A smart heating comprehensive scheduling method and system based on operation data analysis

Abstract

The application relates to the technical field of data scheduling, and particularly discloses a wisdom heat supply comprehensive scheduling method and system based on operation data analysis, which comprises the following steps: performing data preprocessing on a historical operation data set of a monitoring point in a target regional heat supply pipe network system in a historical same period, so as to obtain standard historical operation data; inputting the standard historical operation data and corresponding actual heat load values into a preset initial neural network model for iterative training, so as to construct a load prediction model; inputting real-time operation data into the load prediction model, so as to generate a heat load demand curve; performing multi-objective decision optimization on current operation state data, so as to obtain a Pareto optimal solution set; performing priority screening on the Pareto optimal solution set, so as to obtain a candidate scheduling scheme and issue the candidate scheduling scheme to an execution mechanism of the target regional heat supply pipe network system; and collecting feedback operation data of the execution mechanism in real time, so as to update the load prediction model; and the wisdom heat supply comprehensive scheduling efficiency can be improved.

Description

Technical Field

[0001] This invention relates to the field of data scheduling technology, and in particular to a smart heating integrated scheduling method and system based on operational data analysis. Background Technology

[0002] In the current field of smart heating dispatching, traditional dispatching methods have low efficiency in utilizing heating network operation data. They lack systematic processing and in-depth analysis of historical operation data from the same period, making it difficult to extract effective patterns from massive amounts of raw data. This results in insufficient accuracy in heat load forecasting and an inability to accurately match the actual heat demand of users. Furthermore, existing dispatching methods largely rely on manual experience to formulate plans, failing to incorporate dynamic adjustments based on real-time operation data. This can easily lead to problems such as excessive heating energy consumption and imbalances in network operation. Therefore, the overall scientific and rational nature of dispatching urgently needs improvement.

[0003] Existing smart heating dispatching technologies have significant shortcomings in multi-objective optimization and dynamic adaptation. They only consider a few indicators such as achieving heating standards, failing to take into account multiple needs such as energy consumption control and stable pipeline operation. This limits the optimization dimensions and makes it difficult to form a comprehensively optimal dispatching scheme. Furthermore, once the dispatching model is built, it lacks an effective real-time update mechanism, making it impossible to adjust model parameters based on actual operational feedback data from the heating pipeline network. This leads to a gradual decline in model adaptability and a decreasing degree of matching between the dispatching scheme and the actual heating scenario, ultimately affecting the overall dispatching efficiency and heating service quality of smart heating. Therefore, improving the overall dispatching efficiency of smart heating has become an urgent problem to be solved. Summary of the Invention

[0004] In view of the above problems, the purpose of this invention is to provide a smart heating integrated scheduling method and system based on operational data analysis, so as to solve the problems in existing smart heating scheduling technology, such as insufficient data support due to the lack of standardized preprocessing process for historical operational data of heating pipe networks, low prediction accuracy due to the lack of scientific parameter adjustment in load prediction model construction and the absence of a dynamic update mechanism based on actual operational feedback, scheduling decisions that do not comprehensively consider multiple objectives such as heating energy consumption, room temperature compliance rate, and pipe network stability, and lack a scientific screening method that combines dynamic hydraulic balance constraints, and the lack of closed-loop management in the overall scheduling process, which makes it impossible to optimize scheduling strategies and models based on execution feedback, ultimately resulting in low efficiency of smart heating integrated scheduling and poor matching degree between scheduling schemes and actual heating scenarios.

[0005] This invention provides a smart heating integrated scheduling method based on operational data analysis, comprising: Step 1: Perform data preprocessing on the historical operation dataset of the monitoring points in the target area's heating network system during the same historical period to obtain the standard historical operation data of the target area's heating network system; Step 2: Input the standard historical operating data and the corresponding actual heat load values ​​into the preset initial neural network model for iterative training to build a load prediction model for the target area heating network system; Step 3: Input the real-time operation data of the target area heating network system into the load prediction model to generate the heat load demand curve of the target area heating network system; Step 4: Based on the heat load demand curve, perform multi-objective decision optimization on the current operating status data of the heating network system in the target area to obtain the Pareto optimal solution set of the heating network system in the target area; Step 5: Based on the dynamic hydraulic balance constraints of the target region's heating network system, prioritize the Pareto optimal solution set to obtain candidate scheduling schemes for the target region's heating network system. Step 6: Distribute the candidate scheduling scheme to the execution agency of the heating network system in the target area, and collect the feedback operation data of the execution agency in real time to update the load prediction model.

[0006] Preferably, in step 1, standard historical operating data of the target area's heating network system are obtained, and the process is as follows: Obtain historical operational datasets of monitoring points in the heating network system of the target region during the same historical time period; By removing outliers from the historical operation dataset, we obtain the cleaned historical operation data of the heating network system in the target region. The historical operating data after cleaning is transformed to a unified dimension to obtain the unified dimension historical operating data of the heating network system in the target area. By aligning the dimension-unified historical operation data with the time dimension, the time-aligned historical operation data of the heating network system in the target region is obtained. Missing time periods are filled in the time-series aligned historical operation data to obtain the standard historical operation data of the heating network system in the target area.

[0007] Preferably, in step 2, a load prediction model for the target area's heating network system is constructed, and the process is as follows: The standard historical operating data is divided into the training input dataset, and the actual heat load values ​​corresponding to the standard historical operating data are divided into the training target dataset. The training input dataset is input into a preset initial neural network model. Through the forward propagation process of the initial neural network model, the predicted heat load value corresponding to the training input dataset is obtained. The predicted heat load value is compared with the corresponding actual heat load value in the training target dataset, and the internal parameters of the initial neural network model are adjusted based on the comparison results. The adjusted neural network model is used as the updated model, and the training process is repeated until the updated model meets the preset convergence condition. The updated model that meets the convergence criteria is identified as the load prediction model for the target region's heating network system.

[0008] Preferably, the internal parameters of the initial neural network model are adjusted based on the comparison results, and the process is as follows: The comparison results are extracted to show the error between the predicted heat load value and the corresponding actual heat load value of the initial neural network model under the current iteration training, as well as the error value under the previous iteration training. Obtain the output values ​​of the hidden layer nodes in the initial neural network model under the current iteration of training, as well as the adjustment amount of the connection weights between the hidden layer nodes and the output layer nodes in the previous iteration; The adjustment amount of the connection weights between hidden layer nodes and output layer nodes in the initial neural network model is calculated under the current training iteration. The formula for calculating the adjustment amount under the current training iteration is as follows: ; In the formula, This represents the adjustment amount of the connection weights between hidden layer nodes and output layer nodes in the initial neural network model under the current iteration of training. This represents the preset base learning rate. This represents the error between the predicted heat load value and the corresponding actual heat load value under the current iteration of training. This represents the error between the predicted heat load value and the corresponding actual heat load value from the previous training iteration. Represents the hidden layer nodes in the initial neural network model The output value under the current training iteration. This represents the preset momentum coefficient. Represents the hidden layer nodes in the previous iteration To the output layer node The adjustment amount of the connection weights between them. This indicates a preset trend factor; The connection weights of the hierarchical nodes in the initial neural network model are dynamically updated based on the adjustment of the connection weights between the hidden layer nodes and the output layer nodes in the current iteration of training.

[0009] Preferably, in step 3, the real-time operating data of the target area heating network system is input into the load prediction model to generate the heat load demand curve of the target area heating network system. The process is as follows: According to the preset time step, the real-time operation data of the target area heating network system is divided into real-time operation data units corresponding to consecutive time points; Input the real-time operational data units corresponding to different time points into the load prediction model to obtain the predicted heat load value of the target area's heating network system. The predicted heat load values ​​are smoothly arranged in chronological order to obtain the heat load demand curve of the heating network system in the target area.

[0010] Preferably, in step 4, based on the heat load demand curve, multi-objective decision optimization is performed on the current operating status data of the target area heating network system to obtain the Pareto optimal solution set of the target area heating network system. The process is as follows: The heating energy consumption of the target area's heating network system, the rate of compliance of user-end room temperature, and the stability of network operation are taken as the optimization target set. Determine the operational constraints of the heating network system in the target area. These constraints include the upper limit of network pressure, the range of water supply temperature regulation, and the limit on the rate of flow change. Based on the heat load demand curve and the current operating status data of the target area's heating network system, a candidate heating scheduling parameter group for the target area's heating network system is generated. Based on the optimization target set, the expected performance of the candidate heating scheduling parameter group is evaluated, and the target performance vector of the candidate heating scheduling parameter group is obtained. The target performance vector is compared pairwise internally to select the Pareto optimal solution set for the target regional heating network system from the candidate heating scheduling parameter set.

[0011] Preferably, the Pareto optimal solution set for the target area's heating network system is selected from the candidate heating scheduling parameter set, and the process is as follows: Obtain the target performance vector of the candidate heating scheduling parameter group, which contains the expected performance values ​​of the candidate heating scheduling parameter group on the three optimization objectives; The target performance vectors of the candidate heating scheduling parameter groups are compared with the expected performance values ​​of the target performance vectors of other candidate heating scheduling parameter groups. If, for a candidate heating scheduling parameter group, there is no other candidate heating scheduling parameter group whose target performance vector is better than the target performance vector of the candidate heating scheduling parameter group in all three optimization objectives, then the candidate heating scheduling parameter group is determined as the Pareto optimal candidate solution. The candidate heating scheduling parameter sets that will be identified as Pareto optimal candidate schemes will be collected to form the Pareto optimal solution set of the target area heating network system.

[0012] Preferably, in step 5, the Pareto optimal solution set is prioritized to obtain candidate scheduling schemes for the target area heating network system. The process is as follows: Obtain the pipeline hydraulic balance simulation data corresponding to the candidate heating scheduling parameter group in the Pareto optimal solution set, and compare the pipeline hydraulic balance simulation data with the preset dynamic hydraulic balance constraint threshold. Based on the comparison results, candidate heating scheduling parameter groups whose pipeline hydraulic balance simulation data are all within the dynamic hydraulic balance constraint threshold range are determined as qualified hydraulic balance schemes. According to the preset scheduling priority rules, the qualified hydraulic balance schemes are prioritized and the highest-ranked qualified hydraulic balance scheme is determined as the candidate scheduling scheme for the heating network system in the target area.

[0013] Preferably, in step 6, the candidate scheduling scheme is sent to the execution agency of the target area heating network system, and the feedback operation data of the execution agency is collected in real time to update the load prediction model. The process is as follows: The candidate scheduling schemes are formatted and encoded to obtain the scheduling instruction sequence of the target area heating network system; The sequence of dispatch instructions is sent to the actuators of the heating network system in the target area to control the actuators to operate according to the candidate dispatch scheme; Real-time acquisition of feedback operational data generated by the actuator during operation, including actual operating parameters of the heating network system in the target area; The feedback operation data is compared with the expected operation parameters in the candidate scheduling scheme to determine the deviation trend between the feedback operation data and the expected operation parameters; Based on the deviation trend, the internal parameters of the load forecasting model are dynamically adjusted to update the load forecasting model.

[0014] This invention also provides a smart heating integrated scheduling system based on operational data analysis, the system comprising: The data preprocessing module is used to preprocess the historical operation dataset of the monitoring points in the heating network system of the target region within the same historical time period to obtain the standard historical operation data of the heating network system of the target region. The prediction model training module is used to input standard historical operating data and corresponding actual heat load values ​​into a preset initial neural network model for iterative training in order to build a load prediction model for the target region's heating network system. The real-time load forecasting module is used to input the real-time operating data of the heating network system in the target area into the load forecasting model and generate the heat load demand curve of the heating network system in the target area. The multi-objective optimization solution module is used to perform multi-objective decision optimization on the current operating status data of the heating network system in the target area based on the heat load demand curve, and obtain the Pareto optimal solution set of the heating network system in the target area. The scheduling scheme screening module is used to prioritize the Pareto optimal solution set based on the dynamic hydraulic balance constraints of the target area heating network system, and obtain candidate scheduling schemes for the target area heating network system. The scheduling and execution module is used to distribute candidate scheduling schemes to the execution agencies of the target area heating network system and collect feedback operation data from the execution agencies in real time to update the load prediction model.

[0015] As can be seen from the above technical solution, the present invention provides a smart heating integrated scheduling method and system based on operational data analysis. By standardizing and preprocessing historical operational data, a reliable data foundation is laid for load forecasting. Combined with scientific training and dynamic parameter updates of a neural network model, the accuracy of heat load forecasting is significantly improved, and a precise heat load demand curve is generated. Based on multi-objective decision optimization and combined with Pareto optimal solution set screening, the system comprehensively considers heating energy consumption, room temperature compliance rate, and network stability. Simultaneously, priority screening is performed based on dynamic hydraulic balance constraints, making the scheduling scheme more scientific, reasonable, and aligned with actual heating needs. After scheduling execution, feedback data is collected in real time and the load forecasting model is updated, forming a closed-loop management of the entire scheduling process, continuously optimizing model adaptability and scheduling strategies. Overall, the efficiency of smart heating integrated scheduling is improved, effectively balancing heating effect, energy consumption control, and network operational stability, reducing energy waste, ensuring user heating experience, and making the heating network system operate more efficiently, intelligently, and stably. Attached Figure Description

[0016] Other objects and results of the invention will become more apparent and readily understood by referring to the following description taken in conjunction with the accompanying drawings, and with a more complete understanding of the invention. In the drawings: Figure 1 This is a flowchart illustrating a smart heating integrated scheduling method based on operational data analysis according to an embodiment of the present invention. Figure 2 This is a functional module diagram of a smart heating integrated dispatching system based on operational data analysis, according to an embodiment of the present invention. Figure 3 This is a specific implementation result of the prediction model training module in a smart heating integrated scheduling system based on operational data analysis, according to an embodiment of the present invention. Detailed Implementation

[0017] In existing smart heating dispatching technologies, the lack of standardized preprocessing procedures for historical operation data of heating pipe networks leads to insufficient data support. The parameter adjustments in the load prediction model lack scientific rigor and there is no dynamic update mechanism based on actual operation feedback, resulting in low prediction accuracy. Dispatch decisions do not comprehensively consider multiple objectives such as heating energy consumption, room temperature compliance rate, and pipe network stability, and lack a scientific screening method that incorporates dynamic hydraulic balance constraints. At the same time, the overall dispatching process lacks closed-loop management and cannot optimize dispatching strategies and models based on execution feedback. Ultimately, this results in low overall efficiency of smart heating dispatching and poor matching between dispatching schemes and actual heating scenarios.

[0018] To address the aforementioned problems, this invention provides a smart heating integrated scheduling method and system based on operational data analysis. The specific embodiments of this invention will be described in detail below with reference to the accompanying drawings.

[0019] To illustrate the intelligent heating integrated scheduling method and system based on operational data analysis provided by this invention, Figure 1 An exemplary illustration is provided for a smart heating integrated scheduling method based on operational data analysis according to an embodiment of the present invention; Figure 2 An exemplary illustration is provided for a smart heating integrated scheduling system based on operational data analysis according to an embodiment of the present invention.

[0020] The following description of exemplary embodiments is merely illustrative and is in no way intended to limit the invention or its application or use. Techniques and equipment known to those skilled in the art may not be discussed in detail, but where appropriate, such techniques and equipment should be considered part of the specification.

[0021] Reference Figure 1 The diagram shown is a flowchart illustrating a smart heating integrated scheduling method based on operational data analysis according to an embodiment of the present invention. In this embodiment, the smart heating integrated scheduling method based on operational data analysis includes: Step 1: Perform data preprocessing on the historical operation dataset of the monitoring points in the target area's heating network system during the same historical period to obtain the standard historical operation data of the target area's heating network system; In this embodiment of the invention, in step 1, standard historical operating data of the target area's heating network system are obtained, and the process is as follows: Obtain historical operational datasets of monitoring points in the heating network system of the target region during the same historical time period; By removing outliers from the historical operation dataset, we obtain the cleaned historical operation data of the heating network system in the target region. The historical operating data after cleaning is transformed to a unified dimension to obtain the unified dimension historical operating data of the heating network system in the target area. By aligning the dimension-unified historical operation data with the time dimension, the time-aligned historical operation data of the heating network system in the target region is obtained. Missing time periods are filled in the time-series aligned historical operation data to obtain the standard historical operation data of the heating network system in the target area.

[0022] From all the monitoring points deployed within the target area's heating network system, determine the historical time period from which data needs to be extracted. This historical time period differs from the current analysis period in year but corresponds completely in date and time. According to the distribution location of the monitoring points and the number of the data acquisition device, retrieve all the operational data continuously collected by each monitoring point within the historical time period. Summarize and arrange the various types of operational data from different monitoring points in chronological order of collection time to ensure that each data point corresponds to a unique monitoring point and a unique collection time. Finally, form a complete and comprehensive historical operational dataset of the monitoring points within the target area's heating network system during the historical time period.

[0023] According to the preset normal data fluctuation range, each data record in the historical operation dataset is checked one by one. The value of a single data record is compared with the values ​​of the same data type at multiple consecutive time points at the same monitoring point. Data records whose values ​​exceed the long-term stable fluctuation range of the data type at the monitoring point are directly identified as abnormal data. All identified abnormal data records are completely removed from the dataset, and no abnormal data fragments are retained. Only valid data records whose values ​​conform to the normal fluctuation pattern and are consistent with the trend of adjacent data are retained. Finally, complete and standardized historical operation data of the target area heating network system after cleaning are obtained.

[0024] The units of measurement used for different operating parameters such as temperature, pressure, flow rate, and energy consumption in the historical operating data after cleaning are identified one by one. A unique standard unit of measurement is determined for each type of physical quantity. Data records using non-standard units of measurement are adjusted according to the actual conversion relationship of the physical quantity, so that the values ​​of all physical quantities of the same type correspond to a unified standard unit of measurement. During the conversion process, the actual physical meaning of the data is not changed. After all data is converted, a unified historical operating data of the target area heating network system with consistent numerical meaning and unified units is formed.

[0025] Using a unified and fixed time scale as a benchmark, the collection time of each data point in the historical operation data with unified dimensions is standardized. Data collected from different monitoring points at similar time points are normalized to the same standard time node, ensuring that all monitoring points have corresponding data records at the same standard time point. This eliminates time misalignment caused by differences in collection clocks or collection frequencies between different monitoring points, enabling all data to correspond one-to-one on the time axis. After time normalization, a time-series aligned historical operation data of the target regional heating network system with complete time series correspondence is formed.

[0026] A comprehensive time-series alignment of historical operational data is conducted to identify time periods lacking valid data records. For each identified missing time period, based on the trend of valid data changes at both ends of the missing time period for that type of data at that monitoring point, data is extrapolated for each standard time point within the missing time period. The extrapolated reasonable values ​​are then filled into the corresponding missing time positions to ensure that the entire data sequence is continuous and uninterrupted in time. After all missing time periods are filled, standard historical operational data of the target region's heating network system with complete time data continuity is formed.

[0027] Step 2: Input the standard historical operating data and the corresponding actual heat load values ​​into the preset initial neural network model for iterative training to build a load prediction model for the target area heating network system; In this embodiment of the invention, in step 2, a load prediction model for the target area's heating network system is constructed, and the process is as follows: The standard historical operating data is divided into the training input dataset, and the actual heat load values ​​corresponding to the standard historical operating data are divided into the training target dataset. The training input dataset is input into a preset initial neural network model. Through the forward propagation process of the initial neural network model, the predicted heat load value corresponding to the training input dataset is obtained. The predicted heat load value is compared with the corresponding actual heat load value in the training target dataset, and the internal parameters of the initial neural network model are adjusted based on the comparison results. The adjusted neural network model is used as the updated model, and the training process is repeated until the updated model meets the preset convergence condition. The updated model that meets the convergence criteria is identified as the load prediction model for the target region's heating network system.

[0028] Based on the comparison results, the internal parameters of the initial neural network model are adjusted as follows: The comparison results are extracted to show the error between the predicted heat load value and the corresponding actual heat load value of the initial neural network model under the current iteration training, as well as the error value under the previous iteration training. Obtain the output values ​​of the hidden layer nodes in the initial neural network model under the current iteration of training, as well as the adjustment amount of the connection weights between the hidden layer nodes and the output layer nodes in the previous iteration; The adjustment amount of the connection weights between hidden layer nodes and output layer nodes in the initial neural network model is calculated under the current training iteration. The formula for calculating the adjustment amount under the current training iteration is as follows: ; In the formula, This represents the adjustment amount of the connection weights between hidden layer nodes and output layer nodes in the initial neural network model under the current iteration of training. This represents the preset base learning rate. This represents the error between the predicted heat load value and the corresponding actual heat load value under the current iteration of training. This represents the error between the predicted heat load value and the corresponding actual heat load value from the previous training iteration. Represents the hidden layer nodes in the initial neural network model The output value under the current training iteration. This represents the preset momentum coefficient. Represents the hidden layer nodes in the previous iteration To the output layer node The adjustment amount of the connection weights between them. This indicates a preset trend factor; The connection weights of the hierarchical nodes in the initial neural network model are dynamically updated based on the adjustment of the connection weights between the hidden layer nodes and the output layer nodes in the current iteration of training.

[0029] Based on the timestamps, network areas, monitoring point numbers, and attribute categories of various operating parameters corresponding to the standard historical operating data, each data point is broken down and precisely categorized field by field. All historical operating parameters that can directly reflect the real-time operating status of the heating network, including but not limited to network-related operating parameters, are integrated into the feature information upon which the model input depends. These parameters are arranged strictly and orderly according to the time sequence from earliest to latest and the corresponding relationship of the monitoring points, ensuring that each input data has a unique corresponding time and monitoring point information. This forms a training input dataset that is structurally complete, temporally corresponding, and dimensionally unified. At the same time, the actual heat load values ​​that completely match the training input dataset in terms of time nodes, monitoring point locations, and parameter correspondences are collected one by one and arranged neatly in the same temporal order as the training input dataset, forming a training target dataset specifically used for benchmarking and guiding the model's learning direction during model training.

[0030] Each data point in the completed training input dataset is sequentially and smoothly fed into the pre-built initial neural network model according to a predetermined data transmission order and field correspondence rules. After entering the model system from the input layer node of the initial neural network model, the data is passed forward layer by layer strictly following the pre-set hierarchical connection relationship within the model. First, it is stably transmitted to the hidden layer through the connection path between the input layer node and the hidden layer node. Then, the hidden layer node processes the incoming multi-dimensional and multi-type running data layer by layer, extracts features, and integrates information. After processing, it continues to be passed forward to the output layer node according to the established connection path. Finally, the output layer node comprehensively summarizes and outputs all the information processed by the hidden layer to generate a heat load prediction result that completely corresponds to the input data. This prediction result is the predicted heat load value that accurately corresponds to the training input dataset.

[0031] Each predicted heat load value output by the initial neural network model is compared precisely and meticulously with the actual heat load values ​​at the same time point, under the same monitoring point number, and with the same data correspondence in the training target dataset. For each set of corresponding predicted and actual values, direct comparisons are made of numerical magnitude, numerical change range, and consistency of overall change trend. The specific numerical deviations, differences in change range, and trend deviations between each set of data are fully recorded, forming detailed and traceable comparison results.

[0032] In the detailed comparison results generated after a comprehensive comparison between the predicted and actual heat load values, all error values ​​between the predicted and actual heat load values ​​output by the initial neural network model during the entire iterative training process are extracted separately. These error values ​​are then uniformly summarized and recorded in a standardized manner. At the same time, all error values ​​generated during the previous iterative training process are extracted, independently organized, and saved. The error values ​​of this iterative training and the previous iterative training are organized separately to provide complete and reliable data support for the subsequent calculation and adjustment of connection weights.

[0033] Throughout the entire process of forward data transmission and internal processing during this iteration of training, the specific values ​​output by each hidden layer node within the initial neural network model are collected in real time and completely after receiving data from the previous layer and completing its own data processing operations. The current output values ​​of all hidden layer nodes are comprehensively collected and organized in an orderly manner. At the same time, the specific adjustment amounts of each pair of connection weights between the hidden layer nodes and the output layer nodes during the previous iteration of training are accurately retrieved. The historical adjustment amounts are completely obtained and combined with the currently collected hidden layer output values ​​to make full preparations for subsequent weight adjustment calculations.

[0034] By combining all the error values ​​obtained in this iteration of training, all the error values ​​obtained in the previous iteration of training, the output values ​​of all hidden layer nodes in the current iteration of training, and the adjustment amount of the connection weights between hidden layer nodes and output layer nodes in the previous iteration of training, strictly following the forward propagation logic of data from the input layer to the output layer within the neural network and the influence rules of the backward propagation of error from the output layer to the hidden layer, we step by step deduce, calculate layer by layer, and accurately calculate the specific values ​​that each set of connection weights between hidden layer nodes and output layer nodes in the initial neural network model in the current iteration of training needs to be adjusted, ensuring that the adjustment amount of each set of weights is completely matched with the current model error state.

[0035] Using the adjustment amount of the connection weights between the hidden layer nodes and the output layer nodes, which is obtained through complete deduction and precise calculation under the current iteration training, each original pair of connection weights between the corresponding layer nodes in the initial neural network model is updated one by one. The specific values ​​of the original connection weights are changed strictly according to the numerical direction and magnitude of the adjustment amount, ensuring that each pair of related connection weights is accurately and correspondingly corrected, and the dynamic update operation of the connection weights between all related layer nodes in the model is completed.

[0036] In the formula for calculating the adjustment amount of the connection weights between hidden layer nodes and output layer nodes, the base learning rate comes from a fixed value pre-set before model training, used to control the magnitude of each weight adjustment. The error value is calculated from the difference between the predicted heat load value output by the model in the current iteration and the corresponding actual heat load value, obtained by directly comparing the magnitudes of the two values. The previous iteration error value comes from the difference between the predicted heat load value output by the model in the previous training round and the corresponding actual heat load value, representing the error result of the previous training round. The hidden layer node output value comes from the specific value output by the hidden layer node after receiving and processing data in the current iteration, representing the intermediate result of the model's internal processing. The momentum coefficient comes from a fixed value pre-set before model training, used to incorporate the influence of the previous round of weight adjustments. The previous iteration weight adjustment amount comes from the adjustment amount of the connection weights between hidden layer nodes and output layer nodes calculated in the previous training round, representing the weight adjustment result of the previous training round. The trend factor comes from a fixed value pre-set before model training, used to incorporate the influence of the difference between the current error and the previous error.

[0037] This formula is used to calculate the adjustment amount of the connection weights between the hidden layer nodes and the output layer nodes in the current iteration of training of the initial neural network model. By integrating the error influence of the basic learning rate control, the historical adjustment influence of the momentum coefficient control, and the error change influence of the trend factor control, the specific value that each set of connection weights needs to be adjusted is determined in a comprehensive manner, which guides the direction and magnitude of the model weight update, so that the predicted heat load value output by the model gradually approaches the actual heat load value.

[0038] As the number of training iterations increases, the weight adjustment calculated by this formula will gradually decrease. When the model error stabilizes within the preset range and no longer changes significantly, the weight adjustment will approach zero, the model weight update will stop, and the model will eventually reach a convergent state.

[0039] The initial neural network model after updating all connection weights is redefined as the update model. The training input dataset is then smoothly fed into the update model again according to the original data transmission rules, arrangement order, and field correspondence. The complete training process is repeated, including forward data transmission to obtain the predicted heat load value, detailed comparison of the predicted heat load value with the actual heat load value, extraction of the error value between the current iteration and the previous iteration, obtaining the current output value of the hidden layer node and the previous connection weight adjustment, calculation of the connection weight adjustment in the current iteration, and updating all relevant connection weights within the model. This training process is continuously and cyclically executed until the prediction result output by the update model stably meets the pre-set convergence conditions.

[0040] After the updated model has undergone multiple consecutive iterations of training, the error between the output predicted heat load value and the corresponding actual heat load value remains stable within the preset range, without significant fluctuations, abrupt changes, or large variations. The degree of fit between the predicted and actual results remains stable over a long period without significant improvement. When the model output state enters a stable and unchanging stage, it is determined that the updated model fully meets the preset convergence conditions. The updated model that has successfully reached the convergence conditions at this time is officially identified as a load prediction model that can be accurately adapted to and stably applied to the heating network system of the target region.

[0041] Step 3: Input the real-time operation data of the target area heating network system into the load prediction model to generate the heat load demand curve of the target area heating network system; In this embodiment of the invention, in step 3, the real-time operating data of the target area heating network system is input into the load prediction model to generate the heat load demand curve of the target area heating network system. The process is as follows: According to the preset time step, the real-time operation data of the target area heating network system is divided into real-time operation data units corresponding to consecutive time points; Input the real-time operational data units corresponding to different time points into the load prediction model to obtain the predicted heat load value of the target area's heating network system. The predicted heat load values ​​are smoothly arranged in chronological order to obtain the heat load demand curve of the heating network system in the target area.

[0042] Based on a pre-set and fixed time step, all real-time operating data continuously collected from the heating network system in the target area at the current moment is segmented. Data segments are extracted sequentially from the starting moment with the time step as the interval. Each extracted data segment completely contains the operating parameter information of all monitoring points in the network at the corresponding time point, ensuring that adjacent data segments are closely connected in time and without overlap or omission, ultimately forming real-time operating data units corresponding to consecutive time points.

[0043] The divided real-time operational data units are sequentially input into the trained, converged, and determined load prediction model in chronological order from earliest to latest. After entering the model, each real-time operational data unit is passed through and processed by the nodes at each level within the model. Based on the learned correspondence between operational data and heat load, the model analyzes and extrapolates the current input real-time operational data unit, and finally outputs the prediction result corresponding to that time point. The prediction results corresponding to all time points are summarized to obtain the predicted heat load value of the target area's heating network system.

[0044] All the obtained predicted heat load values ​​are arranged in chronological order according to their corresponding time. After the arrangement is completed, the predicted heat load values ​​at adjacent time points are smoothed to keep the changes between adjacent values ​​continuous and smooth, avoid abrupt jumps, and make the overall value change present a stable and coherent state. All the smoothed predicted heat load values ​​are combined together in chronological order to form a heat load demand curve of the target area heating network system that can completely reflect the heat load of the heating network over time.

[0045] Step 4: Based on the heat load demand curve, perform multi-objective decision optimization on the current operating status data of the heating network system in the target area to obtain the Pareto optimal solution set of the heating network system in the target area; In this embodiment of the invention, in step 4, based on the heat load demand curve, multi-objective decision optimization is performed on the current operating status data of the target area heating network system to obtain the Pareto optimal solution set of the target area heating network system. The process is as follows: The heating energy consumption of the target area's heating network system, the rate of compliance of user-end room temperature, and the stability of network operation are taken as the optimization target set. Determine the operational constraints of the heating network system in the target area. These constraints include the upper limit of network pressure, the range of water supply temperature regulation, and the limit on the rate of flow change. Based on the heat load demand curve and the current operating status data of the target area's heating network system, a candidate heating scheduling parameter group for the target area's heating network system is generated. Based on the optimization target set, the expected performance of the candidate heating scheduling parameter group is evaluated, and the target performance vector of the candidate heating scheduling parameter group is obtained. The target performance vector is compared pairwise internally to select the Pareto optimal solution set for the target regional heating network system from the candidate heating scheduling parameter set.

[0046] The process of selecting the Pareto optimal solution set for the target area's heating network system from the candidate heating scheduling parameter set is as follows: Obtain the target performance vector of the candidate heating scheduling parameter group, which contains the expected performance values ​​of the candidate heating scheduling parameter group on the three optimization objectives; The target performance vectors of the candidate heating scheduling parameter groups are compared with the expected performance values ​​of the target performance vectors of other candidate heating scheduling parameter groups. If, for a candidate heating scheduling parameter group, there is no other candidate heating scheduling parameter group whose target performance vector is better than the target performance vector of the candidate heating scheduling parameter group in all three optimization objectives, then the candidate heating scheduling parameter group is determined as the Pareto optimal candidate solution. The candidate heating scheduling parameter sets that will be identified as Pareto optimal candidate schemes will be collected to form the Pareto optimal solution set of the target area heating network system.

[0047] The total heating energy consumption, the proportion of all users whose indoor temperatures are stable at the set standard, and the overall performance of the heating network system in the target area during actual scheduling and operation are uniformly integrated and standardized and classified to form an optimization target set for measuring the quality of scheduling schemes and guiding the direction of optimization.

[0048] Based on the equipment design limits and safety operation requirements of the target area's heating network and industry standards, each boundary restriction that the network must strictly adhere to during actual operation is clearly defined and locked. The maximum upper limit of the internal pressure of the network, the legal adjustment range formed by the minimum and maximum limits of the adjustable water supply temperature, and the maximum rate of change of the medium flow rate within the network per unit time are all determined as insurmountable operational constraints of the target area's heating network system.

[0049] By combining the existing and readily available heat load demand curves, which reflect the magnitude and trend of heating demand at different time points, and matching all current operating status data of the target region's heating network system, including pressure, temperature, flow rate, equipment status, valve opening, and circulating pump operating status, and under the premise of fully meeting and not exceeding any operating constraints, multiple sets of candidate heating scheduling parameters for the target region's heating network system are generated through a reasonable combination of different adjustment methods and control strategies. These parameters are practically executable and can be used for subsequent evaluation and screening.

[0050] Using the three core optimization directions of minimizing heating energy consumption, maximizing the rate of achieving the target room temperature at the user end, and optimizing the stability of pipeline operation as the complete evaluation criteria, a comprehensive deduction and quantitative judgment are made on the expected effects that each individual candidate heating scheduling parameter group can achieve after being put into actual operation. The specific expected performance values ​​of the parameter group in the three dimensions of heating energy consumption control, user end room temperature achievement, and pipeline operation stability are obtained respectively. The expected performance values ​​of these three dimensions are combined in a fixed order to form a target performance vector of the candidate heating scheduling parameter group that can fully characterize the comprehensive performance of the parameter group.

[0051] From the results obtained after the expected evaluation of each group of candidate heating scheduling parameters, the specific expected performance values ​​of the parameters in terms of the three optimization objectives of heating energy consumption, user-end room temperature compliance rate, and network operation stability are completely extracted. These three independent yet interconnected expected performance values ​​together constitute the target performance vector of the candidate heating scheduling parameter group that corresponds to the parameter group.

[0052] Select any one set from all candidate heating scheduling parameter sets as the current comparison object, and compare the expected performance values ​​of the three optimization targets contained in its target performance vector with the expected performance values ​​of the three targets in the target performance vectors of all other candidate heating scheduling parameter sets item by item and dimension by dimension. Completely record the differences in the advantages and disadvantages of each optimization target in each comparison, and form a complete comparison result covering all combinations.

[0053] For the currently selected candidate heating scheduling parameter group, a full-domain investigation is carried out based on all comparison results. Each candidate heating scheduling parameter group is checked to see if there is any other candidate heating scheduling parameter group whose target performance vector is better than the target performance vector of the current candidate heating scheduling parameter group in all three optimization objectives: heating energy consumption performance, user-end room temperature compliance rate performance, and pipeline operation stability performance. If, after a complete traversal and verification, it is confirmed that there is no other parameter group that meets this condition, the current candidate heating scheduling parameter group is immediately officially determined and identified as the Pareto optimal candidate scheme.

[0054] All candidate heating scheduling parameter groups corresponding to the Pareto optimal candidate schemes that have been successfully identified through the above comparison and verification process are uniformly collected and organized, and arranged in an orderly manner according to fixed rules to ensure that each optimal scheme is completely retained without duplication or omission. All selected schemes are combined to form the Pareto optimal solution set of the target area heating network system that can be directly used for heating scheduling decisions.

[0055] Step 5: Based on the dynamic hydraulic balance constraints of the target region's heating network system, prioritize the Pareto optimal solution set to obtain candidate scheduling schemes for the target region's heating network system. In this embodiment of the invention, in step 5, the Pareto optimal solution set is prioritized to obtain candidate scheduling schemes for the target area heating network system. The process is as follows: Obtain the pipeline hydraulic balance simulation data corresponding to the candidate heating scheduling parameter group in the Pareto optimal solution set, and compare the pipeline hydraulic balance simulation data with the preset dynamic hydraulic balance constraint threshold. Based on the comparison results, candidate heating scheduling parameter groups whose pipeline hydraulic balance simulation data are all within the dynamic hydraulic balance constraint threshold range are determined as qualified hydraulic balance schemes. According to the preset scheduling priority rules, the qualified hydraulic balance schemes are prioritized and the highest-ranked qualified hydraulic balance scheme is determined as the candidate scheduling scheme for the heating network system in the target area.

[0056] Each set of candidate heating scheduling parameters is sequentially extracted from the Pareto optimal solution set. Based on all scheduling-related information included in the set parameters, such as the set value of water supply flow, the adjustment value of water supply temperature, the opening degree of valves in each branch network, the operating frequency of circulating pumps, and the pressure control value of each node, a full-area hydraulic state simulation is performed on the heating network system in the target area. During the simulation, the pressure distribution, flow distribution, pressure difference change, and resistance matching are calculated node by node and pipe segment by pipe segment. The simulation calculation results of all nodes and pipe segments are integrated and summarized to form the network hydraulic balance simulation data that can completely reflect the overall hydraulic state of the network under the set parameters. Then, each hydraulic index in the network hydraulic balance simulation data is precisely compared with the dynamic hydraulic balance constraint threshold set in advance according to the network structure, equipment bearing capacity, safe operation specifications, and real-time adjustment needs, to confirm whether each simulation data is within the reasonable range of the corresponding threshold.

[0057] Based on the results of comparing the simulated hydraulic balance data of the pipeline network with the dynamic hydraulic balance constraint thresholds item by item, a strict qualification judgment is carried out on each group of candidate heating scheduling parameters. All pressure, flow, differential pressure and resistance indicators in the simulated hydraulic balance data of the pipeline network corresponding to the parameters are checked in full coverage. Only when all hydraulic indicators are within the upper and lower limits of the dynamic hydraulic balance constraint thresholds, and no indicator exceeds the threshold or deviates from the reasonable range, and the entire pipeline network maintains a stable and balanced hydraulic state in the simulation, is the group of candidate heating scheduling parameters officially recognized as a qualified hydraulic balance scheme.

[0058] According to the scheduling priority rules formulated in advance based on the actual heating guarantee tasks, energy consumption control targets, user indoor comfort requirements, and pipeline operation safety levels, all qualified hydraulic balance schemes that have passed the judgment are systematically prioritized. In the ranking, heating energy consumption performance is the primary basis, with schemes with lower energy consumption having higher priority. If the energy consumption is the same, the user-end room temperature compliance rate is the secondary basis, with schemes with higher compliance rates ranking higher. If the first two indicators are the same, the pipeline operation stability is the final basis, with schemes with better stability ranking higher. After a complete ranking layer by layer, the qualified hydraulic balance scheme with the highest priority at the top of the ranking is finally determined as the candidate scheduling scheme for the target area's heating pipeline system.

[0059] Step 6: Distribute the candidate scheduling scheme to the execution agency of the heating network system in the target area, and collect the feedback operation data of the execution agency in real time to update the load prediction model.

[0060] In this embodiment of the invention, in step 6, the candidate scheduling scheme is sent to the execution mechanism of the target area heating network system, and the feedback operation data of the execution mechanism is collected in real time to update the load prediction model. The process is as follows: The candidate scheduling schemes are formatted and encoded to obtain the scheduling instruction sequence of the target area heating network system; The sequence of dispatch instructions is sent to the actuators of the heating network system in the target area to control the actuators to operate according to the candidate dispatch scheme; Real-time acquisition of feedback operational data generated by the actuator during operation, including actual operating parameters of the heating network system in the target area; The feedback operation data is compared with the expected operation parameters in the candidate scheduling scheme to determine the deviation trend between the feedback operation data and the expected operation parameters; Based on the deviation trend, the internal parameters of the load forecasting model are dynamically adjusted to update the load forecasting model.

[0061] All scheduling and control information included in the finalized candidate scheduling scheme, such as water supply temperature settings for each time period, water supply flow regulation for each area, valve opening adjustment for each branch pipeline, start / stop control and operation status regulation of each circulating pump, and pressure control for each key pipeline node, is converted and standardized item by item according to a unified and fixed format that can be fully identified and unambiguously resolved by the on-site actuators of the heating pipeline network. Each item of scheduling and control information is converted into a standard instruction unit with a clear execution subject, clear execution action, and clear execution requirements. Then, according to the execution time sequence of the scheduling instructions, the area to which the controlled object belongs, and the linkage logic of the adjustment action, all standard instruction units are arranged and combined in an orderly manner to form a scheduling instruction sequence of the target area heating pipeline network system that is structurally complete, has a clear time sequence, and can be directly transmitted and executed downwards.

[0062] The sequence of scheduling instructions, verified for both format and logic, is transmitted via a dedicated and stable communication channel. Each instruction is sent precisely and sequentially to all corresponding field actuators within the target region's heating network system. These actuators include main heating network circulating pumps, branch network regulating valves, heat exchange station temperature control devices, network pressure regulating equipment, distributed pump station control modules, and user-end control units. This ensures that each scheduling instruction is accurately delivered to the designated actuator, enabling all actuators to respond synchronously upon receiving the instruction and strictly adhere to the operating parameters, adjustment range, timing, and duration specified in the candidate scheduling scheme to perform corresponding adjustment operations, maintaining the overall stable operation of the network according to the plan.

[0063] Throughout the entire operating cycle in which all actuators continuously execute adjustment operations according to the sequence of dispatch instructions, various data acquisition terminals installed at the main pipe nodes, branch pipe nodes, heat exchange station outlets, user inlets, and various actuator bodies of the heating network continuously, frequently, and synchronously collect all feedback operating data generated by the actuators during adjustment and operation in real time. This feedback operating data fully includes all actual operating parameters that can reflect the true operating status, such as real-time supply water pressure, real-time return water pressure, real-time supply water temperature, real-time return water temperature, real-time network flow, actual opening values ​​of each valve, actual operating speed and current of each circulating pump, and actual indoor temperature at the user end in each area.

[0064] The actual operating parameters in each piece of feedback operating data acquired in real time are compared precisely item by item, point by point, and time by time with the expected operating parameters pre-set in the corresponding time period and location in the candidate scheduling scheme. The actual parameters and expected parameters of each pair are continuously tracked and compared to determine whether the actual operating parameters are consistently higher than expected, consistently lower than expected, or exhibit intermittent fluctuations and deviations relative to the expected operating parameters. At the same time, the rate of change, magnitude, and duration of deviation are determined to completely and accurately determine the overall deviation trend between the feedback operating data and the expected operating parameters.

[0065] Based on the established deviation trend of the feedback operation data from the expected operation parameters, and considering the direction, magnitude, and duration of the difference between the actual operation status and the expected status of the plan, targeted dynamic adjustments are made to the data feature extraction method, parameter correlation judgment rules, intermediate level information transmission method, and final result output correction logic within the load forecasting model. This allows the model to better align with the actual operating characteristics of the target region's heating network system when processing real-time operation data and generating forecast results, gradually reducing the deviation between the forecast output and the actual results. Through this continuous correction method based on actual operation feedback, the load forecasting model is updated and optimized.

[0066] As can be seen from the above embodiments, the intelligent heating integrated scheduling method based on operational data analysis provided by this invention lays a reliable data foundation for load forecasting by standardizing and preprocessing historical operational data. Combined with scientific training and dynamic parameter updates of a neural network model, it significantly improves the accuracy of heat load forecasting and accurately generates heat load demand curves. Based on multi-objective decision optimization and combined with Pareto optimal solution set screening, it comprehensively considers heating energy consumption, room temperature compliance rate, and network stability. Simultaneously, it relies on dynamic hydraulic balance constraints for priority screening, making the scheduling scheme more scientific, reasonable, and aligned with actual heating needs. After scheduling execution, feedback data is collected in real time and the load forecasting model is updated, forming a closed-loop management of the entire scheduling process, continuously optimizing model adaptability and scheduling strategies. Overall, it improves the efficiency of intelligent heating integrated scheduling, effectively balancing heating effect, energy consumption control, and network operational stability, reducing energy waste, ensuring user heating experience, and making the heating network system operate more efficiently, intelligently, and stably.

[0067] like Figure 2 The diagram shown is a functional block diagram of a smart heating integrated scheduling system 100 based on operational data analysis provided in an embodiment of the present invention. It includes a data preprocessing module 101, a prediction model training module 102, a real-time load prediction module 103, a multi-objective optimization solution module 104, a scheduling scheme screening module 105, and a scheduling execution module 106.

[0068] In this embodiment, the functions of each module are as follows: The data preprocessing module 101 is used to preprocess the historical operation dataset of the monitoring points in the target area heating network system during the same historical period to obtain the standard historical operation data of the target area heating network system. The prediction model training module 102 is used to input standard historical operating data and corresponding actual heat load values ​​into a preset initial neural network model for iterative training in order to construct a load prediction model for the target area heating network system. The real-time load prediction module 103 is used to input the real-time operation data of the target area heating network system into the load prediction model to generate the heat load demand curve of the target area heating network system. The multi-objective optimization solution module 104 is used to perform multi-objective decision optimization on the current operating status data of the heating network system in the target area based on the heat load demand curve, so as to obtain the Pareto optimal solution set of the heating network system in the target area. The scheduling scheme screening module 105 is used to prioritize the Pareto optimal solution set based on the dynamic hydraulic balance constraints of the target area heating network system to obtain candidate scheduling schemes for the target area heating network system. The scheduling execution module 106 is used to send candidate scheduling schemes to the execution agencies of the target area heating network system and collect feedback operation data from the execution agencies in real time to update the load prediction model.

[0069] As can be seen from the above embodiments, the intelligent heating integrated scheduling system based on operational data analysis provided by this invention lays a reliable data foundation for load forecasting by standardizing and preprocessing historical operational data. Combined with scientific training and dynamic parameter updates of a neural network model, it significantly improves the accuracy of heat load forecasting and accurately generates heat load demand curves. Based on multi-objective decision optimization and combined with Pareto optimal solution set screening, it comprehensively considers heating energy consumption, room temperature compliance rate, and network stability. Simultaneously, it relies on dynamic hydraulic balance constraints for priority screening, making the scheduling scheme more scientific, reasonable, and aligned with actual heating needs. After scheduling execution, feedback data is collected in real time and the load forecasting model is updated, forming a closed-loop management of the entire scheduling process, continuously optimizing model adaptability and scheduling strategies. Overall, it improves the efficiency of intelligent heating integrated scheduling, effectively balancing heating effect, energy consumption control, and network operational stability, reducing energy waste, ensuring user heating experience, and making the heating network system operate more efficiently, intelligently, and stably.

[0070] It will be apparent to those skilled in the art that the present invention is not limited to the details of the exemplary embodiments described above, and that the present invention can be implemented in other specific forms without departing from the spirit or essential characteristics of the present invention.

[0071] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and are not intended to limit it. Although the present invention has been described in detail with reference to preferred embodiments, those skilled in the art should understand that modifications or equivalent substitutions can be made to the technical solutions of the present invention without departing from the spirit and scope of the technical solutions of the present invention.

Claims

1. A smart heating integrated scheduling method based on operational data analysis, characterized in that, The method includes: Step 1: Perform data preprocessing on the historical operation dataset of the monitoring points in the target area's heating network system during the same historical period to obtain the standard historical operation data of the target area's heating network system; Step 2: Input the standard historical operating data and the corresponding actual heat load values ​​into the preset initial neural network model for iterative training to build a load prediction model for the target area heating network system; Step 3: Input the real-time operation data of the target area heating network system into the load prediction model to generate the heat load demand curve of the target area heating network system; Step 4: Based on the heat load demand curve, perform multi-objective decision optimization on the current operating status data of the heating network system in the target area to obtain the Pareto optimal solution set of the heating network system in the target area; Step 5: Based on the dynamic hydraulic balance constraints of the target region's heating network system, prioritize the Pareto optimal solution set to obtain candidate scheduling schemes for the target region's heating network system. Step 6: Distribute the candidate scheduling scheme to the execution agency of the heating network system in the target area, and collect the feedback operation data of the execution agency in real time to update the load prediction model.

2. The intelligent heating integrated scheduling method based on operational data analysis as described in claim 1, characterized in that, In step 1, standard historical operating data of the target area's heating network system are obtained, and the process is as follows: Obtain historical operational datasets of monitoring points in the heating network system of the target region during the same historical time period; By removing outliers from the historical operation dataset, we obtain the cleaned historical operation data of the heating network system in the target region. The historical operating data after cleaning is transformed to a unified dimension to obtain the unified dimension historical operating data of the heating network system in the target area. By aligning the dimension-unified historical operation data with the time dimension, the time-aligned historical operation data of the heating network system in the target region is obtained. Missing time periods are filled in the time-series aligned historical operation data to obtain the standard historical operation data of the heating network system in the target area.

3. The intelligent heating integrated scheduling method based on operational data analysis as described in claim 1, characterized in that, In step 2, a load prediction model for the target area's heating network system is constructed, and the process is as follows: The standard historical operating data is divided into the training input dataset, and the actual heat load values ​​corresponding to the standard historical operating data are divided into the training target dataset. The training input dataset is input into a preset initial neural network model. Through the forward propagation process of the initial neural network model, the predicted heat load value corresponding to the training input dataset is obtained. The predicted heat load value is compared with the corresponding actual heat load value in the training target dataset, and the internal parameters of the initial neural network model are adjusted based on the comparison results. The adjusted neural network model is used as the updated model, and the training process is repeated until the updated model meets the preset convergence condition. The updated model that meets the convergence criteria is identified as the load prediction model for the target region's heating network system.

4. The intelligent heating integrated scheduling method based on operational data analysis as described in claim 3, characterized in that, Based on the comparison results, the internal parameters of the initial neural network model are adjusted as follows: The comparison results are extracted to show the error between the predicted heat load value and the corresponding actual heat load value of the initial neural network model under the current iteration training, as well as the error value under the previous iteration training. Obtain the output values ​​of the hidden layer nodes in the initial neural network model under the current iteration of training, as well as the adjustment amount of the connection weights between the hidden layer nodes and the output layer nodes in the previous iteration; The adjustment amount of the connection weights between hidden layer nodes and output layer nodes in the initial neural network model is calculated under the current training iteration. The formula for calculating the adjustment amount under the current training iteration is as follows: ； In the formula, This represents the adjustment amount of the connection weights between hidden layer nodes and output layer nodes in the initial neural network model under the current iteration of training. This represents the preset base learning rate. This represents the error between the predicted heat load value and the corresponding actual heat load value under the current iteration of training. This represents the error between the predicted heat load value and the corresponding actual heat load value from the previous training iteration. Represents the hidden layer nodes in the initial neural network model The output value in the current training iteration. This represents the preset momentum coefficient. Represents the hidden layer nodes in the previous iteration To the output layer node The adjustment amount of the connection weights between them. This indicates a preset trend factor; The connection weights of the hierarchical nodes in the initial neural network model are dynamically updated based on the adjustment of the connection weights between the hidden layer nodes and the output layer nodes in the current iteration of training.

5. The intelligent heating integrated scheduling method based on operational data analysis as described in claim 1, characterized in that, In step 3, the real-time operating data of the target area's heating network system is input into the load prediction model to generate the heat load demand curve of the target area's heating network system. The process is as follows: According to the preset time step, the real-time operation data of the target area heating network system is divided into real-time operation data units corresponding to consecutive time points; Input the real-time operational data units corresponding to different time points into the load prediction model to obtain the predicted heat load value of the target area's heating network system. The predicted heat load values ​​are smoothly arranged in chronological order to obtain the heat load demand curve of the heating network system in the target area.

6. The intelligent heating integrated scheduling method based on operational data analysis as described in claim 1, characterized in that, In step 4, based on the heat load demand curve, multi-objective decision optimization is performed on the current operating status data of the target area's heating network system to obtain the Pareto optimal solution set of the target area's heating network system. The process is as follows: The heating energy consumption of the target area's heating network system, the rate of compliance of user-end room temperature, and the stability of network operation are taken as the optimization target set. Determine the operational constraints of the heating network system in the target area. These constraints include the upper limit of network pressure, the range of water supply temperature regulation, and the limit on the rate of flow change. Based on the heat load demand curve and the current operating status data of the target area's heating network system, a candidate heating scheduling parameter group for the target area's heating network system is generated. Based on the optimization target set, the expected performance of the candidate heating scheduling parameter group is evaluated to obtain the target performance vector of the candidate heating scheduling parameter group. The target performance vector is compared pairwise internally to select the Pareto optimal solution set for the target regional heating network system from the candidate heating scheduling parameter set.

7. The intelligent heating integrated scheduling method based on operational data analysis as described in claim 6, characterized in that, The process of selecting the Pareto optimal solution set for the target area's heating network system from the candidate heating scheduling parameter set is as follows: Obtain the target performance vector of the candidate heating scheduling parameter group, which contains the expected performance values ​​of the candidate heating scheduling parameter group on the three optimization objectives; The target performance vectors of the candidate heating scheduling parameter groups are compared with the expected performance values ​​of the target performance vectors of other candidate heating scheduling parameter groups. If, for a candidate heating scheduling parameter group, there is no other candidate heating scheduling parameter group whose target performance vector is better than the target performance vector of the candidate heating scheduling parameter group in all three optimization objectives, then the candidate heating scheduling parameter group is determined as the Pareto optimal candidate solution. The candidate heating scheduling parameter sets that will be identified as Pareto optimal candidate schemes will be collected to form the Pareto optimal solution set of the target area heating network system.

8. The intelligent heating integrated scheduling method based on operational data analysis as described in claim 1, characterized in that, In step 5, the Pareto optimal solution set is prioritized to obtain candidate scheduling schemes for the target area's heating network system. The process is as follows: Obtain the pipeline hydraulic balance simulation data corresponding to the candidate heating scheduling parameter group in the Pareto optimal solution set, and compare the pipeline hydraulic balance simulation data with the preset dynamic hydraulic balance constraint threshold. Based on the comparison results, candidate heating scheduling parameter groups whose pipeline hydraulic balance simulation data are all within the dynamic hydraulic balance constraint threshold range are determined as qualified hydraulic balance schemes. According to the preset scheduling priority rules, the qualified hydraulic balance schemes are prioritized and the highest-ranked qualified hydraulic balance scheme is determined as the candidate scheduling scheme for the heating network system in the target area.

9. The intelligent heating integrated scheduling method based on operational data analysis as described in claim 1, characterized in that, In step 6, the candidate scheduling scheme is sent to the execution agency of the target area heating network system, and the feedback operation data of the execution agency is collected in real time to update the load prediction model. The process is as follows: The candidate scheduling schemes are formatted and encoded to obtain the scheduling instruction sequence of the target area heating network system; The sequence of dispatch instructions is sent to the actuators of the heating network system in the target area to control the actuators to operate according to the candidate dispatch scheme; Real-time acquisition of feedback operational data generated by the actuator during operation, including actual operating parameters of the heating network system in the target area; The feedback operation data is compared with the expected operation parameters in the candidate scheduling scheme to determine the deviation trend between the feedback operation data and the expected operation parameters; Based on the deviation trend, the internal parameters of the load forecasting model are dynamically adjusted to update the load forecasting model.

10. A smart heating integrated dispatching system based on operational data analysis, characterized in that, The system, used to implement the intelligent heating integrated scheduling method based on operational data analysis as described in claims 1-9, comprises: The data preprocessing module is used to preprocess the historical operation dataset of the monitoring points in the heating network system of the target region within the same historical time period to obtain the standard historical operation data of the heating network system of the target region. The prediction model training module is used to input standard historical operating data and corresponding actual heat load values ​​into a preset initial neural network model for iterative training in order to build a load prediction model for the target region's heating network system. The real-time load forecasting module is used to input the real-time operating data of the heating network system in the target area into the load forecasting model and generate the heat load demand curve of the heating network system in the target area. The multi-objective optimization solution module is used to perform multi-objective decision optimization on the current operating status data of the heating network system in the target area based on the heat load demand curve, and obtain the Pareto optimal solution set of the heating network system in the target area. The scheduling scheme screening module is used to prioritize the Pareto optimal solution set based on the dynamic hydraulic balance constraints of the target area heating network system, and obtain candidate scheduling schemes for the target area heating network system. The scheduling and execution module is used to distribute candidate scheduling schemes to the execution agencies of the target area heating network system and collect feedback operation data from the execution agencies in real time to update the load prediction model.

Images