Heat supply load prediction system

Abstract

The invention discloses a heat supply load prediction system, and relates to the technical field of heat supply prediction. The data acquisition and fusion module is used for acquiring multi-source heterogeneous data in real time and integrating the acquired multi-source heterogeneous data into a data set with a uniform format; the feature engineering module extracts key features influencing the heat supply load, and a feature subset with the highest prediction value is screened out through a feature selection method and serves as input of a deep learning model; constructing a deep long-short-term memory network model based on an attention mechanism, and performing model training through the training data set; the real-time prediction and adaptive adjustment module receives new data acquired by the data acquisition and fusion module in real time, and inputs the new data into the trained deep long and short term memory network model based on the attention mechanism for real-time prediction of the heat supply load; by integrating multi-source heterogeneous data and utilizing an advanced deep learning model, accurate prediction of the heat supply load is realized, and the method has real-time learning and adaptive adjustment capabilities.

Description

Technical Field

[0001] The present invention relates to the technical field of heating load prediction, and more particularly to a heating load prediction system. Background Art

[0002] With the acceleration of the urbanization process and the continuous improvement of people's requirements for the comfort of the living environment, the heating system, as an important part of urban infrastructure, its operating efficiency and energy utilization efficiency have received increasing attention. Heating load prediction, as a key link in the management and optimization of the heating system, is of great significance for realizing heating on demand, reducing energy consumption, and improving energy utilization efficiency.

[0003] However, heating load prediction faces many challenges. First of all, the heating load is affected by many factors, including outdoor meteorological conditions, building characteristics, user behavior, and the operating status of the heating system itself. The interaction between these factors makes the heating load show high complexity and nonlinearity.

[0004] Secondly, heating load data usually has the characteristics of time series, that is, historical data has a certain reference value for predicting future loads. However, time series data often has problems such as noise, outliers, and missing values, which will have a negative impact on the accuracy and stability of the prediction model.

[0005] In addition, with the rapid development of the Internet of Things technology, the heating system is gradually developing towards the direction of intelligence and networking.

[0006] However, the existing heating load prediction methods have problems such as high computational complexity and limited prediction accuracy when dealing with large-scale and high-dimensional data.

[0007] Therefore, developing a heating load prediction system based on multi-dimensional data fusion and deep learning algorithms to achieve accurate prediction and intelligent management of heating loads is a technical problem that needs to be solved urgently by those skilled in the art. Summary of the Invention

[0008] In view of this, the present invention provides a heating load prediction system, which uses data fusion technology to integrate multi-source heterogeneous data, improves data quality and reliability; uses feature engineering methods to extract key features, reduces model complexity; introduces deep learning algorithms to construct a prediction model, and improves prediction accuracy and generalization ability.

[0009] To achieve the above object, the present invention adopts the following technical solutions:

[0010] A heating load prediction system includes a data collection and fusion module, a feature engineering module, a deep learning model construction and training module, a real-time prediction and adaptive adjustment module, and a visualization and decision support module;

[0011] The data acquisition and fusion module collects in real time multi-source heterogeneous data composed of historical load data, outdoor meteorological data, building characteristic data, and user behavior data of the heating system, and integrates the collected multi-source heterogeneous data into a dataset in a unified format through a data fusion algorithm;

[0012] The feature engineering module receives the dataset from the data acquisition and fusion module, extracts the key features affecting the heating load using statistical analysis and machine learning techniques, and filters out the most predictive feature subset through a feature selection method as the input of the deep learning model;

[0013] The deep learning model construction and training module constructs a deep long short-term memory network model based on the attention mechanism based on the feature subset provided by the feature engineering module, and trains the deep long short-term memory network model based on the attention mechanism through a training dataset;

[0014] The real-time prediction and adaptive adjustment module receives the new data collected in real time by the data acquisition and fusion module, inputs it into the trained deep long short-term memory network model based on the attention mechanism for real-time prediction of the heating load; meanwhile, the real-time prediction and adaptive adjustment module can also dynamically adjust the parameters of the deep learning model according to the deviation between the prediction result and the actual load, realizing online learning and adaptive optimization of the deep learning model;

[0015] The visualization and decision support module receives the prediction results of the real-time prediction and adaptive adjustment module, displays the prediction results, model performance evaluation indicators, and historical load trends through a visualization interface, and generates a heating system scheduling plan and energy-saving optimization suggestions based on the prediction results for users' reference.

[0016] Preferably, in the above-mentioned heating load prediction system, the feature selection method adopted by the feature engineering module adopts the recursive feature elimination method or the model-based feature selection method to filter out the most predictive feature subset.

[0017] Preferably, in the above-mentioned heating load prediction system, the deep long short-term memory network model based on the attention mechanism includes an input layer, an LSTM layer, an attention layer, a fully connected layer, and an output layer, and adopts cross-validation and early stopping strategies to prevent model overfitting.

[0018] Preferably, in the above-mentioned heating load prediction system, the deep learning model construction and training module also introduces a multi-task learning framework to simultaneously predict the heating load and its change trend.

[0019] Preferably, in the above-mentioned heating load prediction system, the deep learning model construction and training module adopts a transfer learning strategy to transfer knowledge from the dataset in the field of heating load prediction, accelerating the model training process.

[0020] Preferably, in the above-mentioned heating load prediction system, the heating system scheduling scheme includes adjusting the heating temperature, adjusting the flow rate, or adjusting the operation time; the energy-saving optimization suggestions include replacing matching energy-saving equipment and optimizing the building insulation performance.

[0021] Preferably, in the above-mentioned heating load prediction system, the visualization and decision support module provides a visualization interface that uses a combination of charts, dashboards, and maps for intelligent comprehensive display.

[0022] Preferably, in the above-mentioned heating load prediction system, the visualization and decision support module is also provided with a user interaction unit. The user interaction unit receives user input instructions, and the instructions include: adjusting prediction parameters and selecting a prediction period; the user interaction unit can also feedback the prediction results and model performance information.

[0023] Preferably, in the above-mentioned heating load prediction system, the visualization and decision support module also includes an abnormal data processing unit for automatically detecting and processing outliers or missing values in the dataset. The abnormal data processing unit uses the box plot method to identify abnormal data. For the identified abnormal data, processing methods such as deletion, filling, or marking as outliers can be taken. In addition, this abnormal data processing unit can also generate an abnormal data report for users to review and process and can automatically detect and process abnormal data.

[0024] As can be seen from the above technical solutions, compared with the prior art, the present invention discloses a heating load prediction system. This system can comprehensively consider various influencing factors, handle the complexity and nonlinearity of time series data, improve the prediction accuracy and stability; at the same time, use the Internet of Things technology to realize real-time data collection and transmission, providing strong support for the intelligent management and optimization of the heating system.

[0025] Data collection and fusion: Real-time collection of multi-source heterogeneous data through the Internet of Things technology, and integration into a dataset in a unified format using data fusion algorithms, effectively improving the quality and reliability of the data and providing a solid foundation for subsequent predictions.

[0026] Feature engineering: Extract key features through statistical analysis methods and machine learning techniques, and screen out the most predictive feature subsets, reducing the model complexity and improving the prediction efficiency. At the same time, the feature selection process helps to identify the key factors affecting the heating load, providing a scientific basis for optimizing the heating system.

[0027] Deep learning model: It adopts a deep long short-term memory network model based on the attention mechanism, which can capture the long-term dependencies of time series data and dynamically adjust the contribution degrees of different time steps to the prediction results, thereby improving the prediction accuracy. In addition, a multi-task learning framework and a transfer learning strategy are introduced to further enhance the generalization ability and prediction accuracy of the model.

[0028] Real-time prediction and adaptive adjustment: By collecting new data in real time and making predictions, while monitoring the deviation between the prediction results and the actual load, the online learning algorithm is used to dynamically adjust the model parameters, realizing the adaptive optimization of the model. This helps to improve the intelligent level of the system and enables it to more flexibly cope with various complex situations.

[0029] Visualization and decision support: By displaying the prediction results and model performance evaluation indicators through a visualization interface, and generating decision support information such as heating system scheduling plans and energy-saving optimization suggestions, it helps users to more intuitively understand the system status and prediction results, so as to make more informed decisions. Brief description of the drawings

[0030] In order to more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the following will briefly introduce the drawings required for use in the description of the embodiments or the prior art. Obviously, the drawings in the following description are only the embodiments of the present invention. For those of ordinary skill in the art, without creative efforts, other drawings can also be obtained according to the provided drawings.

[0031] Figure 1 The attached drawing is a flowchart of the present invention. Detailed implementation manners

[0032] The following will clearly and completely describe the technical solutions in the embodiments of the present invention with reference to the drawings in the embodiments of the present invention. Obviously, the described embodiments are only a part of the embodiments of the present invention, rather than all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those of ordinary skill in the art without creative efforts shall fall within the protection scope of the present invention.

[0033] An embodiment of the present invention discloses a heating load prediction system, including a data acquisition and fusion module, a feature engineering module, a deep learning model construction and training module, a real-time prediction and adaptive adjustment module, and a visualization and decision support module;

[0034] The data acquisition and fusion module collects multi-source heterogeneous data composed of historical load data, outdoor meteorological data, building characteristic data, and user behavior data of the heating system in real time, and integrates the collected multi-source heterogeneous data into a dataset in a unified format through a data fusion algorithm;

[0035] The feature engineering module receives the dataset from the data collection and fusion module, extracts the key features affecting the heating load using statistical analysis and machine learning techniques, and selects the most predictive feature subset through feature selection methods as the input for the deep learning model;

[0036] The deep learning model construction and training module constructs a deep long short-term memory network model based on the attention mechanism based on the feature subset provided by the feature engineering module, and trains the deep long short-term memory network model based on the attention mechanism through the training dataset;

[0037] The real-time prediction and adaptive adjustment module receives the new data collected in real time by the data collection and fusion module, inputs it into the trained deep long short-term memory network model based on the attention mechanism for real-time prediction of the heating load; at the same time, the real-time prediction and adaptive adjustment module can also dynamically adjust the parameters of the deep learning model according to the deviation between the prediction result and the actual load, realizing the online learning and adaptive optimization of the deep learning model;

[0038] The visualization and decision support module receives the prediction results of the real-time prediction and adaptive adjustment module, displays the prediction results, model performance evaluation indicators and historical load trends through the visualization interface, and generates a heating system scheduling plan and energy-saving optimization suggestions based on the prediction results for the user's reference.

[0039] To further optimize the above technical solution, the data collection and fusion module uses Internet of Things technology and data fusion algorithms for real-time collection and integration of multi-source heterogeneous data, and the specific steps are as follows:

[0040] S1: Data preprocessing

[0041] Data cleaning: First, clean the raw data collected from the heating system, meteorological station, building management system and user behavior tracking system, and remove duplicates, missing values and outliers;

[0042] Data format unification: Convert the cleaned data into a unified format, including timestamp, data type, unit, for subsequent processing;

[0043] S2: Data standardization

[0044] Normalization processing: Use the Min-Max normalization method to scale the data to the range of [0,1] to eliminate the influence of different data magnitudes on the fusion result. The normalization formula is:

[0045]

[0046] where x is the original data, x min and x maxare the minimum and maximum values of this data column, x norm is the normalized data;

[0047] Z-score standardization: For data with obvious normal distribution characteristics, the Z-score standardization method is adopted to convert it into a standard normal distribution with a mean of 0 and a standard deviation of 1. The standardization formula is:

[0048]

[0049] where x is the original data, μ is the mean of this data column, σ is the standard deviation of this data column, and z is the normalized data;

[0050] S3: Data fusion

[0051] Adopt the data fusion formula to integrate multi-source heterogeneous data. The data fusion formula is based on the entropy weight method and the correlation analysis between data, automatically determines the weights of each data source, and realizes the efficient fusion of data. The data fusion formula is as follows:

[0052]

[0053] where D fused is the fused data, n is the number of data sources, w i is the weight of the i-th data source, is the normalized data of the i-th data source.

[0054] To further optimize the above technical solution, the feature selection method adopted by the feature engineering module is the recursive feature elimination method or the model-based feature selection method to screen out the most predictive feature subset.

[0055] To further optimize the above technical solution, the deep long short-term memory network model based on the attention mechanism includes an input layer, an LSTM layer, an attention layer, a fully connected layer, and an output layer, and adopts cross-validation and early stopping strategies to prevent the model from overfitting.

[0056] To further optimize the above technical solution, the deep learning model construction and training module also introduces a multi-task learning framework to predict the heating load and its change trend simultaneously.

[0057] To further optimize the above technical solution, the deep learning model construction and training module adopts a transfer learning strategy to transfer knowledge from the dataset in the field of heating load prediction to accelerate the model training process.

[0058] To further optimize the above technical solution, the heating system scheduling plan includes adjusting the heating temperature, adjusting the flow rate, or adjusting the operation time; the energy-saving optimization suggestions include replacing the matching energy-saving equipment and optimizing the building insulation performance.

[0059] To further optimize the above technical solution, the visualization interface provided by the visualization and decision support module uses a combination of charts, dashboards, and maps for intelligent integrated display.

[0060] To further optimize the above technical solution, the visualization and decision support module also sets up a user interaction unit. The user interaction unit receives user input instructions, which include: adjusting prediction parameters and selecting a prediction period; the user interaction unit can also feedback prediction results and model performance information.

[0061] To further optimize the above technical solution, the visualization and decision support module also includes an abnormal data processing unit, which is used to automatically detect and process outliers or missing values in the dataset. The abnormal data processing unit uses the box plot method to identify abnormal data. For the identified abnormal data, processing methods such as deletion, filling, or marking as outliers can be adopted. In addition, the abnormal data processing unit can also generate an abnormal data report for users to review and process and can automatically detect and process abnormal data.

[0062] Technical principle:

[0063] This heating load prediction scheme is based on multi-dimensional data fusion and deep learning algorithms. Its core principles include:

[0064] Data fusion: Using Internet of Things technology to collect multi-source heterogeneous data from the heating system, meteorological station, building, and user side in real time, and integrating them into a high-quality dataset through data fusion algorithms, providing a solid foundation for prediction.

[0065] Feature engineering: Applying statistical analysis and machine learning techniques to extract key features, screening the most predictive feature subset, reducing the model complexity, and improving the prediction efficiency.

[0066] Deep learning prediction: Constructing a deep long short-term memory network model based on the attention mechanism to capture the long-term dependence relationship of time series data and achieve accurate prediction. At the same time, introducing multi-task learning and transfer learning strategies to enhance the model generalization ability.

[0067] Adaptive optimization: Monitoring the prediction deviation through an online learning algorithm, dynamically adjusting the model parameters, and achieving adaptive optimization. Combining with the visualization and decision support module to provide intelligent management and optimization suggestions.

[0068] Robustness enhancement: Using abnormal data processing technology to identify and process outliers, regularly evaluating and optimizing the model performance, and ensuring the stability and reliability of the system.

[0069] In summary, this scheme realizes the accurate prediction and intelligent management of heating load by integrating multi-dimensional data, optimizing feature selection, applying deep learning algorithms, and adaptive optimization strategies.

[0070] The various embodiments in this specification are described in a progressive manner. Each embodiment focuses on the differences from other embodiments. For the same or similar parts among the embodiments, reference can be made to each other. For the devices disclosed in the embodiments, since they correspond to the methods disclosed in the embodiments, the description is relatively simple. For related parts, reference can be made to the description in the method section.

[0071] The above description of the disclosed embodiments enables those skilled in the art to implement or use the present invention. Various modifications to these embodiments will be obvious to those skilled in the art. The general principles defined herein can be implemented in other embodiments without departing from the spirit or scope of the present invention. Therefore, the present invention will not be limited to these embodiments shown herein, but rather will be accorded the widest scope consistent with the principles and novel features disclosed herein.

Claims

1. A heating load prediction system, characterized in that: It includes data collection and fusion module, feature engineering module, deep learning model construction and training module, real-time prediction and adaptive adjustment module, and visualization and decision support module; The data collection and fusion module collects multi-source heterogeneous data consisting of historical load data of the heating system, outdoor meteorological data, building characteristic data, and user behavior data in real time, and integrates the collected multi-source heterogeneous data into a data set in a unified format through a data fusion algorithm; The feature engineering module receives the data set from the data acquisition and fusion module, uses statistical analysis and machine learning techniques to extract key features that affect the heating load, and selects the feature subset with the most predictive value through a feature selection method as the input of the deep learning model; The deep learning model construction and training module constructs a deep long short-term memory network model based on the attention mechanism based on the feature subset provided by the feature engineering module, and trains the deep long short-term memory network model based on the attention mechanism through a training data set; The real-time prediction and adaptive adjustment module receives the new data collected in real time by the data acquisition and fusion module, and inputs it into the trained deep long short-term memory network model based on the attention mechanism to perform real-time prediction of the heating load; at the same time, the real-time prediction and adaptive adjustment module can also dynamically adjust the parameters of the deep learning model according to the deviation between the prediction result and the actual load, so as to realize online learning and adaptive optimization of the deep learning model; The visualization and decision support module receives the prediction results of the real-time prediction and adaptive adjustment module, displays the prediction results, model performance evaluation indicators and historical load trends through a visualization interface, and generates a heating system scheduling plan and energy-saving optimization suggestions based on the prediction results for user reference.

2. A heating load prediction system according to claim 1, characterized in that: The data collection and fusion module uses the Internet of Things technology and data fusion algorithm to collect and integrate multi-source heterogeneous data in real time. The specific steps are as follows: S1: Data preprocessing Data cleaning: First, the raw data collected from the heating system, weather station, building management system, and user behavior tracking system were cleaned to remove duplicates, missing values, and outliers; Data format unification: Convert the cleaned data into a unified format, including timestamp, data type, and unit, for subsequent processing; S2: Data Standardization Normalization: The Min-Max normalization method is used to scale the data to the range of [0,1] to eliminate the impact of different data magnitudes on the fusion results. The normalization formula is: Among them, x is the original data, x min and x max are the minimum and maximum values ​​of the data column, respectively. norm is the normalized data; Z-score standardization: For data with obvious normal distribution characteristics, the Z-score standardization method is used to convert it into a standard normal distribution with a mean of 0 and a standard deviation of 1. The standardization formula is: Among them, x is the original data, μ is the mean of the data column, σ is the standard deviation of the data column, and z is the standardized data; S3: Data Fusion The data fusion formula is used to integrate multi-source heterogeneous data. The data fusion formula is based on the entropy weight method and the correlation analysis between data. It automatically determines the weight of each data source to achieve efficient data fusion. The data fusion formula is as follows: Among them, D fused is the fused data, n is the number of data sources, w i is the weight of the i-th data source, is the normalized data of the i-th data source.

3. A heating load prediction system according to claim 1, characterized in that: The feature selection method adopted by the feature engineering module adopts a recursive feature elimination method or a model-based feature selection method to screen out the feature subset with the most predictive value.

4. A heating load prediction system according to claim 1, characterized in that: The deep long short-term memory network model based on the attention mechanism includes an input layer, an LSTM layer, an attention layer, a fully connected layer and an output layer, and adopts cross-validation and early stopping strategies to prevent model overfitting.

5. A heating load prediction system according to claim 1, characterized in that: The deep learning model construction and training module also introduces a multi-task learning framework to simultaneously predict the heating load and its changing trend.

6. A heating load prediction system according to claim 5, characterized in that: The deep learning model construction and training module adopts a transfer learning strategy to transfer knowledge from data sets in the field of heating load prediction to accelerate the model training process.

7. A heating load prediction system according to claim 1, characterized in that: The heating system scheduling plan includes adjusting the heating temperature, adjusting the flow rate or adjusting the operating time; the energy-saving optimization suggestions include replacing matching energy-saving equipment and optimizing the building insulation performance.

8. A heating load prediction system according to claim 1, characterized in that: The visualization interface provided by the visualization and decision support module uses a combination of charts, dashboards, and maps for intelligent comprehensive display.

9. A heating load prediction system according to claim 1, characterized in that: The visualization and decision support module is also provided with a user interaction unit, and the user interaction unit receives user input instructions, and the instructions include: adjusting prediction parameters and selecting a prediction period; the user interaction unit can also feedback prediction results and model performance information.

10. A heating load prediction system according to claim 1, characterized in that: The visualization and decision support module also includes an abnormal data processing unit for automatically detecting and processing abnormal values ​​or missing values ​​in a data set. The abnormal data processing unit uses a box plot method to identify abnormal data. For the identified abnormal data, deletion, filling or marking as abnormal values ​​can be adopted. In addition, the abnormal data processing unit can also generate an abnormal data report for user review and processing, and can automatically detect and process abnormal data.

Images