A method and system for modeling and simulating an electric heat grid

Abstract

The application relates to the technical field of data processing, in particular to a power and heat network model simulation method and system. The method comprises the following steps: collecting multi-dimensional operation parameters of a power side, a heat side and a coupling unit in real time respectively; constructing a multi-scale observation window according to a fluctuation degree of the operation parameters, quantifying fluctuation significance comprehensively to form a parameter fluctuation sequence, and then clustering different historical periods; constructing a training set and a test set based on multi-dimensional operation parameters in all historical periods in each cluster to obtain a power and heat network model through neural network training, which is used for power and heat network simulation. The application trains the neural network through the construction of the multi-scale observation window and the hierarchical balanced sampling strategy, solves the problem of large simulation deviation caused by incomplete working condition coverage and ignoring time-varying characteristics of the traditional model, and improves the accuracy and stability of power and heat network model simulation.

Description

Technical Field

[0001] This application relates to the field of data processing technology, specifically to a simulation method and system for power heating network models. Background Technology

[0002] The power-heat network model is an integrated model that combines the operational characteristics of power systems and regional heating networks. By analyzing the energy conversion mechanisms of coupled power and heat units, it constructs a comprehensive model covering power generation, heating, energy transmission, and load response. This model comprehensively considers key factors such as generator output, grid flow, and power load in the power subsystem, and heat source supply, hydraulic and thermal characteristics of the pipeline network, and heat load in the thermal subsystem. This enables precise analysis of the energy conversion, transmission, and distribution processes of the two systems. Based on this, the model can effectively support system simulation, facilitate optimized operation scheduling and efficient emergency response to faults, thereby significantly improving the intelligent operation efficiency of integrated energy systems.

[0003] However, in the construction and application of actual power heating network models, the simulation models often deviate significantly from the actual system operation due to multiple factors such as energy output fluctuations, user load changes, data transmission packet loss, and dynamic changes in pipeline topology. At the same time, traditional modeling methods often ignore the nonlinear and time-varying characteristics of electro-thermal coupling units, making it difficult to accurately capture the thermoelectric conversion law under different operating conditions. This fails to meet the collaborative simulation requirements of the rapid dynamics of the power system and the slow dynamics of the heating network, resulting in a large difference between the model simulation results and the actual operating state characteristics, which seriously reduces the accuracy and stability of power heating network model simulation. Summary of the Invention

[0004] To address the aforementioned technical problems, the purpose of this application is to provide a simulation method and system for power heating network models. The specific technical solution adopted is as follows: In a first aspect, embodiments of this application provide a simulation method for a power heating network model, the method comprising the following steps: Real-time acquisition of multi-dimensional operating parameters from the power side, thermal side, and coupling unit; Based on the degree of fluctuation of each type of operating parameter within the historical cycle, a minimum time unit that can characterize its fluctuation characteristics is set as the basic observation window for each type of operating parameter. On the basis of the basic observation window, multi-scale observation windows are constructed by expanding by integer multiples until the entire historical cycle is covered. By combining the dispersion of each type of operating parameter under each observation window, and the correlation between each type of operating parameter and any other type of operating parameter, the fluctuation significance of each type of operating parameter under each observation window is quantified, which is used to reflect the fluctuation intensity of each type of operating parameter within a specific time window and the degree of coupling with other parameters. Based on the mean significance of the fluctuations of each type of operating parameter across all observation windows, a parameter fluctuation sequence corresponding to the historical period is constructed for clustering different historical periods; Training and testing sets are constructed based on the multidimensional operating parameters of all historical periods in each cluster to train the neural network and obtain the power heating network model for power heating network simulation.

[0005] Preferably, the multi-dimensional operating parameters on the power side include the active and reactive power of the generator, node voltage, line transmission power, and power load; the multi-dimensional operating parameters on the heat side include the heat supplied by the heat source, the supply and return water temperatures of the pipeline nodes, pipeline flow rate, pipeline pressure, and heat load demand; the multi-dimensional operating parameters of the coupling unit include the input and output power of the generator, the input and output power of the electric boiler, and the operating power of the heat pump.

[0006] Preferably, the method for setting the basic observation window for each type of operating parameter is as follows: The result of subtracting the normalized value of the fluctuation of each operating parameter in the historical period from 1 is denoted as the weighting factor. The product of the number of each type of operating parameter in the historical period and the weight factor is used as the input to the rounding function, and the output value is used as the size of the basic observation window for each type of operating parameter. That is, the basic observation window needs to contain the number of data points of the output value.

[0007] Preferably, the method for constructing the multi-scale observation window is as follows: If it exists Then the observation window sizes for the x-th type of running parameters at multiple scales are as follows: Conversely, if ,but Then the observation window sizes for the x-th type of running parameters at multiple scales are as follows: ,in, Represents positive integers. This represents the size of the basic observation window for the x-th type of operating parameter. This indicates the number of x-th type of operating parameters within a historical period.

[0008] Preferably, the significance of the fluctuation of each type of operating parameter under each observation window is positively correlated with the degree of dispersion of each type of operating parameter under each observation window, and negatively correlated with the correlation between each type of operating parameter and any other type of operating parameter.

[0009] Preferably, constructing the parameter fluctuation sequence corresponding to the historical period includes: The average fluctuation significance values ​​of all types of operating parameters on the power side, the average fluctuation significance values ​​of all types of operating parameters on the heat side, and the average fluctuation significance values ​​of all types of operating parameters of the coupling unit within the historical period are respectively formed into a first sequence, a second sequence, and a third sequence; The sequence obtained by splicing the first, second, and third sequences is used as the parameter fluctuation sequence for the corresponding historical period.

[0010] Preferably, the distance metric used in the clustering process is the Euclidean distance between parameter fluctuation sequences of different historical periods.

[0011] Preferably, the construction of training and testing sets based on multidimensional operating parameters across all historical periods in each cluster includes: All class running parameters at each time point are combined into a running parameter sequence; the number of all running parameter sequences in all historical periods in each cluster is counted and recorded as the total number of running parameter sequences in each cluster; the proportion of the total number of running parameter sequences in each cluster in the sum of the total number of running parameter sequences in all clusters is calculated and multiplied by the preset total number of samples, which is used as the number of samples to be selected from each cluster. Select a corresponding number of running parameter sequences as samples from each cluster, traverse all clusters to obtain all samples, and divide all samples into training set and test set according to a preset ratio.

[0012] Preferably, the process of training the neural network to obtain the power heating network model includes: Multidimensional key indicators of the power heating network are acquired in real time. All key indicators at each time point are combined into a key indicator sequence. The training set and test set are used as the learning samples and validation samples of the neural network, respectively. The key indicator sequence is used as the output label of the neural network to train the neural network. The trained neural network model is used as the power heating network model.

[0013] Secondly, embodiments of this application also provide a power heating network model simulation system, including a memory, a processor, and a computer program stored in the memory and running on the processor. When the processor executes the computer program, it implements the steps of any of the above-described power heating network model simulation methods.

[0014] This application has at least the following beneficial effects: This application sets a basic observation window based on the fluctuation of operating parameters within a historical period, and constructs a multi-scale observation window covering the entire period based on this, achieving precise matching of the response sensitivity and time characteristics of different operating parameters. This method not only uses a small-scale window to capture the instantaneous high-frequency response of operating parameters to changes in operating conditions, but also reveals slow dynamic trends such as thermodynamic inertia through a large-scale window, forming a complete analytical framework from micro-fluidities to macro-evolution. This effectively avoids feature loss or noise interference caused by single-time-scale analysis, and provides a high-resolution time-series data foundation for accurately extracting the deep coupling characteristics of power and heat under complex operating conditions. Furthermore, this application quantifies the significance of fluctuations by comprehensively analyzing the dispersion of operating parameters and the correlation between parameters within the observation window. This index is positively correlated with the dispersion and negatively correlated with the correlation, which can effectively identify the drastic fluctuations of operating parameters and the decoupling of coupling relationships at a specific time scale, thereby achieving in-depth capture of the complex operating characteristics of the power-heat network system. Furthermore, this application constructs a parameter fluctuation sequence that integrates the characteristics of the power side, thermal side, and coupled units by calculating the mean significance of fluctuations of various operating parameters under multi-scale observation windows. It also uses a condensed hierarchical clustering algorithm based on Euclidean distance to accurately divide historical cycles, thereby realizing the automatic identification and classification of complex operating conditions. Based on this, a hierarchical balanced sampling strategy is adopted to select training and test sets that cover the characteristics of all operating conditions according to the proportion of each cluster. This effectively solves the problem of small sample overwhelmed by uneven distribution of operating conditions in traditional sampling, ensuring the comprehensiveness and representativeness of the model training data. Finally, based on the clustering results, this application adopts a hierarchical balanced sampling strategy to select training and testing sets that cover the characteristics of all operating conditions according to the proportion of each cluster. It also integrates multi-dimensional key indicators of the power side, thermal side and coupling units as output labels of the neural network to train the neural network and build a power heating network model that can accurately map complex nonlinear relationships. By introducing a feedback mechanism of actual operating data, the dataset is continuously updated and the model is iteratively optimized, thereby ensuring the high-precision simulation and stability of the power heating network model under all operating conditions. Attached Figure Description

[0015] To more clearly illustrate the technical solutions and advantages in the embodiments of this application or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of this application. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0016] Figure 1 A flowchart illustrating the steps of a power heating network model simulation method provided in one embodiment of this application; Figure 2 A flowchart of historical periodic clustering steps provided for one embodiment of this application. Detailed Implementation

[0017] To further illustrate the technical means and effects adopted by this application to achieve the intended purpose of the invention, the following, in conjunction with the accompanying drawings and preferred embodiments, details the specific implementation, structure, features, and effects of a power heating network model simulation method and system proposed in this application. In the following description, different "one embodiment" or "another embodiment" do not necessarily refer to the same embodiment. Furthermore, specific features, structures, or characteristics in one or more embodiments can be combined in any suitable form.

[0018] Unless otherwise defined, all technical and scientific terms used herein have the same meaning as commonly understood by one of ordinary skill in the art to which this application pertains.

[0019] The following description, in conjunction with the accompanying drawings, details the specific scheme of the power heating network model simulation method and system provided in this application.

[0020] Please see Figure 1 The diagram illustrates a flowchart of a power heating network model simulation method according to an embodiment of this application. The method includes the following steps: Step S1: Collect multi-dimensional operating parameters of the power side, thermal side and coupling unit in real time.

[0021] In this embodiment, to construct the power-heat network model, multi-dimensional operating parameters are collected in real time through a SCADA system for the power subsystem, the heat subsystem, and the coupling unit. This includes power-side data, heat-side data, and coupling unit data. The power-side multi-dimensional operating parameters include active and reactive power of the generating unit, node voltage, line transmission power, and power load. The heat-side multi-dimensional operating parameters include heat source heating capacity, supply and return water temperatures at network nodes, pipeline flow rate, pipeline pressure, and heat load demand. The coupling unit multi-dimensional operating parameters include the unit's input and output power, the electric boiler's input and output power, and the heat pump's operating power. In this embodiment, the data acquisition frequency is set to 5 minutes per acquisition. In practical applications, as other implementation methods, the implementer can set the data acquisition frequency according to specific circumstances; this embodiment does not impose any special limitations.

[0022] The process of using a SCADA system to acquire data from power, thermal, and coupling units is a well-known technology and will not be described in detail here.

[0023] Data acquisition and transmission are susceptible to noise interference, which can reduce data quality. Therefore, preprocessing is performed on the acquired data. Specifically, the 3σ criterion is used to identify and remove abnormal data, and linear interpolation is used to fill in missing data. Furthermore, in order to convert parameters of different dimensions and orders of magnitude to a unified scale for subsequent analysis, the maximum-minimum normalization method is used to normalize the above data respectively.

[0024] The processes of identifying and removing abnormal data from the collected data using the 3σ criterion, interpolating the data using linear interpolation, and normalizing the data using the maximum-minimum normalization method are all well-known techniques and will not be elaborated further.

[0025] Step S2: Construct a multi-scale observation window based on the degree of fluctuation of the operating parameters, comprehensively quantify the significance of the fluctuations to form a parameter fluctuation sequence, and then cluster different historical cycles.

[0026] In the operation of an electric-thermal system, changes on the power and thermal sides are interconnected, and the operating state of the coupled unit directly affects the operation of both systems. However, due to the complex and variable actual operating environment, the correlation and influence between the power side, thermal side, and coupled unit vary significantly under different operating conditions. Therefore, to address the impact of complex operating conditions, this study analyzes the fluctuation of operating parameters during historical operation to construct a multi-scale observation window, comprehensively quantifies the significance of these fluctuations to form a parameter fluctuation sequence, and then clusters different historical periods to effectively extract operating state data features reflecting the characteristics of complex operating conditions. This allows for the accurate capture of the correlation and influence patterns of the electric-thermal-coupled unit under conditions such as extreme weather, load changes, equipment coordinated switching, and fluctuations in renewable energy output. It clarifies the mechanism by which changes in operating conditions affect the system's operating state, providing core feature data that reflects the characteristics of all operating conditions for the construction of the electric-thermal network model. This avoids reduced model simulation accuracy due to complex and variable actual operating scenarios. The specific process is as follows: S2.1: Based on the degree of fluctuation of each type of operating parameter within the historical cycle, set the smallest time unit that can characterize its fluctuation characteristics as the basic observation window for each type of operating parameter, and build a multi-scale observation window by expanding it through integer multiples until it covers the entire historical cycle.

[0027] Considering that different operating parameters exhibit varying sensitivities to complex operating conditions during actual operation, and that their temporal characteristics affected by continuous changes in operating conditions differ, a minimum time unit capable of characterizing the fluctuation characteristics of each type of operating parameter within a historical period is defined as the basic observation window for each type of operating parameter. Furthermore, multi-scale observation windows are constructed by expanding these basic observation windows through integer multiples until they cover the entire historical period. The specific process is as follows: First, subtract the normalized value of the fluctuation of each operating parameter in the historical period from 1, and denote the result as the weighting factor. The product of the number of each type of operating parameter in the historical period and the weight factor is used as the input to the rounding function, and the output value is used as the size of the basic observation window for each type of operating parameter. That is, the basic observation window needs to contain the number of data points of the output value.

[0028] In this embodiment, the historical period is set to 3 days to accurately capture the dynamic coupling characteristics and operating patterns of the power system and the thermal system across multiple time scales. On the one hand, the 3-day span is sufficient to cover the significant state changes of the thermal system caused by large thermal inertia, including the entire process of hydraulic condition adjustments and temperature lag responses caused by fluctuations in user-side heat demand. On the other hand, this length can completely encompass the typical daily load peak-valley variation curves in the power system and the evolution of peak-shaving strategies over multiple consecutive days. By setting 3 days as a complete historical period sample, the continuous behavioral characteristics and short-term fluctuation patterns exhibited by the electric-thermal coupling unit in response to supply and demand balance adjustments under different seasons and weather conditions can be effectively extracted. This avoids the loss of key inertial characteristics due to an excessively short time span or the introduction of excessive redundant interference due to an excessively long time span, providing the most representative data foundation for subsequent fluctuation characteristic analysis and model training. In practical applications, as other implementation methods, implementers can also set their own methods according to specific circumstances. This embodiment does not impose any special restrictions.

[0029] It should be noted that there are many methods to measure the degree of fluctuation of operating parameters. In this embodiment, the variance of each operating parameter in the historical period is used as the degree of fluctuation of each operating parameter in the historical period. In practical applications, as other implementation methods, implementers may also use other methods such as standard deviation or coefficient of variation to measure the degree of data dispersion, depending on the specific circumstances. This embodiment does not impose any special restrictions on the selection of methods to measure the degree of data dispersion.

[0030] It should be noted that there are many methods for normalizing data. In this embodiment, the maximum and minimum value normalization method is used to normalize the fluctuation degree and the result of 1 minus the normalized value of the fluctuation degree, respectively, and map their range to [0,1]. In practical applications, as other implementation methods, implementers may also use other normalization methods according to specific circumstances. This embodiment does not impose any special restrictions.

[0031] Based on the size of the basic observation window, it can be understood that the basic observation window is the smallest analytical unit set for the different sensitivity of operating parameters to changes in operating conditions. It is used to characterize the ability of operating parameters to capture instantaneous fluctuation details, reflecting how short a time scale is needed for operating parameters to highlight their changing characteristics under complex operating conditions. Its calculation logic is to multiply the total amount of data in the historical period by the normalization weight factor of "1 minus the normalization value of fluctuation" and then round up. This calculation is mainly affected by the fluctuation degree of the operating parameter. The greater the fluctuation degree, the smaller the calculated basic observation window size, reflecting that the parameter responds more violently and sensitively to changes in operating conditions, and can capture high-frequency sudden change characteristics through a shorter time window. This helps the model accurately locate the instantaneous disturbance of the system. Conversely, the smaller the fluctuation degree of the operating parameter, the larger the basic observation window, reflecting that the parameter changes slowly and has strong time stability. It requires a larger time span to accumulate meaningful feature changes, which helps to filter out small random noise and focus on the long-term trend of the parameter.

[0032] Furthermore, based on the basic observation window, multi-scale observation windows are constructed by expanding them in integer multiples until they cover the entire historical cycle. Specifically: If it exists Then the observation window sizes for the x-th type of running parameters at multiple scales are as follows: Conversely, if ,but Then the observation window sizes for the x-th type of running parameters at multiple scales are as follows: ,in, Represents positive integers. This represents the size of the basic observation window for the x-th type of operating parameter. This indicates the number of x-th type of operating parameters within a historical period.

[0033] The multi-scale observation window can be understood as a time-series analysis framework that expands hierarchically based on the basic observation window. It is used to characterize the cumulative effect and evolution of operating parameters over different time lengths, reflecting the complete dynamic process of operating parameters from microscopic instantaneous fluctuations to macroscopic trend changes. Its calculation logic is based on the basic observation window and expands gradually in integer multiples (such as 1, 2 to k times) until it covers the total amount of data for the entire historical period. This construction process is mainly affected by both the size of the basic observation window and the total amount of periodic data. The larger the size of the basic observation window, the wider the span of the multi-scale observation window at each level, reflecting that the analysis perspective is more biased towards the medium- and long-term slow dynamic process. Although it may blur high-frequency details, it can effectively reveal the inertial characteristics and periodic patterns of operating parameters. Conversely, the smaller the size of the basic observation window, the more refined and richer the multi-scale observation window is, reflecting that the analysis capability is more biased towards capturing microscopic rapid changes and can respond more sensitively to instantaneous changes in operating conditions, but it may also increase the sensitivity to noise.

[0034] Thus, this embodiment sets a basic observation window based on the fluctuation of operating parameters within a historical cycle, and constructs a multi-scale observation window covering the entire cycle based on this, achieving precise matching of the sensitivity and time characteristics of different operating parameters. This method not only uses a small-scale window to capture the instantaneous high-frequency response of operating parameters to changes in operating conditions, but also reveals slow dynamic trends such as thermodynamic inertia through a large-scale window, forming a complete analytical framework from micro-fluctuations to macro-evolution. This effectively avoids feature loss or noise interference caused by single-time-scale analysis, providing a high-resolution time-series data foundation for accurately extracting the deep coupling characteristics of power and heat under complex operating conditions.

[0035] S2.2: By combining the dispersion of each type of operating parameter under each observation window, and the correlation between each type of operating parameter and any other type of operating parameter, the fluctuation significance of each type of operating parameter under each observation window is quantified, which is used to reflect the fluctuation intensity of each type of operating parameter within a specific time window and the degree of coupling with other parameters.

[0036] Based on the above-mentioned multi-scale observation window division, further, by comprehensively considering the dispersion of each type of operating parameter under each observation window, and the correlation between each type of operating parameter and any other type of operating parameter, the fluctuation significance of each type of operating parameter under each observation window is quantified. This reflects the fluctuation intensity of each type of operating parameter within a specific time window and its coupling with other parameters. Specifically: In this embodiment, the significance of the fluctuation of each type of operating parameter under each observation window is positively correlated with the degree of dispersion of each type of operating parameter under each observation window, and negatively correlated with the correlation between each type of operating parameter and any other type of operating parameter.

[0037] It should be noted that there are many methods to measure the correlation between operating parameters. In this embodiment, the absolute value of the Pearson correlation coefficient between each type of operating parameter and any other type of operating parameter is used as the quantitative method for measuring the correlation between each type of operating parameter and any other type of operating parameter. In practical applications, as other implementation methods, implementers may also use other correlation measurement methods such as Spearman correlation coefficient or Kendall rank correlation coefficient according to specific circumstances. This embodiment does not impose any special restrictions on the selection of correlation measurement methods between operating parameters.

[0038] The calculation method for the Pearson correlation coefficient is a well-known technique, and its specific calculation process will not be elaborated here.

[0039] It should be understood that a positive correlation means that the dependent variable increases as the independent variable increases, and the dependent variable decreases as the independent variable decreases. The specific relationship can be additive or multiplicative, etc., and is determined by the actual application. This application does not impose any special restrictions. A negative correlation means that the dependent variable decreases as the independent variable increases, and the dependent variable increases as the independent variable decreases. The relationship can be subtractive or divisive, etc., and is determined by the actual application.

[0040] Preferably, as one implementation method, in this embodiment, the significance of the fluctuation of the x-th type of operating parameter within the observation window y is... The expression is: In the formula, This indicates the degree of dispersion of the x-th type of operating parameter under the observation window y; This represents the mean absolute value of the Pearson correlation coefficient between the x-th class of operating parameters and all other classes of operating parameters under the observation window y; This represents a preset constant greater than 0, used to prevent the denominator from being 0. In this embodiment... The value of is 0.01. Under the premise of ensuring that the denominator is not 0 and does not excessively affect the calculation result, the implementer can also set it according to the specific situation. This embodiment does not impose any special restrictions.

[0041] It should be noted that there are many methods to measure the dispersion of data. In this embodiment, the variance of each type of operating parameter under each observation window is used as the dispersion of each type of operating parameter under each observation window. In practical applications, as other implementation methods, implementers may also use other methods such as standard deviation or coefficient of variation to measure the dispersion of data, depending on the specific circumstances. This embodiment does not impose any special restrictions on the selection of methods to measure the dispersion of data.

[0042] Based on the significance of fluctuations, it can be understood that the significance of fluctuations characterizes the drastic changes in operating parameters and the stability of their correlation with other parameters. It reflects whether the operating parameter has significant fluctuation-dominant and decoupling characteristics under the current operating conditions. The significance of fluctuations is positively correlated with the dispersion of the operating parameter within the observation window and negatively correlated with the correlation between the operating parameter and all other parameters. This calculation is mainly affected by the dispersion of the operating parameter itself and the correlation between operating parameters. If the dispersion of the current class of operating parameters is greater and the correlation is smaller, the significance of fluctuations is higher, reflecting that the current class of operating parameters exhibits violent and disordered fluctuations within the window and that the normal coupling relationship with other parameters is broken. This often indicates that the power heating network is in extreme operating conditions or on the verge of failure, which requires the model to give higher attention weight to the current class of operating parameters. Conversely, if the dispersion of the current class of operating parameters is smaller and the correlation is greater, the significance of fluctuations is lower, reflecting that the current class of operating parameters is operating stably and is in a normal coupled operating state. The model can use the current class of operating parameters as background parameters for auxiliary analysis.

[0043] Thus, this embodiment quantifies the significance of fluctuations by comprehensively analyzing the dispersion of operating parameters and the correlation between parameters within the observation window. This index is positively correlated with the dispersion and negatively correlated with the correlation, which can effectively identify the drastic fluctuations of operating parameters and the decoupling of coupling relationships at a specific time scale, thereby achieving in-depth capture of the complex operating characteristics of the power-heat network system.

[0044] S2.3: Based on the mean significance of the fluctuations of each type of operating parameter across all observation windows, construct the parameter fluctuation sequence for the corresponding historical period, which is used to cluster all types of operating parameters under different historical periods.

[0045] Based on the above calculations and analysis, and considering the persistent differences in various operating parameters over historical periods and their variation characteristics at different observation scales, the response characteristics of various operating parameters to changes in operating conditions are precisely quantified. The fluctuation significance of each type of operating parameter under each observation window is quantified. Furthermore, this embodiment constructs a parameter fluctuation sequence for the corresponding historical period based on the average fluctuation significance of each type of operating parameter under all observation windows. This sequence is used to cluster all types of operating parameters under different historical periods. The specific process is as follows: First, the average fluctuation significance values ​​of all types of operating parameters on the power side, the average fluctuation significance values ​​of all types of operating parameters on the heat side, and the average fluctuation significance values ​​of all types of operating parameters of the coupling unit within the historical period are respectively formed into a first sequence, a second sequence, and a third sequence. The sequence obtained by splicing the first, second, and third sequences is used as the parameter fluctuation sequence for the corresponding historical period.

[0046] Furthermore, all historical cycles within a preset time period are used as input to the clustering algorithm. The distance metric is set as the Euclidean distance between parameter fluctuation sequences of different historical cycles, and the number of clusters is determined by the elbow rule. All clusters are output, and each cluster contains different historical cycles.

[0047] It should be noted that the preset duration is a positive integer multiple of the historical period; in this embodiment, the preset duration is 3 years. It should be noted that there are many commonly used clustering algorithms. In this embodiment, the agglomerative layer clustering algorithm is used for clustering. In practical applications, as other implementation methods, implementers may also use density clustering algorithms or other clustering methods according to specific circumstances. This embodiment does not impose any special restrictions on the selection of clustering algorithms.

[0048] The process of clustering using the agglomerative layer clustering algorithm, the calculation of Euclidean distance, and the determination of the number of clusters using the elbow rule are all well-known techniques and will not be elaborated further.

[0049] Preferably, the flowchart of the historical periodic clustering steps provided in this embodiment is as follows: Figure 2 As shown.

[0050] All class running parameters at each time point are combined into a running parameter sequence; the number of all running parameter sequences in all historical periods in each cluster is counted and recorded as the total number of running parameter sequences for each cluster; the proportion of the total number of running parameter sequences for each cluster in the sum of the total number of running parameter sequences for all clusters is calculated and multiplied by the preset total number of samples, which is used as the number of samples to be selected from each cluster; the corresponding number of running parameter sequences are selected from each cluster as samples; all clusters are traversed to obtain all samples; all samples are divided into training set and test set according to a preset ratio.

[0051] It should be noted that the preset total number of samples and the preset ratio are both set manually. In this embodiment, the preset total number of samples is 10,000, and the preset ratio is 7:3. The preset total number of samples is set to 10,000 to construct a dataset with sufficient statistical significance, ensuring that it can cover all extreme operating conditions, seasonal patterns, and equipment failure characteristics that may occur within a 3-year historical period. This significantly improves the generalization ability and robustness of the neural network and avoids overfitting due to insufficient sample size. At the same time, the preset ratio of 7:3 (i.e., 7,000 training samples and 3,000 test samples) is the golden ratio in the field of deep learning. 70% of the data is used for training, which allows the model to fully learn the nonlinear mapping relationship and deep features of the complex power heating network coupling system, while 30% of the independent data is reserved for rigorous testing. This allows for an objective evaluation of the model's prediction accuracy and generalization performance on unseen data, ensuring the reliability of the simulation results in practical applications.

[0052] Based on the clustering process, it can be understood that this is a strategy of hierarchical balanced sampling based on the inherent distribution characteristics of the data. It aims to solve the problem of the flooding of niche working condition samples and the bias in model training caused by uneven data distribution. The process first uses agglomerative hierarchical clustering algorithm to divide massive historical state data into several clusters with similar fluctuation characteristics and working condition attributes, so that the data in each cluster represents a specific operating mode (such as peak in extremely cold weather, stable during the spring and autumn transition season, etc.).

[0053] Thus, this embodiment constructs a parameter fluctuation sequence that integrates the characteristics of the power side, thermal side, and coupled unit by calculating the mean significance of the fluctuations of various operating parameters under multi-scale observation windows. It also uses the agglomerative hierarchical clustering algorithm to accurately divide the historical cycle based on Euclidean distance, thereby realizing the automatic identification and classification of complex operating conditions.

[0054] S3: Based on the multi-dimensional operating parameters of all historical periods in each cluster, a training set and a test set are constructed to train the neural network to obtain the power heating network model for power heating network simulation.

[0055] All class running parameters at each time point are combined into a running parameter sequence; the number of all running parameter sequences in all historical periods in each cluster is counted and recorded as the total number of running parameter sequences for each cluster; the proportion of the total number of running parameter sequences for each cluster in the sum of the total number of running parameter sequences for all clusters is calculated and multiplied by the preset total number of samples, which is used as the number of samples to be selected from each cluster; the corresponding number of running parameter sequences are selected from each cluster as samples; all clusters are traversed to obtain all samples; all samples are divided into training set and test set according to a preset ratio.

[0056] It should be noted that the preset total number of samples and the preset ratio are both set manually. In this embodiment, the preset total number of samples is 10,000, and the preset ratio is 7:3. The preset total number of samples is set to 10,000 to construct a dataset with sufficient statistical significance, ensuring that it can cover all extreme operating conditions, seasonal patterns, and equipment failure characteristics that may occur within a 3-year historical period. This significantly improves the generalization ability and robustness of the neural network and avoids overfitting due to insufficient sample size. At the same time, the preset ratio of 7:3 (i.e., 7,000 training samples and 3,000 test samples) is the golden ratio in the field of deep learning. 70% of the data is used for training, which allows the model to fully learn the nonlinear mapping relationship and deep features of the complex power heating network coupling system, while 30% of the independent data is reserved for rigorous testing. This allows for an objective evaluation of the model's prediction accuracy and generalization performance on unseen data, ensuring the reliability of the simulation results in practical applications.

[0057] Based on the sample selection process, it can be understood that a predetermined number of samples are drawn from each cluster according to the proportion and weight of each cluster in the total dataset. This ensures that the final training set and test set are evenly distributed across various operating conditions, strengthens the learning of common features and weakens random interference, enabling the trained power heating network model to accurately adapt to complex changes under all operating conditions, and helps to improve the simulation accuracy and stability of the power heating network model.

[0058] Based on the analysis and calculations in the above steps, training and test sets are obtained. Furthermore, in this embodiment, training and test sets are constructed based on the multi-dimensional operating parameters of all historical periods in each cluster to train the neural network and obtain a power heating network model for power heating network simulation. The specific process is as follows: Multidimensional key indicators of the power heating network are acquired in real time. All key indicators at each time point are combined into a key indicator sequence. The training set and test set are used as the learning samples and validation samples of the neural network, respectively. The key indicator sequence is used as the output label of the neural network. The loss function is the weighted sum of mean squared error (MSE) and mean absolute percentage error (MAPE). The optimizer is the Adam optimizer. The neural network is trained and the trained neural network model is used as the power heating network model.

[0059] In this embodiment, the multidimensional key indicators of the power heating network include the load rate and load ramp rate of each unit, the coefficient of variation of the pipeline flow, the lag time of the pipeline temperature, the network loss rate, and the heating compliance rate. The sum of the weights of the mean square error (MSE) and the mean absolute percentage error (MAPE) is 1. Implementers need to adjust this according to the type of input data for model training. This embodiment does not impose any special restrictions.

[0060] It should be noted that there are many commonly used neural networks. In this embodiment, a Long Short-Term Memory (LSTM) network is used for training. In practical applications, as other implementation methods, implementers may also choose other neural networks according to specific circumstances. This embodiment does not impose any special restrictions.

[0061] The process of training the Long Short-Term Memory network, the mean squared error, the mean absolute percentage error, and the working principle of the Adam optimizer are all well-known techniques and will not be elaborated further.

[0062] Furthermore, the constructed power heating network model is used for simulation applications under complex operating conditions in multiple scenarios. Specifically, the multi-dimensional operating parameters of the power side, heat side, and coupling units at the current moment are used as inputs to the power heating network model, and the key indicators at the current moment are output. Then, the actual scheduling scheme is adjusted based on the key indicators. For example, under the peak operating conditions of low temperature heat load in winter, if the simulation shows that the load rate of the CHP unit is low, resulting in low water supply temperature, the scheduling strategy can be adjusted to "prioritize heat supply and moderately constrain power generation", while starting the electric boiler to supplement heat supply and coordinating the release of heat energy by the thermal storage device.

[0063] Furthermore, after applying the power heating network model to simulations, subsequent optimization and feedback adjustments are made based on the simulation results. Actual operating data is fed back into the power heating network model to update its dataset. Based on the updated data, dynamic optimization of the power heating network model is achieved, ensuring that the model can accurately match the actual operating state of the power heating network.

[0064] Thus, this embodiment selects training and testing sets covering all operating conditions based on the clustering results using a hierarchical balanced sampling strategy, according to the proportion of each cluster. It also integrates multi-dimensional key indicators from the power side, thermal side, and coupling units as output labels for the neural network, trains the neural network, and constructs a power heating network model that can accurately map complex nonlinear relationships. Furthermore, by introducing a feedback mechanism based on actual operating data, it achieves continuous updating of the dataset and iterative optimization of the model, thereby ensuring high-precision simulation and stability of the power heating network model under all operating conditions.

[0065] Based on the same inventive concept as the above method, this application embodiment also provides a power heating network model simulation system, including a memory, a processor, and a computer program stored in the memory and running on the processor. When the processor executes the computer program, it implements the steps of any one of the above-described power heating network model simulation methods.

[0066] It should be noted that the order of the embodiments described above is merely for descriptive purposes and does not represent the superiority or inferiority of the embodiments. Furthermore, specific embodiments of this specification have been described above. Additionally, the processes depicted in the accompanying drawings do not necessarily require a specific or sequential order to achieve the desired results. In some implementations, multitasking and parallel processing are possible or may be advantageous.

[0067] The various embodiments in this specification are described in a progressive manner. The same or similar parts between the various embodiments can be referred to each other. Each embodiment focuses on describing the differences from other embodiments.

[0068] The above description is only a preferred embodiment of this application and is not intended to limit this application. Any modifications, equivalent substitutions, improvements, etc., made within the principles of this application should be included within the protection scope of this application.

Claims

1. A simulation method for a power heating network model, characterized in that, The method includes the following steps: Real-time acquisition of multi-dimensional operating parameters from the power side, thermal side, and coupling unit; Based on the degree of fluctuation of each type of operating parameter within the historical cycle, a minimum time unit that can characterize its fluctuation characteristics is set as the basic observation window for each type of operating parameter. On the basis of the basic observation window, multi-scale observation windows are constructed by expanding by integer multiples until the entire historical cycle is covered. By combining the dispersion of each type of operating parameter under each observation window, and the correlation between each type of operating parameter and any other type of operating parameter, the fluctuation significance of each type of operating parameter under each observation window is quantified, which is used to reflect the fluctuation intensity of each type of operating parameter within a specific time window and the degree of coupling with other parameters. Based on the mean significance of the fluctuations of each type of operating parameter across all observation windows, a parameter fluctuation sequence corresponding to the historical period is constructed for clustering different historical periods; Training and testing sets are constructed based on the multidimensional operating parameters of all historical periods in each cluster to train the neural network and obtain the power heating network model for power heating network simulation.

2. The power heating network model simulation method as described in claim 1, characterized in that, The multi-dimensional operating parameters on the power side include the active and reactive power of the generating unit, node voltage, line transmission power, and power load; the multi-dimensional operating parameters on the heat side include the heat supplied by the heat source, the supply and return water temperatures of the pipeline nodes, pipeline flow rate, pipeline pressure, and heat load demand; the multi-dimensional operating parameters of the coupling unit include the input and output power of the generating unit, the input and output power of the electric boiler, and the operating power of the heat pump.

3. The power heating network model simulation method as described in claim 1, characterized in that, The method for setting the basic observation window for each type of operating parameter is as follows: The result of subtracting the normalized value of the fluctuation of each operating parameter in the historical period from 1 is denoted as the weighting factor. The product of the number of each type of operating parameter in the historical period and the weight factor is used as the input to the rounding function, and the output value is used as the size of the basic observation window for each type of operating parameter. That is, the basic observation window needs to contain the number of data points of the output value.

4. The power heating network model simulation method as described in claim 3, characterized in that, The method for constructing the multi-scale observation window is as follows: If it exists Then the observation window sizes for the x-th type of running parameters at multiple scales are as follows: Conversely, if ,but Then the observation window sizes for the x-th type of running parameters at multiple scales are as follows: ,in, Represents positive integers. This represents the size of the basic observation window for the x-th type of operating parameter. This indicates the number of x-th type of operating parameters within a historical period.

5. The power heating network model simulation method as described in claim 1, characterized in that, The significance of the fluctuation of each type of operating parameter under each observation window is positively correlated with the degree of dispersion of each type of operating parameter under each observation window, and negatively correlated with the correlation between each type of operating parameter and any other type of operating parameter.

6. The power heating network model simulation method as described in claim 1, characterized in that, The construction of the parameter fluctuation sequence corresponding to the historical period includes: The average fluctuation significance values ​​of all types of operating parameters on the power side, the average fluctuation significance values ​​of all types of operating parameters on the heat side, and the average fluctuation significance values ​​of all types of operating parameters of the coupling unit within the historical period are respectively formed into a first sequence, a second sequence, and a third sequence; The sequence obtained by splicing the first, second, and third sequences is used as the parameter fluctuation sequence for the corresponding historical period.

7. The power heating network model simulation method as described in claim 1, characterized in that, The distance metric used in the clustering process is the Euclidean distance between parameter fluctuation sequences of different historical periods.

8. The power heating network model simulation method as described in claim 1, characterized in that, The construction of training and test sets based on multidimensional operating parameters across all historical periods in each cluster includes: All class running parameters at each time point are combined into a running parameter sequence; the number of all running parameter sequences in all historical periods in each cluster is counted and recorded as the total number of running parameter sequences in each cluster; the proportion of the total number of running parameter sequences in each cluster in the sum of the total number of running parameter sequences in all clusters is calculated and multiplied by the preset total number of samples, which is used as the number of samples to be selected from each cluster. Select a corresponding number of running parameter sequences as samples from each cluster, traverse all clusters to obtain all samples, and divide all samples into training set and test set according to a preset ratio.

9. The power heating network model simulation method as described in claim 1, characterized in that, The process of training the neural network to obtain the power heating network model includes: Multidimensional key indicators of the power heating network are acquired in real time. All key indicators at each time point are combined into a key indicator sequence. The training set and test set are used as the learning samples and validation samples of the neural network, respectively. The key indicator sequence is used as the output label of the neural network to train the neural network. The trained neural network model is used as the power heating network model.

10. A power heating network model simulation system, comprising a memory, a processor, and a computer program stored in the memory and running on the processor, characterized in that, When the processor executes the computer program, it implements the steps of the power heating network model simulation method as described in any one of claims 1-9.

Images