Load forecasting and scheduling method adaptive to hot user behaviors

Abstract

The invention discloses a load forecasting and scheduling method adaptive to heat user behaviors, relates to the technical field of load forecasting, solves the technical problem that all factors influencing the load are difficult to capture when a linear method is used for processing complex problems in the prior art, and specifically comprises the following steps: collecting historical load data of a source side pipe network, and analyzing a heat user behavior mode; establishing a pipe network load simulation model based on the historical load data and the heat user behavior mode; simulating a real-time operation state of a pipe network by using a digital twinborn technology, and updating heat user behavior data in real time; according to the simulation result and the heat user behavior data updated in real time, the load change of the pipe network is predicted in real time; according to the prediction result, self-adaptive regulation and control scheduling is carried out so as to adapt to the real-time changeable switching requirements of the heat users; and the regulation and control strategy is optimized through a machine learning algorithm, and optimization of a scheduling plan is ensured.

Description

Technical Field

[0001] The present invention relates to the technical field of load forecasting, and in particular to a load forecasting and scheduling method adapted to thermal user behavior. Background Art

[0002] At present, load forecasting technology is mainly based on the following methods:

[0003] Artificial intelligence technology: In recent years, artificial intelligence technology has been widely used in load forecasting, especially in new power systems. These technologies include deep learning, neural networks, machine learning, etc., which are used to process complex and changing load data.

[0004] Traditional forecasting models: including time series method, regression analysis method, etc. These methods still have a place in short-term load forecasting.

[0005] Gas load forecasting technology: With the development of the natural gas industry, gas load forecasting technology is also constantly improving. These technologies are not only related to the investment benefits and reliability of the pipeline network, but also to the ability to plan and optimize the operation of the gas pipeline network.

[0006] Some patents of prior art that are closest to the above patent technology:

[0007] Smart Grid Short-Term Load Forecasting Method (CN114169645A): This technology provides a smart grid short-term load forecasting method, which combines data preprocessing technology and load forecasting technology, and uses the LMD algorithm to combine reinforcement learning and deep learning to reduce the error of smart grid short-term load forecasting data.

[0008] Electricity load forecasting method based on neural network (CN114444750A): This invention discloses an electricity load forecasting method based on neural network. It uses relevant data affecting the electricity load to establish an electricity load forecasting model based on BP neural network to improve the efficiency and accuracy of electricity load forecasting.

[0009] Load forecasting analysis method for typical urban cable network systems: This technology builds a spatial load forecasting platform for urban power grids based on power GIS, proposes a spatial load forecasting method and forecasting model for urban power grids, and achieves the optimal output of predicted load accuracy.

[0010] Short-term load forecasting method for regional distribution network: This invention relates to a short-term load forecasting method for regional distribution network based on twin network classification method, grey wolf algorithm and long short-term memory network, which effectively solves the problems of low efficiency, high operating cost and lack of adaptability of existing forecasting models.

[0011] In view of the above technical defects, a solution is now proposed. Summary of the invention

[0012] The purpose of the present invention is to solve the above-mentioned problems and propose a load forecasting and scheduling method adapted to the behavior of heat users. The specific technical problems solved are as follows:

[0013] 1. High data requirements: It relies heavily on historical data and requires a large amount of accurate and detailed data.

[0014] 2. Simple model: Using linear methods to deal with complex problems makes it difficult to capture all factors that affect the load.

[0015] 3. Insufficient accuracy: The prediction results have a low fit with the actual load and the error is large.

[0016] 4. Difficult to adapt to changes: The load of the power system is affected by many factors, such as weather and seasonal changes. Existing technologies are difficult to adapt to the uncertainty of these changes.

[0017] The purpose of the present invention can be achieved through the following technical solutions:

[0018] A load forecasting and scheduling method adapted to heat user behavior comprises the following steps:

[0019] Step 1: Collect historical load data of the source-side pipe network and analyze the behavior patterns of heat users;

[0020] Step 2: Based on the historical load data and the behavior pattern of heat users, a pipe network load simulation model is established;

[0021] Step 3: Use digital twin technology to simulate the real-time operation status of the pipeline network and update the heat user behavior data in real time;

[0022] Step 4: Based on the simulation results and real-time updated heat user behavior data, real-time prediction of network load changes is performed;

[0023] Step 5: Based on the prediction results, adaptive control and scheduling are performed to adapt to the real-time and changing switching needs of hot users;

[0024] Step 6: Optimize the control strategy through machine learning algorithms to ensure the optimization of the scheduling plan.

[0025] As a preferred embodiment of the present invention, the specific steps of step 1 are as follows:

[0026] Step 11: Data collection: determine data collection points, such as user access points and key nodes; use sensors and smart meters to collect load data at each collection point; transmit the collected data to the central database for storage and processing;

[0027] Step 12: Data preprocessing: clean the data, remove outliers and noise; standardize the data to ensure the consistency of different data sources; perform data aggregation, such as summarizing load data by hour or day;

[0028] Step 13: Hot user behavior pattern analysis: Use clustering algorithms to classify user behaviors and identify different behavior patterns;

[0029] Clustering Algorithms: ;in, is the i-th cluster, k is the number of clusters, is the set of points in the jth cluster, yes The center point of

[0030] Apply association rule learning algorithms to find the correlation between user behaviors; Association rule learning: ; where A and B are item sets, is the frequency of item set A appearing in all transactions, It represents the probability of B happening when A happens;

[0031] Time Series Analysis: ;in, is the observed value at time t, c is a constant term, is the white noise error term, is the autoregressive coefficient, is the moving average coefficient, p and q are the orders of autoregression and moving average respectively; the time series analysis method is used to analyze the changing trend of user behavior over time.

[0032] As a preferred embodiment of the present invention, the specific steps of step 2 are as follows:

[0033] Step 21: Use the seasonal autoregressive integrated moving average model (SARIMA) to characterize the load characteristics and build a pipeline network load simulation model:

[0034] ; Where L is the lag operator, s is the seasonal period, and (P, D, Q) are the orders of seasonal autoregression, difference, and moving average;

[0035] According to the selected mathematical model, the construction of the pipeline network load simulation model is completed:

[0036] Step 22: estimating parameters of the selected model based on the historical load data and the thermal user behavior pattern;

[0037] The least squares method was used to fit the model parameters;

[0038] Verify and adjust the model to ensure that its prediction performance meets the actual application requirements;

[0039] Step 23: The goal of parameter estimation is to find a set of parameters ( , , , ), so that the model best fits the historical load data; this can be achieved by minimizing the error term This is achieved by the sum of squares;

[0040] Least squares fitting: The objective function of the least squares method can be expressed as: ;

[0041] Model validation and tuning: Model validation usually involves calculating the difference between the predicted values ​​and the actual observed values, such as the mean squared error (MSE) or root mean squared error (RMSE): ;

[0042] Where T is the number of test data points, is the load value predicted by the model; by comparing the MSE or RMSE of different parameter sets, the best model parameters can be selected and used to predict future loads, and the model can be further adjusted and optimized according to actual application requirements.

[0043] As a preferred embodiment of the present invention, the application of digital twin technology in step 3 further includes:

[0044] Step 31: Build a digital twin platform;

[0045] Step 32: Create a virtual copy of the pipe network on the digital twin platform;

[0046] Step 33: Input the real-time load data into the virtual copy;

[0047] Step 34: Simulate the operation status of the pipeline network through the virtual copy.

[0048] As a preferred embodiment of the present invention, the real-time prediction of the change in pipe network load in step 4 further includes:

[0049] Step 41: Analyze the data provided by the digital twin platform;

[0050] Step 42: Use machine learning algorithms to predict future loads;

[0051] Step 43: Evaluate the accuracy of the prediction results.

[0052] As a preferred implementation of the present invention, the specific implementation is as follows:

[0053] Collect real-time data: collect real-time load data from sensors and monitoring systems in the pipe network; also collect behavioral data of heat users, such as frequency and duration of use;

[0054] Data integration: Integrate real-time load data with heat user behavior data; use data preprocessing techniques to ensure data quality, such as removing noise and filling missing values;

[0055] Model input update: input the integrated real-time data into the established pipe network load simulation model; update the model parameters to reflect the latest user behavior and load conditions;

[0056] Load forecasting: Use simulation models to predict the network load over a period of time in the future; apply time series analysis, machine learning or other forecasting techniques to generate forecast results;

[0057] Result analysis: Analyze the forecast results to identify possible load peaks or valleys; adjust the network operation strategy based on the forecast results;

[0058] As a preferred implementation of the present invention, the implementation of the adaptive control scheduling in step 5 further includes:

[0059] The implementation of the adaptive control scheduling further includes:

[0060] Step 511: Formulate a scheduling plan based on the prediction results;

[0061] Step 512: Implement the dispatching plan and adjust the pipe network operation parameters;

[0062] Step 513: Monitor the scheduling effect and make adjustments based on the actual situation;

[0063] The formulation of the scheduling plan further includes:

[0064] Step 521: Determine a high-priority hot user;

[0065] Step 522: Allocate necessary resources to the high-priority hot users;

[0066] Step 523: Adjust resources for non-priority users;

[0067] The step of monitoring the scheduling effect further includes:

[0068] Step 531: Setting monitoring indicators;

[0069] Step 532: Collecting pipe network operation data in real time;

[0070] Step 533: Compare the prediction results with the actual operation data to evaluate the scheduling effect.

[0071] As a preferred implementation of the present invention, a feedback control mechanism may be added to step three in the implementation process of the adaptive control scheduling:

[0072] Performance evaluation: is the forecast load at time t, is the actual load, ;

[0073] Parameter Tuning: Parameter tuning can be done using gradient descent or other optimization algorithms to minimize the prediction error: ;in, represents the model parameters, is the learning rate, is the MSE with respect to the parameter Through this feedback control mechanism, the prediction model self-corrects and adapts to changes in heat user behavior and fluctuations in actual load conditions, thereby improving the accuracy of the prediction and the efficiency of the network operation.

[0074] As a preferred embodiment of the present invention, the step of optimizing the control strategy using the machine learning algorithm in step 6 further includes:

[0075] Step 61: Collect control execution result data;

[0076] Step 62: Analyze the deviation between the control execution result and the prediction result;

[0077] Step 63: Adjust the machine learning algorithm parameters based on the deviation analysis results;

[0078] Step 64: Retrain the prediction model to reduce the error of future predictions.

[0079] Compared with the prior art, the present invention has the following beneficial effects:

[0080] 1. In the present invention, the role of thermal user behavior patterns in the pipeline network load simulation model is crucial. Specifically, the analysis results of the thermal user behavior patterns can provide the following effects: Enhanced prediction accuracy: By analyzing the behavior patterns of thermal users, the simulation model can more accurately predict the load demand within a specific time period, especially during peak hours. Provide dynamic adaptability: The analysis of thermal user behavior patterns enables the simulation model to dynamically adapt to changes in user behavior, thereby adjusting predictions in a real-time environment. Optimize resource allocation: The model can perform resource allocation and scheduling plans in advance based on the load changes predicted by the thermal user behavior patterns to meet upcoming demand. Improve risk management: By predicting load fluctuations that may be caused by thermal users, the simulation model helps pipeline network operators identify potential risks and take preventive measures. In general, thermal user behavior patterns provide a mechanism for simulation models to understand and predict how user behavior affects pipeline network loads, thereby making pipeline network operations more efficient and reliable.

[0081] 2. In the present invention, user behavior adaptability: considering the differences in user behavior, it is possible to more accurately predict load demand; dynamic adjustment capability: it is possible to dynamically adjust the prediction model according to real-time data to improve the accuracy of the prediction. Economic operation: while meeting user needs, the economic operation of the heating system is achieved. Optimized scheduling strategy: by optimizing the scheduling model, the peak shaving and valley filling effect of the distribution network is improved, which is beneficial to the economic and safe operation of the system. Improved accuracy: by adapting to user behavior, the accuracy of load forecasting is improved. Enhanced responsiveness: optimizing the scheduling strategy, and improving the ability to respond to changes in grid load. Economic benefits: through more accurate prediction and scheduling, operating costs are reduced and economic benefits are improved. System stability: reducing prediction errors and improving the stability and reliability of the power system BRIEF DESCRIPTION OF THE DRAWINGS

[0082] In order to facilitate understanding by those skilled in the art, the present invention is further described below with reference to the accompanying drawings.

[0083] Figure 1 is a flow chart of the method of the present invention;

[0084] Figure 2 It is a principle flow chart of the present invention. DETAILED DESCRIPTION

[0085] In order to enable those skilled in the art to better understand the scheme of the present invention, the technical scheme in the embodiments of the present invention will be clearly and completely described below in conjunction with the drawings in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0086] Reference to "embodiments" herein means that a particular feature, structure, or characteristic described in conjunction with the embodiments may be included in at least one embodiment of the present invention. The appearance of the phrase in various places in the specification does not necessarily refer to the same embodiment, nor is it an independent or alternative embodiment that is mutually exclusive with other embodiments. It is explicitly and implicitly understood by those skilled in the art that the embodiments described herein may be combined with other embodiments.

[0087] Embodiment 1

[0088] See also Figure 1 As shown, a load forecasting and scheduling method adapted to the behavior of heat users includes the following steps:

[0089] Step 1: Collect historical load data of the source-side pipe network and analyze the behavior patterns of heat users;

[0090] Step 2: Based on the historical load data and the behavior pattern of heat users, a pipe network load simulation model is established;

[0091] Step 3: Use digital twin technology to simulate the real-time operation status of the pipeline network and update the heat user behavior data in real time;

[0092] Step 4: Based on the simulation results and real-time updated heat user behavior data, real-time prediction of network load changes is performed;

[0093] Step 5: Based on the prediction results, adaptive control and scheduling are performed to adapt to the real-time and changing switching needs of hot users;

[0094] Step 6: Optimize the control strategy through machine learning algorithms to ensure the optimization of the scheduling plan;

[0095] Analyzing hot user behavior patterns in step 1 usually involves the following specific steps:

[0096] Data collection: First, collect usage data of heat users in different time periods, including information such as users' activity frequency, duration, load changes, etc.

[0097] Behavior recognition: Through data mining technology, the behavior patterns of hot users are identified, such as high-frequency activities or load peaks in a specific time period.

[0098] Pattern classification: Classify the identified behavior patterns to determine which ones are normal and which ones are abnormal.

[0099] Correlation analysis: Correlation analysis is performed to discover relationships between different behavior patterns, such as whether certain patterns occur after specific events.

[0100] Predictive model construction: Use statistics and machine learning algorithms to build a predictive model based on historical behavior data to predict the future behavior trends of hot users.

[0101] Model validation: Test the accuracy of the prediction model through actual data and adjust the model parameters as needed;

[0102] See also Figure 2 As shown, step 1 collects historical load data of the source-side pipe network and analyzes the behavior patterns of heat users; the specific steps are as follows:

[0103] Step 11: Data collection: determine data collection points, such as user access points, key nodes, etc.; use sensors, smart meters and other equipment to collect load data at each collection point; transmit the collected data to the central database for storage and processing;

[0104] Step 12: Data preprocessing: clean the data, remove outliers and noise; standardize the data to ensure the consistency of different data sources; perform data aggregation, such as summarizing load data by hour or day;

[0105] Step 13: Hot user behavior pattern analysis: Use clustering algorithms to classify user behaviors and identify different behavior patterns;

[0106] Clustering algorithms, such as K-means: ;in, is the i-th cluster, k is the number of clusters, is the set of points in the jth cluster, yes The center point of

[0107] Apply association rule learning algorithms to find correlations between user behaviors;

[0108] Association rule learning, such as Apriori algorithm: ; where A and B are item sets, is the frequency of item set A appearing in all transactions, It represents the probability of B happening when A happens;

[0109] Time series analysis, such as ARIMA models: ;in, is the observed value at time t, c is a constant term, is the white noise error term, is the autoregressive coefficient, is the moving average coefficient, p and q are the orders of autoregression and moving average respectively; use time series analysis method to analyze the changing trend of user behavior over time;

[0110] In step 2, a pipe network load simulation model is established based on the historical load data and the heat user behavior pattern; the specific steps are as follows:

[0111] Step 21: Use the seasonal autoregressive integrated moving average model (SARIMA) to characterize the load characteristics and build a pipeline network load simulation model:

[0112] ; where L is the lag operator, s is the seasonal period, and (P, D, Q) are the orders of seasonal autoregression, difference, and moving average.

[0113] According to the selected mathematical model, the construction of the pipeline network load simulation model is completed:

[0114] Step 22: estimating parameters of the selected model based on the historical load data and the thermal user behavior pattern;

[0115] The least squares method was used to fit the model parameters;

[0116] The model is validated and adjusted to ensure that its prediction performance meets the actual application requirements.

[0117] Step 23: The goal of parameter estimation is to find a set of parameters ( , , , ), so that the model best fits the historical load data; this can be achieved by minimizing the error term This is achieved by summing the squares of .

[0118] Least squares fitting: The objective function of the least squares method can be expressed as: ;

[0119] Model validation and tuning: Model validation usually involves calculating the difference between the predicted values ​​and the actual observed values, such as the mean squared error (MSE) or root mean squared error (RMSE): ;

[0120] Where T is the number of test data points, is the load value predicted by the model.

[0121] By comparing the MSE or RMSE of different parameter sets, the best model parameters can be selected. These parameters can then be used to predict future loads and further adjust and optimize the model according to actual application requirements;

[0122] The application of the digital twin technology described in step 3 further includes:

[0123] Step 31: Build a digital twin platform;

[0124] Step 32: Create a virtual copy of the pipe network on the digital twin platform;

[0125] Step 33: Input the real-time load data into the virtual copy;

[0126] Step 34: Simulate the operation status of the pipe network through the virtual copy;

[0127] The real-time prediction of the change in pipe network load in step 4 further includes:

[0128] Step 41: Analyze the data provided by the digital twin platform;

[0129] Step 42: Use machine learning algorithms to predict future loads;

[0130] Step 43: Evaluate the accuracy of the prediction results.

[0131] The details are as follows:

[0132] Collect real-time data (following the data collection in the previous section): collect real-time load data from the sensors and monitoring systems of the pipe network. At the same time, collect the behavior data of heat users, such as frequency of use, duration, etc.

[0133] Data integration: Integrate real-time load data with thermal user behavior data. Use data preprocessing techniques to ensure data quality, such as removing noise and filling missing values.

[0134] Model input update: Input the integrated real-time data into the established pipe network load simulation model; update the model parameters to reflect the latest user behavior and load conditions.

[0135] Load forecasting: Use simulation models to predict the network load over a period of time in the future. Apply time series analysis, machine learning or other forecasting techniques to generate forecast results.

[0136] Result analysis: Analyze the forecast results and identify possible load peaks or valleys. Adjust the network operation strategy based on the forecast results, such as adjusting pressure, flow, etc.

[0137] The implementation of the adaptive control scheduling described in step 5 further includes:

[0138] The implementation of the adaptive control scheduling further includes:

[0139] Step 511: Formulate a scheduling plan based on the prediction results;

[0140] Step 512: Implement the dispatching plan and adjust the pipe network operation parameters;

[0141] Step 513: Monitor the scheduling effect and make adjustments based on actual conditions.

[0142] The formulation of the scheduling plan further includes:

[0143] Step 521: Determine a high-priority hot user;

[0144] Step 522: Allocate necessary resources to the high-priority hot users;

[0145] Step 523: Adjust resources for non-priority users.

[0146] The step of monitoring the scheduling effect further includes:

[0147] Step 531: Setting monitoring indicators;

[0148] Step 532: Collecting pipe network operation data in real time;

[0149] Step 533: Compare the prediction results with the actual operation data to evaluate the scheduling effect.

[0150] In the implementation process of the adaptive control scheduling, a feedback control mechanism can also be added in step 3:

[0151] Performance evaluation: is the forecast load at time t, is the actual load, ;

[0152] Parameter Tuning: Parameter tuning can be done using gradient descent or other optimization algorithms to minimize the prediction error: ;

[0153] in, represents the model parameters, is the learning rate, is the MSE with respect to the parameter gradient.

[0154] Through this feedback control mechanism, the forecasting model can self-correct and adapt to changes in heat user behavior and fluctuations in actual load conditions, thereby improving the accuracy of forecasts and the efficiency of network operation;

[0155] The step of optimizing the control strategy using the machine learning algorithm in step 6 further includes:

[0156] Step 61: Collect control execution result data;

[0157] Step 62: Analyze the deviation between the control execution result and the prediction result;

[0158] Step 63: Adjust the machine learning algorithm parameters based on the deviation analysis results;

[0159] Step 64: Retrain the prediction model to reduce the error of future predictions;

[0160] Embodiment 2

[0161] Based on the above embodiment, an actual scenario operation simulation is performed, as follows:

[0162] Assume that a city gas company wants to optimize its pipeline network operation and improve its adaptability to changes in heat user behavior and the accuracy of load forecasting.

[0163] Step 1: Data Collection and Analysis

[0164] Smart sensors are installed at key nodes of the city gas pipeline network to collect usage data of heat users, including usage time, frequency and load.

[0165] The collected data is transferred to the central processing system and pre-processed, including data cleaning, standardization and aggregation.

[0166] Step 2: User behavior pattern analysis

[0167] Apply clustering algorithms to classify user behavior and identify regular and abnormal usage patterns.

[0168] Association rule learning algorithms are used to analyze the relationships between different user behavior patterns.

[0169] Use time series analysis methods to predict trends and cyclical changes in user behavior.

[0170] Step 3: Load simulation model construction

[0171] Based on historical load data and user behavior patterns, select a suitable mathematical model, such as the SARIMA model, to characterize load characteristics.

[0172] The model parameters are estimated using historical data and the model is optimized through cross-validation.

[0173] Step 4: Digital Twin Technology Application

[0174] Build a digital twin platform and create a virtual copy of the pipeline network on the platform.

[0175] Real-time load data is fed into the virtual replica to simulate the operation status of the network.

[0176] Step 5: Load Forecasting and Scheduling

[0177] The constructed simulation model is combined with the data provided by the digital twin platform to carry out load forecasting.

[0178] Based on the prediction results, the network operating parameters, such as pressure and flow, are automatically adjusted to meet the needs of heat users.

[0179] Step 6: Optimization and feedback

[0180] Through machine learning algorithms, such as reinforcement learning, the control strategy is continuously optimized.

[0181] Set monitoring indicators, collect operating data in real time, evaluate scheduling effects, and adjust prediction models based on feedback.

[0182] Results: By implementing the above methods, the gas company was able to achieve rapid response to heat user behavior, improve the accuracy of load forecasting, optimize resource allocation, reduce operating costs, and improve user satisfaction.

[0183] The above formulas are obtained by collecting a large amount of data and performing software simulation, and a formula close to the actual value is selected. The coefficients in the formula are set by technicians in this field according to actual conditions;

[0184] When in use, the present invention collects historical load data of the source-side pipeline network and analyzes the behavior patterns of heat users; establishes a pipeline network load simulation model based on the historical load data and the heat user behavior patterns; uses digital twin technology to simulate the real-time operation status of the pipeline network and updates the heat user behavior data in real time; performs real-time prediction of pipeline network load changes based on the simulation results and the real-time updated heat user behavior data; performs adaptive regulation and scheduling based on the prediction results to adapt to the real-time and changing switching needs of heat users; optimizes the regulation and control strategy through a machine learning algorithm to ensure the optimization of the scheduling plan.

[0185] The preferred embodiments of the present invention disclosed above are only used to help explain the present invention. The preferred embodiments do not describe all the details in detail, nor do they limit the invention to only specific implementation methods. Obviously, many modifications and changes can be made according to the content of this specification. This specification selects and specifically describes these embodiments in order to better explain the principles and practical applications of the present invention, so that those skilled in the art can understand and use the present invention well. The present invention is limited only by the claims and their full scope and equivalents.

Claims

1. A load forecasting and scheduling method adapted to the behavior of heat users, characterized in that: The following steps are involved: Step 1: Collect historical load data of the source-side pipe network and analyze the behavior patterns of heat users; Step 2: Based on the historical load data and the behavior pattern of heat users, a pipe network load simulation model is established; Step 3: Use digital twin technology to simulate the real-time operation status of the pipeline network and update the heat user behavior data in real time; Step 4: Based on the simulation results and real-time updated heat user behavior data, real-time prediction of network load changes is performed; Step 5: Based on the prediction results, adaptive control and scheduling are performed to adapt to the real-time and changing switching needs of hot users; Step 6: Optimize the control strategy through machine learning algorithms to ensure the optimization of the scheduling plan.

2. A load forecasting and scheduling method adapted to thermal user behavior according to claim 1, characterized in that: Step 1 The specific steps are as follows: Step 11: Data collection: determine data collection points, such as user access points and key nodes; use sensors and smart meters to collect load data at each collection point; transmit the collected data to the central database for storage and processing; Step 12: Data preprocessing: clean the data and remove outliers and noise; Standardize data to ensure consistency across different data sources; Perform data aggregation, such as summarizing load data by hour or day; Step 13: Hot user behavior pattern analysis: Use clustering algorithms to classify user behaviors and identify different behavior patterns; Clustering Algorithms: ;in, is the i-th cluster, k is the number of clusters, is the set of points in the jth cluster, yes The center point of Apply association rule learning algorithms to find the correlation between user behaviors; Association rule learning: ; where A and B are item sets, is the frequency of item set A appearing in all transactions, It represents the probability of B happening when A happens; Time Series Analysis: ;in, is the observed value at time t, c is a constant term, is the white noise error term, is the autoregressive coefficient, is the moving average coefficient, p and q are the orders of autoregression and moving average respectively; time series analysis method is used to analyze the changing trend of user behavior over time.

3. A load forecasting and scheduling method adapted to thermal user behavior according to claim 1, characterized in that: Step 2 The specific steps are as follows: Step 21: Use the seasonal autoregressive integrated moving average model to characterize the load characteristics and build a pipe network load simulation model: ; Where L is the lag operator, s is the seasonal period, and (P, D, Q) are the orders of seasonal autoregression, difference, and moving average; According to the selected mathematical model, the construction of the pipeline network load simulation model is completed: Step 22: estimating parameters of the selected model based on the historical load data and the thermal user behavior pattern; The least squares method was used to fit the model parameters; Verify and adjust the model to ensure that its prediction performance meets the actual application requirements; Step 23: The goal of parameter estimation is to find a set of parameters ( , , , ), so that the model best fits the historical load data; this can be achieved by minimizing the error term This is achieved by the sum of the squares of Least squares fitting: The objective function of the least squares method can be expressed as: ; Model validation and tuning: Model validation usually involves calculating the difference between the predicted values ​​and the actual observed values, such as the mean squared error or root mean squared error: ; Where T is the number of test data points, is the load value predicted by the model; by comparing the MSE or RMSE of different parameter sets, the best model parameters can be selected and used to predict future loads, and the model can be further adjusted and optimized according to actual application requirements.

4. A load forecasting and scheduling method adapted to thermal user behavior according to claim 1, characterized in that: The application of digital twin technology in step 3 further includes: Step 31: Build a digital twin platform; Step 32: Create a virtual copy of the pipe network on the digital twin platform; Step 33: Input the real-time load data into the virtual copy; Step 34: Simulate the operation status of the pipeline network through the virtual copy.

5. A load forecasting and scheduling method adapted to thermal user behavior according to claim 1, characterized in that: The real-time prediction of the network load change in step 4 further includes: Step 41: Analyze the data provided by the digital twin platform; Step 42: Use machine learning algorithms to predict future loads; Step 43: Evaluate the accuracy of the prediction results.

6. A load forecasting and scheduling method adapted to thermal user behavior according to claim 5, characterized in that: The details are as follows: Collect real-time data: collect real-time load data from sensors and monitoring systems in the pipe network; also collect behavioral data of heat users, such as frequency and duration of use; Data integration: Integrate real-time load data with heat user behavior data; Use data preprocessing techniques to ensure data quality, such as removing noise and filling missing values; Model input update: input the integrated real-time data into the established pipe network load simulation model; update the model parameters to reflect the latest user behavior and load conditions; Load forecasting: Use simulation models to predict the network load over a period of time in the future; apply time series analysis, machine learning or other forecasting techniques to generate forecast results; Result analysis: Analyze the forecast results and identify possible load peaks or troughs; adjust the pipeline network operation strategy based on the forecast results.

7. A load forecasting and scheduling method adapted to thermal user behavior according to claim 1, characterized in that: The implementation of the adaptive control scheduling in step 5 further includes: Step 511: Formulate a scheduling plan based on the prediction results; Step 512: Implement the dispatching plan and adjust the pipe network operation parameters; Step 513: Monitor the scheduling effect and make adjustments based on the actual situation; The formulation of the scheduling plan further includes: Step 521: Determine a high-priority hot user; Step 522: Allocate necessary resources to the high-priority hot users; Step 523: Adjust resources for non-priority users; The step of monitoring the scheduling effect further includes: Step 531: Setting monitoring indicators; Step 532: Collecting pipe network operation data in real time; Step 533: Compare the prediction results with the actual operation data to evaluate the scheduling effect.

8. A load forecasting and scheduling method adapted to thermal user behavior according to claim 7, characterized in that: In the implementation process of the adaptive control scheduling, a feedback control mechanism can also be added in step 3: Performance evaluation: is the forecast load at time t, is the actual load, ; Parameter Tuning: Parameter tuning can be done using gradient descent or other optimization algorithms to minimize the prediction error: ;in, represents the model parameters, is the learning rate, is the MSE with respect to the parameter Through this feedback control mechanism, the prediction model self-corrects and adapts to changes in heat user behavior and fluctuations in actual load conditions, thereby improving the accuracy of the prediction and the efficiency of the network operation.

9. A load forecasting and scheduling method adapted to thermal user behavior according to claim 1, characterized in that: Step 6 The step of optimizing the control strategy with the machine learning algorithm further includes: Step 61: Collect control execution result data; Step 62: Analyze the deviation between the control execution result and the prediction result; Step 63: Adjust the machine learning algorithm parameters based on the deviation analysis results; Step 64: Retrain the prediction model to reduce the error of future predictions.

Images