accurately, and flexibly obtain business operation strategies is a pressing issue that needs to be addressed.

[0004] Based on the above technical problems, it is necessary to design a new method for optimizing scheduling and control of heating systems based on a benchmark library and operating condition matching. Summary of the Invention

[0005] The technical problem this invention aims to solve is to overcome the shortcomings of existing technologies and provide an optimized scheduling and control method for heating systems based on an offline benchmark library and operating condition matching. This method utilizes an offline benchmark library to reduce real-time computation, allows for rapid matching of business requirements and operating conditions, and enables the intelligent agent model to quickly map real-time data to operating strategies. Furthermore, the steady-state operating condition database and the offline benchmark library ensure stable operation under normal business requirements and operating conditions, while model training and strategy fusion can address new operating conditions. By balancing the reliability of benchmark strategies with the flexibility of model strategies, the risk of relying on a single decision path is reduced.

[0006] To solve the above-mentioned technical problems, the technical solution of the present invention is as follows: This invention provides an optimized scheduling and control method for heating systems based on a benchmark database and operating condition matching, comprising: Step S1: Obtain historical operating data and object attribute data of the heating system source, grid and load sides, perform time alignment, data preprocessing and steady-state condition filtering, and construct a steady-state condition database for each side of the source, grid and load. Step S2: Obtain the dispatch control business requirements and historical operation strategy data of each dispatch control business on the source, network and load sides of the heating system. Combine the steady-state operating condition database of each source, network and load side to perform understanding and analysis of each dispatch control business requirement and operation strategy, define the characteristic parameters of the business conditions, cluster and classify the business conditions, and analyze the correlation of the business conditions. The dispatch control business requirements include the operation output dispatch control of each heating unit on the source side, the secondary network flow and temperature regulation of each heating station on the network side, and the pump and valve regulation of each building on the load side. Step S3: Based on the simulation model of the heating system, after conducting simulation analysis and verification optimization of multi-objective parameter operation strategies under different operating conditions of various scheduling and control services, construct an offline operation benchmark library that includes scheduling and control service demand types, service operating condition types, service operating condition correlations, and service operation strategies. Step S4: The pre-trained heating system dispatch and control agent obtains business condition characteristic parameters and business requirement data based on the real-time operation data of the source, grid and load sides and the dispatch and control business requirements, and calculates the similarity of dispatch and control business requirements and business condition types with the offline operation benchmark library respectively. If the similarity of business requirements and the similarity of business conditions both exceed the threshold, the match is successful. The business operation strategy under the business requirement and business condition is obtained by running the benchmark library offline, and the operation strategy of other types of scheduling and control services is adaptively adjusted according to the correlation of business conditions. Step S5: The pre-trained heating system scheduling and control agent uses the offline benchmark library as the training sample set to train and learn the scheduling and control models of each business. After constructing the scheduling and control models of each business, it inputs the real-time operation data of the source network and load side and the scheduling and control business requirements to obtain the corresponding business operation strategies. Step S6: Integrate and analyze the business operation strategies in steps S4 and S5 to obtain the final operation strategies for each scheduling and control service.

[0007] Furthermore, step S1 includes: Obtain the heat source output, supply and return water temperature, flow rate, fuel consumption, unit operating status and heat source type, installed capacity, unit equipment efficiency curve, and unit output constraints on the source side of the heating system; Acquire data on pump frequency, valve opening, heat exchange efficiency, network heat metering, network structure parameters, heating station coverage area, pump operating characteristics, and valve regulation characteristics of the heating system network-side heating stations. Obtain the indoor temperature of the building on the load side of the heating system, the flow rate at the end of the secondary network, the supply and return water temperatures, the building type, the building area, and the building heat user load; The acquired data timestamps are converted into a unified format, and the running data is collected uniformly according to the sampling frequency. A unique identifier is added to each time series data, and data association is established through topological relationships. After handling missing values ​​and outliers, key parameters on the source, grid, and load sides are selected as the basis for steady-state determination. Combined with steady-state period filtering, a steady-state operating condition database for each side of the source, grid, and load is constructed. The database structure includes a steady-state operating condition master table, a source-side steady-state data table, a grid-side steady-state data table, a load-side steady-state data table, and an object attribute table.

[0008] Furthermore, the selection of key parameters on the source, grid, and load sides as the basis for steady-state determination includes: heat consumption fluctuations, water supply temperature fluctuations, and unit load rate fluctuations on the source side; temperature difference fluctuations at the inlet and outlet of the heat station, water pump frequency fluctuations, and secondary network flow fluctuations on the load side; building water supply temperature fluctuations and building heat load fluctuations on the load side; and outdoor temperature fluctuations among environmental parameters. The steady-state period selection method includes: using the sliding window method to calculate the fluctuation values ​​of each key parameter within the window; if the fluctuation of all parameters is less than the threshold, the window is determined to be a steady-state period; merging adjacent steady-state windows into a longer steady-state period; and considering the correlation between the steady-state conditions of the source, grid, and load sides, verifying whether each side meets the steady-state conditions in the same period.

[0009] Furthermore, in step S2, the operational requirements and strategies of each scheduling and control operation are analyzed, including: analyzing the operational output scheduling and control requirements of each heating unit on the source side to meet the total heat load demand of the grid side while taking into account multiple objectives such as energy conservation, environmental protection, and economic benefits; and recording the output commands and start-up / shutdown plans of each unit on the source side under steady-state conditions; analyzing the secondary network flow and temperature regulation requirements of each heating station on the grid side to match the secondary network flow and temperature with the heat demand on the load side and eliminate hydraulic imbalance; and recording the secondary network water supply temperature curve, flow regulation commands, and valve opening changes of the heating station under steady-state conditions; analyzing the pump and valve regulation requirements of each building on the load side to meet the indoor temperature of building heat users and respond to grid side load regulation commands; and recording the pump and valve regulation parameters at the building entrance and the valve regulation parameters in front of the heat users; and as well as associating the operational strategy data with the steady-state condition database through timestamps and spatial topology to form a condition-strategy mapping combination for source-grid-load coordinated operation. The definition of service condition characteristic parameters includes: source-side service condition characteristics, grid-side service condition characteristics, and load-side service condition characteristics; the source-side service condition characteristics include outdoor temperature, total grid-side heat load demand, and peak-valley ratio; the grid-side service condition characteristics include outdoor temperature, load demand of heat users under the jurisdiction of the heating station, and flow-temperature difference combination; the load-side service condition characteristics include outdoor temperature, heat load demand of each building, indoor temperature, and building type; Service condition clustering includes: based on the source-side service condition characteristics, network-side service condition characteristics and load-side service condition characteristics, clustering algorithms are used to group similar features in the steady-state service condition database into one category, and service condition clustering is performed on the source-side, network-side and load-side. The business condition correlation analysis includes: calculating the correlation coefficients between source-side business condition characteristics and network-side business condition characteristics, and between network-side business condition characteristics and load-side business condition characteristics; obtaining pairs of business condition characteristic parameters that are strongly correlated among various business condition characteristics; and setting thresholds for changes in each business condition characteristic parameter by combining the business condition characteristic parameters and business condition clustering results. The analysis then selects and simulates slight and significant changes in a certain business condition characteristic parameter, determining whether the business condition characteristic parameters of related services exceed their thresholds. If they do, the business condition type of the related services changes, and the operating strategy changes synchronously; otherwise, the original operating strategy is maintained. Finally, a mapping table of changes in business condition characteristic parameters and changes in related business condition types is formed.

[0010] Furthermore, step S4 includes: Set the prompt word template for the scheduling control agent, including: defining the role of the agent as generating business operation strategies through reasoning based on real-time operation data and business requirements of the source network and load side; and setting the agent's thought chain task steps and intermediate conclusions for each step. The steps for setting up the intelligent agent's thought chain task include: uploading real-time operating data and scheduling control business requirements from the source, network, and load sides; extracting business condition feature parameters and business requirement data; similarity calculation and matching task rules; and outputting business operation strategies. The pre-trained intelligent agent for scheduling and control of the heating system uses set prompt word templates and thought chain task steps to guide the large model to transform real-time operating data of the source, grid and load sides and scheduling and control business requirements into understandable text instructions. After extracting business condition feature parameters and business requirement data, it calls the similarity calculation module to calculate the similarity between real-time scheduling and control business requirements and scheduling and control business requirement types in the offline benchmark library, and the similarity between real-time operating condition feature parameters and business condition types in the offline benchmark library. If both the similarity of business requirements and the similarity of business operating conditions exceed the threshold, the match is successful. The business operation strategy under the business requirement and business operating condition is obtained through the offline benchmark library. Based on the business operating condition correlation analysis results, the mapping table of changes in business operating condition characteristic parameters and changes in associated business operating condition types is used to determine whether the business operating condition affects the operating condition type of other types of scheduling and control services. If it does, the corresponding business operation strategy is searched from the offline benchmark library according to the changed operating condition type. If the operating condition type cannot be matched, the operating strategy is obtained after simulation analysis and verification optimization based on the heating system simulation model.

[0011] Furthermore, the calculation of the similarity between real-time scheduling and control service requirements and the scheduling and control service requirements of the offline benchmark library includes: extracting semantic features of real-time scheduling and control service requirements, assigning different weights to different semantic features using a weighted cosine similarity algorithm, and calculating the feature similarity with the scheduling and control service requirements of the offline benchmark library. The calculation of the similarity between real-time operating condition feature parameters and business operating condition types in the offline benchmark library includes: extracting operating condition features from real-time operating condition feature parameters, assigning different weights to different operating condition features using a weighted cosine similarity algorithm, and calculating the similarity between the operating condition features and the business operating condition types in the offline benchmark library, including the semantic similarity and numerical similarity of each operating condition feature. The different weights of different semantic features and different weights of different working condition features are calculated using a multi-dimensional algorithm to determine the optimal weight values ​​of different semantic features and different working condition features.

[0012] Furthermore, step S5 includes: The pre-trained intelligent agent for scheduling and control of the heating system uses the offline benchmark library as the training sample set. It takes the relevant features of scheduling and control business demand type, business condition type, and business condition correlation as input features and the business operation strategy as output features. After training and learning using machine algorithms, it constructs the business scheduling and control model for each side of the source, network, and load. After inputting the real-time operation data and scheduling control service requirements of the source, grid and load sides into the service scheduling control model of each side of the source, grid and load, the corresponding service operation strategy is output. Among them, after constructing the service scheduling and control models for each side of the source, network, and load, a total loss function is constructed, including the loss of the single-side service scheduling and control model and the cross-side coordination loss, and the model is optimized.

[0013] Furthermore, step S6 includes: The intelligent agent for scheduling and control of the heating system regards the process of integrating and analyzing business operation strategies as a decision-making task, and learns the optimal business operation strategy by interacting with the operating environment of the heating system. The state is defined as real-time running data, scheduling and control business requirements, strategy calculation and analysis related parameters in step S4, and strategy calculation and analysis related parameters in step S5. The action is defined as the business operation strategy output in step S4, the business operation strategy output in step S5, and the business operation strategy of the weighted fusion of steps S4 and S5; the business operation strategy of the weighted fusion of steps S4 and S5 adopts the analytic hierarchy process when calculating the strategy weight. The reward is defined as the heating operation effect after the implementation of the business operation strategy, including energy-saving and environmental protection effect, economic benefit effect, and load target achievement effect; The agent explores the action space through a greedy strategy, selects the strategy corresponding to the action to execute, and evaluates the expected reward of the action. After each action is executed, the current state, action, reward and the state data of the next moment are stored in the experience replay pool, and the next round of iteration begins until the reward is the highest. The corresponding action is then used as the final operation strategy for each scheduling and control service.

[0014] Furthermore, step S4 also includes: If the similarity of business requirements and the similarity of business operating conditions are not matched, the corresponding business operation strategy will be obtained after simulation analysis and verification optimization based on the heating system simulation model, and added to the offline operation benchmark library.

[0015] The beneficial effects of this invention are: (1) This invention obtains historical operating data and object attribute data of the source, grid and load sides of the heating system, performs time alignment, data preprocessing and steady-state condition screening, and constructs a steady-state condition database for each side of the source, grid and load. This can eliminate data noise and errors caused by time misalignment, ensure data reliability, and perform steady-state condition screening, providing high-quality samples for subsequent construction of offline operation benchmark library and condition matching. (2) By acquiring the scheduling and control business requirements and historical operation strategy data of each scheduling and control business on the source, grid and load sides of the heating system, and combining the steady-state operating condition database of each source, grid and load side, this invention can perform analysis and understanding of scheduling and control business requirements and operation strategies, define business condition characteristic parameters, cluster and classify business conditions and analyze the correlation of business conditions. This invention can quantify the complex operating status of the heating system, reduce decision complexity, and mine the correlation of business conditions, laying the foundation for multi-business collaborative scheduling, and making the operation strategy match the specific operating conditions, thus avoiding coarse control. (3) This invention constructs an offline benchmark library that includes scheduling and control service demand types, service condition types, service condition correlations, and service operation strategies after conducting simulation analysis and verification optimization of multi-objective parameter operation strategies under different operating conditions of various scheduling and control services based on the simulation model of the heating system. The benchmark strategies based on simulation verification are scientific and reliable, covering multiple service demands and different operating conditions on the source, grid, and load sides, and reducing real-time computing pressure through offline construction. (4) The present invention calculates the similarity between real-time business requirements, operating conditions and benchmark databases of the pre-trained heating system scheduling and control agent, matches the optimal strategy, and adjusts the collaborative business operation strategy according to the correlation of business operating conditions. This enables the business operation strategy to be highly adapted to the real-time state and avoids system imbalance caused by single business optimization. (5) This invention trains the scheduling control model of each business by using the offline benchmark library as training samples, realizes the direct mapping from real-time input to policy output, improves the level of scheduling automation, and the model has generalization ability and can handle new working conditions not covered by the benchmark library. (6) This invention can take into account the reliability of benchmark strategies and the flexibility of model strategies through the integrated analysis of business operation strategies, thereby reducing the risk of a single decision path.

[0016] Other features and advantages will be set forth in the description which follows, and will be apparent in part from the description, or may be learned by practicing the invention. The objects and other advantages of the invention are realized and obtained through the structures particularly pointed out in the description and the drawings.

[0017] To make the above-mentioned objects, features and advantages of the present invention more apparent and understandable, preferred embodiments are described below in detail with reference to the accompanying drawings. Attached Figure Description

[0018] To more clearly illustrate the specific embodiments of the present invention or the technical solutions in the prior art, the drawings used in the description of the specific 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 from these drawings without creative effort.

[0019] Figure 1 This is a flowchart of an optimized scheduling and control method for a heating system based on a benchmark library and operating condition matching, according to the present invention. Figure 2 This is a flowchart of the method for outputting business operation strategies based on an offline benchmark library according to the present invention; Figure 3 This is a flowchart of the fusion analysis method for business operation strategies of the present invention. Detailed Implementation

[0020] To make the objectives, technical solutions, and advantages of the embodiments of the present invention clearer, the technical solutions of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0021] like Figure 1 As shown, this embodiment 1 provides an optimized scheduling and control method for a heating system based on a benchmark library and operating condition matching, which includes: Step S1: Obtain historical operating data and object attribute data of the heating system source, grid and load sides, perform time alignment, data preprocessing and steady-state condition filtering, and construct a steady-state condition database for each side of the source, grid and load. Step S2: Obtain the dispatch control business requirements and historical operation strategy data of each dispatch control business on the source, network and load sides of the heating system. Combine the steady-state operating condition database of each source, network and load side to perform understanding and analysis of each dispatch control business requirement and operation strategy, define the characteristic parameters of the business conditions, cluster and classify the business conditions, and analyze the correlation of the business conditions. The dispatch control business requirements include the operation output dispatch control of each heating unit on the source side, the secondary network flow and temperature regulation of each heating station on the network side, and the pump and valve regulation of each building on the load side. Step S3: Based on the simulation model of the heating system, after conducting simulation analysis and verification optimization of multi-objective parameter operation strategies under different operating conditions of various scheduling and control services, construct an offline operation benchmark library that includes scheduling and control service demand types, service operating condition types, service operating condition correlations, and service operation strategies. Step S4: The pre-trained heating system dispatch and control agent obtains business condition characteristic parameters and business requirement data based on the real-time operation data of the source, grid and load sides and the dispatch and control business requirements, and calculates the similarity of dispatch and control business requirements and business condition types with the offline operation benchmark library respectively. If the similarity of business requirements and the similarity of business conditions both exceed the threshold, the match is successful. The business operation strategy under the business requirement and business condition is obtained by running the benchmark library offline, and the operation strategy of other types of scheduling and control services is adaptively adjusted according to the correlation of business conditions. Step S5: The pre-trained heating system scheduling and control agent uses the offline benchmark library as the training sample set to train and learn the scheduling and control models of each business. After constructing the scheduling and control models of each business, it inputs the real-time operation data of the source network and load side and the scheduling and control business requirements to obtain the corresponding business operation strategies. Step S6: Integrate and analyze the business operation strategies in steps S4 and S5 to obtain the final operation strategies for each scheduling and control service.

[0022] In this embodiment, step S1 includes: Obtain the heat source output, supply and return water temperature, flow rate, fuel consumption, unit operating status and heat source type, installed capacity, unit equipment efficiency curve, and unit output constraints on the source side of the heating system; Acquire data on pump frequency, valve opening, heat exchange efficiency, network heat metering, network structure parameters, heating station coverage area, pump operating characteristics, and valve regulation characteristics of the heating system network-side heating stations. Obtain the indoor temperature of the building on the load side of the heating system, the flow rate at the end of the secondary network, the supply and return water temperatures, the building type, the building area, and the building heat user load; The acquired data timestamps are converted into a unified format, and the running data is collected uniformly according to the sampling frequency. A unique identifier is added to each time series data, and data association is established through topological relationships. After handling missing values ​​and outliers, key parameters on the source, grid, and load sides are selected as the basis for steady-state determination. Combined with steady-state period filtering, a steady-state operating condition database for each side of the source, grid, and load is constructed. The database structure includes a steady-state operating condition master table, a source-side steady-state data table, a grid-side steady-state data table, a load-side steady-state data table, and an object attribute table.

[0023] In this embodiment, the selection of key parameters on the source-grid-load side as the basis for steady-state determination includes: heat consumption fluctuation, water supply temperature fluctuation, and unit load rate fluctuation on the source side; temperature difference fluctuation, pump frequency fluctuation, and secondary network flow fluctuation on the load side; building water supply temperature fluctuation and building heat load fluctuation on the load side; and outdoor temperature fluctuation among environmental parameters. The steady-state period selection method includes: using the sliding window method to calculate the fluctuation values ​​of each key parameter within the window; if the fluctuation of all parameters is less than the threshold, the window is determined to be a steady-state period; merging adjacent steady-state windows into a longer steady-state period; and considering the correlation between the steady-state conditions of the source, grid, and load sides, verifying whether each side meets the steady-state conditions in the same period.

[0024] In this embodiment, step S2 involves understanding and analyzing the scheduling and control business requirements and operational strategies of each unit. This includes: analyzing the operational output scheduling and control business requirements of each heating unit on the source side to meet the total heat load demand of the grid side while considering multiple objectives such as energy conservation, environmental protection, and economic benefits; simultaneously recording the output commands and start-up / shutdown plans of each unit on the source side under steady-state operating conditions; analyzing the secondary network flow and temperature regulation business requirements of each heating station on the grid side to match the secondary network flow and temperature with the heat demand on the load side and eliminate hydraulic imbalance; simultaneously recording the secondary network water supply temperature curve, flow regulation commands, and valve opening changes of the heating station under steady-state operating conditions on the grid side; analyzing the pump and valve regulation business requirements of each building on the load side to meet the indoor temperature of building heat users and respond to grid side load regulation commands; simultaneously recording the pump and valve regulation parameters at the building entrance and the valve regulation parameters in front of the heat users; and as well as associating the operational strategy data with the steady-state operating condition database through timestamps and spatial topology to form a working condition-strategy mapping combination for source-grid-load coordinated operation. The definition of service condition characteristic parameters includes: source-side service condition characteristics, grid-side service condition characteristics, and load-side service condition characteristics; the source-side service condition characteristics include outdoor temperature, total grid-side heat load demand, and peak-valley ratio; the grid-side service condition characteristics include outdoor temperature, load demand of heat users under the jurisdiction of the heating station, and flow-temperature difference combination; the load-side service condition characteristics include outdoor temperature, heat load demand of each building, indoor temperature, and building type; Service condition clustering includes: based on the source-side service condition characteristics, network-side service condition characteristics and load-side service condition characteristics, clustering algorithms are used to group similar features in the steady-state service condition database into one category, and service condition clustering is performed on the source-side, network-side and load-side. The business condition correlation analysis includes: calculating the correlation coefficients between source-side business condition characteristics and network-side business condition characteristics, and between network-side business condition characteristics and load-side business condition characteristics; obtaining pairs of business condition characteristic parameters that are strongly correlated among various business condition characteristics; and setting thresholds for changes in each business condition characteristic parameter by combining the business condition characteristic parameters and business condition clustering results. The analysis then selects and simulates slight and significant changes in a certain business condition characteristic parameter, determining whether the business condition characteristic parameters of related services exceed their thresholds. If they do, the business condition type of the related services changes, and the operating strategy changes synchronously; otherwise, the original operating strategy is maintained. Finally, a mapping table of changes in business condition characteristic parameters and changes in related business condition types is formed.

[0025] like Figure 2 As shown, in this embodiment, step S4 includes: Set the prompt word template for the scheduling control agent, including: defining the role of the agent as generating business operation strategies through reasoning based on real-time operation data and business requirements of the source network and load side; and setting the agent's thought chain task steps and intermediate conclusions for each step. The steps for setting up the intelligent agent's thought chain task include: uploading real-time operating data and scheduling control business requirements from the source, network, and load sides; extracting business condition feature parameters and business requirement data; similarity calculation and matching task rules; and outputting business operation strategies. The pre-trained intelligent agent for scheduling and control of the heating system uses set prompt word templates and thought chain task steps to guide the large model to transform real-time operating data of the source, grid and load sides and scheduling and control business requirements into understandable text instructions. After extracting business condition feature parameters and business requirement data, it calls the similarity calculation module to calculate the similarity between real-time scheduling and control business requirements and scheduling and control business requirement types in the offline benchmark library, and the similarity between real-time operating condition feature parameters and business condition types in the offline benchmark library. If both the similarity of business requirements and the similarity of business operating conditions exceed the threshold, the match is successful. The business operation strategy under the business requirement and business operating condition is obtained through the offline benchmark library. Based on the business operating condition correlation analysis results, the mapping table of changes in business operating condition characteristic parameters and changes in associated business operating condition types is used to determine whether the business operating condition affects the operating condition type of other types of scheduling and control services. If it does, the corresponding business operation strategy is searched from the offline benchmark library according to the changed operating condition type. If the operating condition type cannot be matched, the operating strategy is obtained after simulation analysis and verification optimization based on the heating system simulation model.

[0026] It should be noted that the heating system scheduling and control agent is a large language model agent, with the large language model as the core component, responsible for understanding and reasoning. The agent also integrates a perception module and an action module. The perception module is used to obtain heating system operation data, and the action module is used to schedule the similarity calculation module and the offline benchmark library to perform similarity calculation and matching, and business operation strategy decision-making.

[0027] In this embodiment, the calculation of the similarity between real-time scheduling and control service requirements and the scheduling and control service requirements of the offline benchmark library includes: extracting semantic features of real-time scheduling and control service requirements, assigning different weights to different semantic features using a weighted cosine similarity algorithm, and calculating the feature similarity with the scheduling and control service requirements of the offline benchmark library. The calculation of the similarity between real-time operating condition feature parameters and business operating condition types in the offline benchmark library includes: extracting operating condition features from real-time operating condition feature parameters, assigning different weights to different operating condition features using a weighted cosine similarity algorithm, and calculating the similarity between the operating condition features and the business operating condition types in the offline benchmark library, including the semantic similarity and numerical similarity of each operating condition feature. The different weights of different semantic features and different weights of different working condition features are calculated using a multi-dimensional algorithm to determine the optimal weight values ​​of different semantic features and different working condition features.

[0028] It should be noted that a weighted cosine similarity algorithm is used to assign different weights to different semantic features, and the feature similarity with the scheduling and control business requirement type of the offline benchmark library is calculated, expressed as: ; a i b is the i-th semantic feature in the real-time scheduling and control service requirements; i For the i-th feature corresponding to the business requirement type in the offline benchmark library; w i is the weight of the i-th feature; n is the number of semantic features; A weighted cosine similarity algorithm is used to assign different weights to features of different operating conditions, and the similarity of the features to the operating condition types of the offline benchmark library is calculated, including the semantic similarity and numerical similarity of each operating condition feature, as expressed as: ; Let X be the feature vector of the current operating condition, and Y be the feature parameter vector of the business operating condition type in the offline benchmark library; yy (X,Y) represents the semantic similarity of the working condition features calculated using the cosine similarity algorithm; sim sz (X,Y) represents the numerical similarity of the working condition features calculated using the cosine similarity algorithm; δ is the weight coefficient of the semantic similarity of the working condition features.

[0029] A multidimensional algorithm is used to calculate and determine the optimal weight values ​​for different semantic features and different working condition features, including: For different semantic features, the weights of each feature are calculated using the analytic hierarchy process (AHP), grey relational analysis, fuzzy comprehensive evaluation method-entropy method, and XGBoost algorithm with attention mechanism. Then, the weights of each semantic feature are combined using a cooperative game mechanism to obtain combined weight values ​​for different semantic features. Each semantic feature corresponds to four weight calculation results, which are then combined using the cooperative game mechanism to give each semantic feature a combined weight. For the calculation of weight values ​​for different working conditions, the same weight calculation method as for different semantic features is adopted.

[0030] In this embodiment, step S5 includes: The pre-trained intelligent agent for scheduling and control of the heating system uses the offline benchmark library as the training sample set. It takes the relevant features of scheduling and control business demand type, business condition type, and business condition correlation as input features and the business operation strategy as output features. After training and learning using machine algorithms, it constructs the business scheduling and control model for each side of the source, network, and load. After inputting the real-time operation data and scheduling control service requirements of the source, grid and load sides into the service scheduling control model of each side of the source, grid and load, the corresponding service operation strategy is output. Among them, after constructing the service scheduling and control models for each side of the source, network, and load, a total loss function is constructed, including the loss of the single-side service scheduling and control model and the cross-side coordination loss, and the model is optimized.

[0031] It should be noted that the scheduling control model uses an offline benchmark library as training samples, but it does not simply memorize the operating conditions and strategies in the samples. Instead, it uses machine learning algorithms to extract the inherent patterns between business requirements, operating condition characteristics, and operating strategies from the samples. When new operating conditions (data not in the benchmark library) are input, the model will analyze and reason about the characteristics of the new data based on the patterns learned during training, and then generate an appropriate operating strategy.

[0032] It should be noted that traditional side-based model training only focuses on the accuracy of a single-sided policy, which may lead to local optima but suboptimal global outcomes. Therefore, in the model training phase, in addition to the single-sided loss, a cross-side collaborative loss is added. The total loss function is expressed as: L = αL 单侧 +βL 协同 L 单侧 Measure the deviation between the output strategy of each side model and the single-sided optimal strategy in the benchmark library; L 协同 The degree of conflict or mismatch between source, grid, and load strategies is measured, including physical consistency loss and target consistency loss; α and β are loss weight coefficients; the physical consistency loss ensures that the strategies meet the physical laws between the source, grid, and load, and the target consistency loss ensures that the strategies on each side coordinate around the same target and avoid target conflict.

[0033] like Figure 3 As shown, in this embodiment, step S6 includes: The intelligent agent for scheduling and control of the heating system regards the process of integrating and analyzing business operation strategies as a decision-making task, and learns the optimal business operation strategy by interacting with the operating environment of the heating system. The state is defined as real-time running data, scheduling and control business requirements, strategy calculation and analysis related parameters in step S4, and strategy calculation and analysis related parameters in step S5. The action is defined as the business operation strategy output in step S4, the business operation strategy output in step S5, and the business operation strategy of the weighted fusion of steps S4 and S5; the business operation strategy of the weighted fusion of steps S4 and S5 adopts the analytic hierarchy process when calculating the strategy weight. The reward is defined as the heating operation effect after the implementation of the business operation strategy, including energy-saving and environmental protection effect, economic benefit effect, and load target achievement effect; The agent explores the action space through a greedy strategy, selects the strategy corresponding to the action to execute, and evaluates the expected reward of the action. After each action is executed, the current state, action, reward and the state data of the next moment are stored in the experience replay pool, and the next round of iteration begins until the reward is the highest. The corresponding action is then used as the final operation strategy for each scheduling and control service.

[0034] It should be noted that the process of obtaining the operation strategies for each scheduling and control service, based on the exploration mechanism of the greedy strategy and the iterative optimization logic of the experience replay pool, includes: Initialization phase: 1) Parameter initialization: Set the greed coefficient and initialize the experience replay pool for storing states, actions, rewards and the next state; initialize the policy network for evaluating the expected reward of actions; 2) Define the status, action, and reward; Iterative exploration phase: 1) Action selection: For the current state, the policy network calculates the expected reward value of three types of actions, selects the action with the highest expected reward value with probability ε, and randomly selects other actions with probability 1-ε to avoid getting trapped in local optima; 2) Strategy Execution and Reward Evaluation: Execute the selected action and calculate the reward; 3) Data storage and status update: Store the status, action, reward and next status into the experience replay pool, enter the next moment, repeat the iterative exploration, and continuously accumulate experience data; Strategy optimization phase: 1) Sample extraction and network update: Randomly extract batches of samples from the experience replay pool and use the samples to update the policy network parameters: Optimize the action evaluation accuracy by minimizing the error between the predicted reward value and the target reward value; 2) Greed coefficient decay: As the number of iterations increases, ε is gradually reduced to decrease the proportion of random exploration and enhance the utilization of the optimal action; Convergence determination and final policy output stage: 1) Convergence condition judgment: When the fluctuation range of the highest reward value is less than the threshold in multiple consecutive iterations, and the action selection output by the policy network tends to be stable, it is judged as convergence; 2) Final strategy determination: After convergence, for each scheduling and control service requirement, input its state, and the policy network outputs the action with the highest reward value in that state. The policy corresponding to the action is taken as the final running policy.

[0035] In this embodiment, step S4 further includes: If the similarity of business requirements and the similarity of business operating conditions are not matched, the corresponding business operation strategy will be obtained after simulation analysis and verification optimization based on the heating system simulation model, and added to the offline operation benchmark library.

[0036] In the several embodiments provided in this application, it should be understood that the disclosed systems and methods can also be implemented in other ways. The system embodiments described above are merely illustrative; for example, the flowcharts and block diagrams in the accompanying drawings illustrate the architecture, functionality, and operation of possible implementations of systems, methods, and computer program products according to various embodiments of the present invention. In this regard, each block in a flowchart or block diagram may represent a module, segment, or portion of code, which contains one or more executable instructions for implementing a specified logical function. It should also be noted that in some alternative implementations, the functions marked in the blocks may occur in a different order than those marked in the drawings. For example, two consecutive blocks may actually be executed substantially in parallel, and they may sometimes be executed in reverse order, depending on the functions involved. It should also be noted that each block in a block diagram and / or flowchart, and combinations of blocks in block diagrams and / or flowcharts, can be implemented using a dedicated hardware-based system that performs the specified function or action, or using a combination of dedicated hardware and computer instructions.

[0037] Furthermore, the functional modules in the various embodiments of this invention can be integrated together to form an independent part, or each module can exist independently, or two or more modules can be integrated to form an independent part. If the function is implemented as a software functional module and sold or used as an independent product, it can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of this invention, or the part that contributes to the prior art, or a part of the technical solution, can be embodied in the form of a software product. This computer software product is stored in a storage medium and includes several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute all or part of the steps of the methods in the various embodiments of this invention. The aforementioned storage medium includes various media capable of storing program code, such as USB flash drives, portable hard drives, read-only memory, random access memory, magnetic disks, or optical disks.

[0038] Based on the above-described preferred embodiments of the present invention, and through the foregoing description, those skilled in the art can make various changes and modifications without departing from the inventive concept. The technical scope of this invention is not limited to the contents of the specification, but must be determined according to the scope of the claims.

Claims

1. A method for optimizing scheduling and control of a heating system based on a benchmark database and operating condition matching, characterized in that, It includes: Step S1: Obtain historical operating data and object attribute data of the heating system source, grid and load sides, perform time alignment, data preprocessing and steady-state condition filtering, and construct a steady-state condition database for each side of the source, grid and load. Step S2: Obtain the dispatch control business requirements and historical operation strategy data of each dispatch control business on the source, network and load sides of the heating system. Combine the steady-state operating condition database of each source, network and load side to perform understanding and analysis of each dispatch control business requirement and operation strategy, define the characteristic parameters of the business conditions, cluster and classify the business conditions, and analyze the correlation of the business conditions. The dispatch control business requirements include the operation output dispatch control of each heating unit on the source side, the secondary network flow and temperature regulation of each heating station on the network side, and the pump and valve regulation of each building on the load side. Step S3: Based on the simulation model of the heating system, after conducting simulation analysis and verification optimization of multi-objective parameter operation strategies under different operating conditions of various scheduling and control services, construct an offline operation benchmark library that includes scheduling and control service demand types, service operating condition types, service operating condition correlations, and service operation strategies. Step S4: The pre-trained heating system dispatch and control agent obtains business condition characteristic parameters and business requirement data based on the real-time operation data of the source, grid and load sides and the dispatch and control business requirements, and calculates the similarity of dispatch and control business requirements and business condition types with the offline operation benchmark library respectively. If the similarity of business requirements and the similarity of business conditions both exceed the threshold, the match is successful. The business operation strategy under the business requirement and business condition is obtained by running the benchmark library offline, and the operation strategy of other types of scheduling and control services is adaptively adjusted according to the correlation of business conditions. Step S5: The pre-trained heating system scheduling and control agent uses the offline benchmark library as the training sample set to train and learn the scheduling and control models of each business. After constructing the scheduling and control models of each business, it inputs the real-time operation data of the source network and load side and the scheduling and control business requirements to obtain the corresponding business operation strategies. Step S6: Integrate and analyze the business operation strategies in steps S4 and S5 to obtain the final operation strategies for each scheduling and control service.

2. The heating system optimization scheduling and control method according to claim 1, characterized in that, Step S1 includes: Obtain the heat source output, supply and return water temperature, flow rate, fuel consumption, unit operating status and heat source type, installed capacity, unit equipment efficiency curve, and unit output constraints on the source side of the heating system; Acquire data on pump frequency, valve opening, heat exchange efficiency, network heat metering, network structure parameters, heating station coverage area, pump operating characteristics, and valve regulation characteristics of the heating system network-side heating stations. Obtain the indoor temperature of the building on the load side of the heating system, the flow rate at the end of the secondary network, the supply and return water temperatures, the building type, the building area, and the building heat user load; The acquired data timestamps are converted into a unified format, and the running data is collected uniformly according to the sampling frequency. A unique identifier is added to each time series data, and data association is established through topological relationships. After handling missing values ​​and outliers, key parameters on the source, grid, and load sides are selected as the basis for steady-state determination. Combined with steady-state period filtering, a steady-state operating condition database for each side of the source, grid, and load is constructed. The database structure includes a steady-state operating condition master table, a source-side steady-state data table, a grid-side steady-state data table, a load-side steady-state data table, and an object attribute table.

3. The heating system optimization scheduling and control method according to claim 2, characterized in that, The selection of key parameters on the source, grid, and load sides as the basis for steady-state determination includes: heat consumption fluctuations, water supply temperature fluctuations, and unit load rate fluctuations on the source side; temperature difference fluctuations at the inlet and outlet of the heating station, pump frequency fluctuations, and secondary network flow fluctuations on the load side; building water supply temperature fluctuations and building heat load fluctuations on the load side; and outdoor temperature fluctuations among environmental parameters. The steady-state period selection method includes: using the sliding window method to calculate the fluctuation values ​​of each key parameter within the window; if the fluctuation of all parameters is less than the threshold, the window is determined to be a steady-state period; merging adjacent steady-state windows into a longer steady-state period; and considering the correlation between the steady-state conditions of the source, grid, and load sides, verifying whether each side meets the steady-state conditions in the same period.

4. The heating system optimization scheduling and control method according to claim 1, characterized in that, In step S2, the operational needs and strategies of each scheduling and control operation are analyzed, including: analyzing the operational output scheduling and control needs of each heating unit on the source side to meet the total heat load demand of the grid side while taking into account multiple objectives such as energy conservation, environmental protection, and economic benefits; and recording the output commands and start-up / shutdown plans of each unit on the source side under steady-state conditions; analyzing the secondary network flow and temperature regulation needs of each heating station on the grid side to match the secondary network flow and temperature with the heat demand on the load side and eliminate hydraulic imbalance; and recording the secondary network water supply temperature curve, flow regulation commands, and valve opening changes of the heating station under steady-state conditions; analyzing the pump and valve regulation needs of each building on the load side to meet the indoor temperature of building heat users and respond to grid side load regulation commands; and recording the pump and valve regulation parameters at the building entrance and the valve regulation parameters in front of the heat users; and as well as associating the operational strategy data with the steady-state condition database through timestamps and spatial topology to form a condition-strategy mapping combination for source-grid-load coordinated operation. The definition of service condition characteristic parameters includes: source-side service condition characteristics, grid-side service condition characteristics, and load-side service condition characteristics; the source-side service condition characteristics include outdoor temperature, total grid-side heat load demand, and peak-valley ratio; the grid-side service condition characteristics include outdoor temperature, load demand of heat users under the jurisdiction of the heating station, and flow-temperature difference combination; the load-side service condition characteristics include outdoor temperature, heat load demand of each building, indoor temperature, and building type; Service condition clustering includes: based on the source-side service condition characteristics, network-side service condition characteristics and load-side service condition characteristics, clustering algorithms are used to group similar features in the steady-state service condition database into one category, and service condition clustering is performed on the source-side, network-side and load-side. The business condition correlation analysis includes: calculating the correlation coefficients between source-side business condition characteristics and network-side business condition characteristics, and between network-side business condition characteristics and load-side business condition characteristics; obtaining pairs of business condition characteristic parameters that are strongly correlated among various business condition characteristics; and setting thresholds for changes in each business condition characteristic parameter by combining the business condition characteristic parameters and business condition clustering results. The analysis then selects and simulates slight and significant changes in a certain business condition characteristic parameter, determining whether the business condition characteristic parameters of related services exceed their thresholds. If they do, the business condition type of the related services changes, and the operating strategy changes synchronously; otherwise, the original operating strategy is maintained. Finally, a mapping table of changes in business condition characteristic parameters and changes in related business condition types is formed.

5. The heating system optimization scheduling and control method according to claim 1, characterized in that, Step S4 includes: Set the prompt word template for the scheduling control agent, including: defining the role of the agent as generating business operation strategies through reasoning based on real-time operation data and business requirements of the source network and load side; and setting the agent's thought chain task steps and intermediate conclusions for each step. The steps for setting up the intelligent agent's thought chain task include: uploading real-time operating data and scheduling control business requirements from the source, network, and load sides; extracting business condition feature parameters and business requirement data; similarity calculation and matching task rules; and outputting business operation strategies. The pre-trained intelligent agent for scheduling and control of the heating system uses set prompt word templates and thought chain task steps to guide the large model to transform real-time operating data of the source, grid and load sides and scheduling and control business requirements into understandable text instructions. After extracting business condition feature parameters and business requirement data, it calls the similarity calculation module to calculate the similarity between real-time scheduling and control business requirements and scheduling and control business requirement types in the offline benchmark library, and the similarity between real-time operating condition feature parameters and business condition types in the offline benchmark library. If both the similarity of business requirements and the similarity of business operating conditions exceed the threshold, the match is successful. The business operation strategy under the business requirement and business operating condition is obtained through the offline benchmark library. Based on the business operating condition correlation analysis results, the mapping table of changes in business operating condition characteristic parameters and changes in associated business operating condition types is used to determine whether the business operating condition affects the operating condition type of other types of scheduling and control services. If it does, the corresponding business operation strategy is searched from the offline benchmark library according to the changed operating condition type. If the operating condition type cannot be matched, the operating strategy is obtained after simulation analysis and verification optimization based on the heating system simulation model.

6. The heating system optimization scheduling and control method according to claim 5, characterized in that, The calculation of the similarity between real-time scheduling and control service requirements and the scheduling and control service requirements of the offline benchmark library includes: extracting the semantic features of real-time scheduling and control service requirements, assigning different weights to different semantic features using a weighted cosine similarity algorithm, and calculating the feature similarity with the scheduling and control service requirements of the offline benchmark library. The calculation of the similarity between real-time operating condition feature parameters and business operating condition types in the offline benchmark library includes: extracting operating condition features from real-time operating condition feature parameters, assigning different weights to different operating condition features using a weighted cosine similarity algorithm, and calculating the similarity between the operating condition features and the business operating condition types in the offline benchmark library, including the semantic similarity and numerical similarity of each operating condition feature. The different weights of different semantic features and different weights of different working condition features are calculated using a multi-dimensional algorithm to determine the optimal weight values ​​of different semantic features and different working condition features.

7. The heating system optimization scheduling and control method according to claim 1, characterized in that, Step S5 includes: The pre-trained intelligent agent for scheduling and control of the heating system uses the offline benchmark library as the training sample set. It takes the relevant features of scheduling and control business demand type, business condition type, and business condition correlation as input features and the business operation strategy as output features. After training and learning using machine algorithms, it constructs the business scheduling and control model for each side of the source, network, and load. After inputting the real-time operation data and scheduling control service requirements of the source, grid and load sides into the service scheduling control model of each side of the source, grid and load, the corresponding service operation strategy is output. Among them, after constructing the service scheduling and control models for each side of the source, network, and load, a total loss function is constructed, including the loss of the single-side service scheduling and control model and the cross-side coordination loss, and the model is optimized.

8. The heating system optimization scheduling and control method according to claim 1, characterized in that, Step S6 includes: The intelligent agent for scheduling and control of the heating system regards the process of integrating and analyzing business operation strategies as a decision-making task, and learns the optimal business operation strategy by interacting with the operating environment of the heating system. The state is defined as real-time running data, scheduling and control business requirements, strategy calculation and analysis related parameters in step S4, and strategy calculation and analysis related parameters in step S5. The action is defined as the business operation strategy output in step S4, the business operation strategy output in step S5, and the business operation strategy of the weighted fusion of steps S4 and S5; the business operation strategy of the weighted fusion of steps S4 and S5 adopts the analytic hierarchy process when calculating the strategy weight. The reward is defined as the heating operation effect after the implementation of the business operation strategy, including energy-saving and environmental protection effect, economic benefit effect, and load target achievement effect; The agent explores the action space through a greedy strategy, selects the strategy corresponding to the action to execute, and evaluates the expected reward of the action. After each action is executed, the current state, action, reward and the state data of the next moment are stored in the experience replay pool, and the next round of iteration begins until the reward is the highest. The corresponding action is then used as the final operation strategy for each scheduling and control service.

9. The heating system optimization scheduling and control method according to claim 1, characterized in that, Step S4 further includes: If the similarity of business requirements and the similarity of business operating conditions are not matched, the corresponding business operation strategy will be obtained after simulation analysis and verification optimization based on the heating system simulation model, and added to the offline operation benchmark library.

Images