Intelligent indoor temperature prediction method based on digital-intelligent double drive

Abstract

The invention discloses an intelligent indoor temperature prediction method based on digital-intelligent double drive, and the method comprises the following steps: employing a multivariable correlation analysis method, and carrying out the evaluation of the correlation between indoor temperature and heat supply metering related data, heat exchange station internal unit related data, meteorological data related factors, building related data, and temperature measurement data; sorting is carried out according to the evaluated correlation result, three variables with the strongest indoor temperature correlation are selected, and clustering classification is carried out on the heat users by adopting a K-means clustering method so as to identify user groups with similar attributes; and based on a clustering classification result, an MLP algorithm is used for modeling for each type of heat users, and indoor temperature prediction of the type of heat users is realized. By accurately predicting the indoor temperature, the heat supply system can adjust operation in advance according to a prediction result, and the situation of excessive heat supply or insufficient heat supply is avoided. Therefore, the utilization efficiency of energy is improved, and the energy consumption and the operation cost can be remarkably reduced.

Description

Technical Field

[0001] The present invention belongs to the technical field of energy prediction and optimization, and particularly relates to an intelligent indoor temperature prediction method based on digital and intelligent dual drive. Background Art

[0002] Indoor temperature prediction is one of the important research directions in the fields of intelligent buildings and energy management. With the continuous improvement of building energy conservation requirements, how to accurately control indoor temperature and optimize energy use has become a key issue in building design and operation management. Traditional heating systems often rely on fixed set temperatures or external weather conditions, lacking precise regulation of the indoor environment and user needs, resulting in energy waste and reduced comfort. Therefore, data-driven indoor temperature prediction methods have gradually become an effective means to improve heating efficiency and user comfort.

[0003] In recent years, with the development of Internet of Things technology, more and more intelligent sensors have been applied to building environment monitoring, capable of real-time collecting various information such as indoor temperature, humidity, heating data, and weather conditions. These large amounts of real-time data provide rich inputs for temperature prediction models. With the support of data science and machine learning technologies, researchers have tried to establish temperature prediction models based on historical data through methods such as regression analysis, neural networks, and K-means clustering. These models can accurately predict indoor temperature changes at different time periods and under different conditions, and then provide optimized regulation schemes for heating systems.

[0004] Indoor temperature prediction technology can not only improve the utilization efficiency of building energy, but also enhance the living comfort of users and reduce energy costs. With the continuous progress of technology and the application of big data, indoor temperature prediction will play an increasingly important role in the field of intelligent buildings.

[0005] However, in the absence of an accurate indoor temperature prediction method, the following problems often exist:

[0006] 1. Energy waste and high operation costs: Without temperature prediction, heating or air-conditioning systems usually rely on fixed schedules or simple external meteorological data for regulation, and cannot be dynamically adjusted according to the actual indoor temperature demand. This may lead to too high or too low temperatures, serious energy waste. The heating system may operate excessively, resulting in unnecessary energy consumption and increased operation costs.

[0007] 2. Inability to cope with dynamic changes and complex environments: Indoor temperature is affected by multiple factors, including weather changes, building insulation performance, human activities, equipment operation conditions, etc. Without temperature prediction, it is difficult for the system to cope with these dynamic changes, and there may be excessive or insufficient heating / cooling, making it difficult to effectively respond to changes in complex environmental factors.

[0008] 3. Decrease in user comfort: Without temperature prediction, the change of indoor temperature cannot be precisely controlled according to actual needs. For example, in the case of weather changes or frequent user activities, the indoor temperature may deviate from the comfortable range, causing discomfort to the occupants or employees, affecting their work efficiency and quality of life. The inability to accurately adjust the temperature control equipment may also lead to a lag in temperature adjustment, affecting comfort. Summary of the Invention

[0009] To solve the above problems, the technical solution adopted by the present invention is: An intelligent indoor temperature prediction method based on digital and intelligent dual drive, including the following steps:

[0010] Adopt the multivariate correlation analysis method to evaluate the correlation between indoor temperature and data related to heat supply metering, data related to units in the heat exchange station, relevant factors of meteorological data, building related data, and temperature measurement data;

[0011] Sort according to the evaluated correlation results, select the three variables with the strongest correlation with indoor temperature, and use the K-means clustering method to cluster and classify heat users to identify user groups with similar attributes;

[0012] Based on the clustering and classification results, use the MLP algorithm to model each type of heat user respectively to achieve the indoor temperature prediction of this type of heat user.

[0013] Further: The evaluation of the correlation is to evaluate the degree of linear correlation.

[0014] Further: The process of sorting according to the evaluated correlation results, selecting the three variables with the strongest correlation with indoor temperature, and using the K-means clustering algorithm to cluster and classify heat users to identify user groups with similar attributes is as follows:

[0015] S21: Take the three variables with the strongest correlation with indoor temperature as the input samples of the K-means clustering algorithm;

[0016] S22: Randomly select k samples as the initial clustering centers;

[0017] S23: For each sample in the dataset, calculate the distance between this sample and the K clustering centers respectively, and assign it to the class corresponding to the clustering center with the smallest distance.

[0018] S24: For each category, recalculate the clustering center of each;

[0019] S25: Repeat the above two steps of S23 and S24 until the clustering centers no longer change, and output the clustering results.

[0020] Further: Based on the clustering classification results, the MLP algorithm is used to model each type of hot user respectively, and the process of realizing the indoor temperature prediction of this type of hot user is as follows:

[0021] S31: According to the clustering results, establish an MLP algorithm model for each category;

[0022] S32: Define the input and output of the MLP algorithm model. The input is the original single-household dataset with the heat exchange station as the unit, and this dataset includes building data, meteorological data, heat supply metering data, temperature measurement data, and heat exchange station unit operation data; the output is the single-household indoor temperature prediction;

[0023] S33: Define the input feature set in groups, perform tensor conversion on the data in the feature set, and normalize the parameters of the dataset after tensor conversion;

[0024] S34: On the basis of normalization, shuffle the dataset, and divide the shuffled dataset into a training set and a test set according to a certain proportion;

[0025] S35: Initialize and define the parameters of the MLP algorithm model, and define the loss function and optimization algorithm of the MLP algorithm model;

[0026] S36: Train according to the set MLP algorithm model parameters. When the termination condition is reached, obtain the trained MLP model network structure and parameters, and output the predicted value of the hot user's room temperature.

[0027] Further: The structure of the MLP algorithm model is determined according to the task characteristics and data characteristics of room temperature prediction;

[0028] The MLP algorithm model includes the number of neurons in the input layer, hidden layer, and output layer, as well as the number of hidden layers and the number of neurons;

[0029] The dimension of the input layer is equal to the number of data characteristics related to room temperature collected;

[0030] The number of hidden layers is set to 2 - 3 layers, and the number of neurons in each layer is between 32 and 128;

[0031] The dimension of the output layer is 1, that is, the predicted room temperature value.

[0032] Further: The loss function of the MLP algorithm model adopts the mean square error, and the optimization algorithm adopts the Adam optimization algorithm.

[0033] Further: The task characteristics of the room temperature prediction include the prediction time scale, the prediction accuracy requirement, and the application scenario of the prediction result;

[0034] In terms of data characteristics, it is determined according to the specific characteristics of heat metering related data, unit related data in the heat exchange station, meteorological data related factors, building related data, and temperature measurement data.

[0035] An intelligent indoor temperature prediction device based on digital and intelligent dual drive, comprising:

[0036] A correlation analysis module: used to evaluate the correlation between indoor temperature and heat metering related data, unit related data in the heat exchange station, meteorological data related factors, building related data, and temperature measurement data by using the multi-variable correlation analysis method;

[0037] A clustering and classification module: used to sort according to the evaluated correlation results, select the three variables with the strongest correlation with indoor temperature, and use the K-means clustering method to cluster and classify heat users to identify user groups with similar attributes;

[0038] A prediction module: used to build models for each type of heat user respectively using the MLP algorithm based on the clustering and classification results to achieve the prediction of the indoor temperature of this type of heat user.

[0039] A readable storage medium stores program modules, and the program modules can be run in a processor to implement the method described in any one of the above.

[0040] An intelligent indoor temperature prediction method based on digital and intelligent dual drive provided by the present invention aims to provide an intelligent energy-saving and carbon reduction strategy driven by big data with supply based on demand, and has the following advantages:

[0041] Optimizing energy consumption and reducing costs: By accurately predicting the indoor temperature, the heating system can adjust its operation in advance according to the prediction results, avoiding overheating or underheating. This not only improves the energy utilization efficiency but also significantly reduces energy consumption and operating costs. For large-scale buildings (such as office buildings, commercial complexes, etc.), temperature prediction can help reduce energy waste and save operating expenses.

[0042] Improving the comfort of living and working environments: Indoor temperature prediction can help accurately control the indoor temperature to ensure that the indoor environment always remains within the comfortable range of users. By adjusting the temperature control equipment in real time, the prediction technology can adjust the temperature change in advance according to dynamic factors such as external weather changes and indoor activities, avoiding affecting the comfort of residents due to too high or too low temperature, and improving the overall experience of living and working environments.

[0043] Intelligent Building Management and Automatic Regulation: Temperature prediction can be combined with a Building Management System (BMS) to form an automatic regulation mechanism. Based on the prediction results, the building management system can achieve intelligent control and automatically adjust the operation modes of air conditioning and heating equipment without manual intervention. This automatic management not only improves the efficiency of building management but also reduces labor costs, promoting the intelligent and green development of buildings. Brief Description of the Drawings

[0044] 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 some embodiments of the present invention. For those of ordinary skill in the art, without creative efforts, other drawings can also be obtained based on these drawings.

[0045] Figure 1 is a classification diagram of data sources;

[0046] Figure 2 is a flowchart of the K-means algorithm;

[0047] Figure 3 is an organizational framework diagram for establishing a room temperature prediction model;

[0048] Figure 4 is a flowchart of the MLP algorithm;

[0049] Figure 5 is an MLP network architecture diagram; Detailed Embodiment

[0050] It should be noted that, without conflict, the embodiments in the present invention and the features in the embodiments can be combined with each other. The following will refer to the drawings and combine the embodiments to describe the present invention in detail.

[0051] To make the objectives, technical solutions, and advantages of the embodiments of the present invention clearer, 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 some embodiments of the present invention, rather than all embodiments. The following description of at least one exemplary embodiment is actually only illustrative and in no way limits the present invention and its application or use. Based on the embodiments of the present invention, all other embodiments obtained by those of ordinary skill in the art without creative efforts belong to the scope of protection of the present invention.

[0052] An intelligent indoor temperature prediction method based on digital and intelligent dual drive includes the following steps:

[0053] S1: Use the multi-variable correlation analysis method to evaluate the correlation between indoor temperature and heating metering related data, unit related data in the heat exchange station, meteorological data related factors, building related data, and temperature measurement data;

[0054] S2: Sort according to the evaluated correlation results, select the three variables with the strongest correlation with indoor temperature, and use the K-means clustering method to cluster and classify heat users to identify user groups with similar attributes;

[0055] S3: Based on the clustering classification results, use the MLP algorithm to model each type of heat user respectively to achieve the prediction of the indoor temperature of this type of heat user.

[0056] The steps S1 / S2 / S3 are executed in sequence;

[0057] Figure 1 It is a data source classification diagram;

[0058] The meteorological data related factors include outdoor temperature, sunshine intensity, wind direction information, wind speed information, precipitation, etc.;

[0059] The unit data includes: primary supply temperature, primary return temperature, primary supply pressure, primary return pressure, primary supply water flow, primary return water flow;

[0060] The heating metering related data includes cumulative flow, cumulative heat, instantaneous heat, instantaneous flow, inlet temperature, outlet temperature;

[0061] The building related data includes: community location, building location, unit, floor, room number, heating area, building materials, orientation, insulation performance;

[0062] The temperature measurement data includes temperature measurement time, temperature measurement value, and indoor equipment usage (such as the on / off state and power of air conditioners and heaters, which directly change the room temperature);

[0063] Furthermore, the evaluation of the correlation is to evaluate the degree of linear correlation. The correlation range is [-1, 1], where the value closer to 0 indicates the weaker the linear correlation; the value closer to 1 or -1 indicates the stronger the linear correlation. A positive value indicates a positive correlation, and a negative value indicates a negative correlation.

[0064] The degree of linear correlation is usually measured by the correlation coefficient, and the most commonly used is the Pearson correlation coefficient. The following are its calculation steps:

[0065] 1. Calculate the mean

[0066] Suppose there are two variables X and Y, and they have n observation values respectively, that is, X = [x1, x2,..., xn and Y = [y_1, y_2,..., y_ n Y = [y1, y2,..., y n .

[0067] Calculate the mean: Calculate the means of variables X and Y respectively:

[0068] 2. Calculate the covariance

[0069] The covariance Cov(X, Y) measures the overall error change trend of two variables, and the formula is:

[0070]

[0071] 3. Calculate the standard deviation

[0072] The formulas for calculating the standard deviations of variables X and Y respectively are: S X and S Y ;

[0073] 4. Calculate the correlation coefficient

[0074]

[0075] The value range of the finally obtained correlation coefficient r is between -1 and 1, and the linear correlation degree is judged based on this. The closer the value is to 0, the weaker the linear correlation; the closer the value is to 1 or -1, the stronger the linear correlation.

[0076] Figure 2 is the flowchart of the K-Means algorithm; K-Means is a commonly used clustering algorithm, and the steps are as follows:

[0077] Determine the number of clusters K: Determine in advance the number K of clusters (categories) into which the data is to be divided. The value of K can be determined according to actual needs;

[0078] Initialize the cluster centers: Randomly select K points in the data space as the initial cluster centers. For example, in the data of a two-dimensional plane, randomly select K coordinate points;

[0079] Assign data points: Calculate the distances from each data point to the K cluster centers (usually using the Euclidean distance), and assign each data point to the cluster represented by the closest cluster center.

[0080] Update the cluster centers: Recalculate the mean of all data points in each cluster, and use this mean as the new cluster center;

[0081] Convergence judgment: Check whether the cluster centers change. If the positions of the cluster centers no longer change or change very little, or the preset number of iterations is reached, the algorithm stops; otherwise, return the assigned data points and continue to repeat the process of assigning data points and updating the cluster centers.

[0082] In summary, the K-means algorithm divides data points into different clusters through continuous iteration, making the data points within each cluster as similar as possible and the data points between different clusters as different as possible.

[0083] Furthermore, the process of sorting according to the evaluated correlation results, selecting the three variables with the strongest correlation with indoor temperature, and using the K-means clustering algorithm to cluster and classify heat users to identify user groups with similar attributes is as follows:

[0084] S21: Use the three variables with the strongest correlation with indoor temperature as the input samples of the K-means clustering algorithm;

[0085] S22: Randomly select k samples as the initial cluster centers C = {c1, c2…c k};

[0086] S23: For each sample x in the dataset, calculate the distance from this sample to the K cluster centers respectively, and

[0087] assign it to the class corresponding to the cluster center with the minimum distance;

[0088] S24: For each category, recalculate the cluster center of each;

[0089]

[0090] where: n is the number of data points included in category c j ;

[0091] S25: Repeat the above two steps of S23 and S24, and perform convergence judgment: Check whether the cluster centers change. If the positions of the cluster centers no longer change or change very little, or the preset number of iterations is reached, until the cluster centers no longer change, the algorithm stops; output the clustering result.

[0092] Combining the clustering of heat users and the prediction of room temperature, first use the K-means clustering algorithm to classify heat users. Heat users have differences in aspects such as house structure, orientation, and usage habits, and these factors will affect the room temperature. Through clustering, heat users with similar characteristics can be grouped into one category, and the group characteristics of different types of heat users can be mined. Then, for each category, use MLP to predict the room temperature respectively, breaking the previous way of unified modeling and prediction for all heat users, being able to consider the characteristics of different user groups, making the prediction model more in line with the actual situation, and improving the prediction accuracy.

[0093] Traditional room temperature prediction methods may not fully consider the differences among heat users, or simply process the overall data, resulting in a large deviation between the prediction result and the actual room temperature. This method clusters and subdivides user groups, allowing the MLP to be trained and predicted for each category, and can better learn the complex relationship between the factors affecting room temperature and room temperature in each category, thereby improving the accuracy of room temperature prediction and providing more reliable data support for the precise control of the heating system, which is of great significance in energy management and optimization.

[0094] Figure 3 It is the organizational framework diagram for establishing the room temperature prediction model;

[0095] Figure 4 It is the flowchart of the MLP algorithm;

[0096] Furthermore: Based on the clustering classification results, the MLP algorithm is used for modeling for each type of heat user respectively, and the process of predicting the indoor temperature of this type of heat user is as follows:

[0097] S31: According to the clustering results, establish the MLP algorithm model for each category;

[0098] S32: Define the input and output of the MLP algorithm model. The input is the original single-household dataset with the heat exchange station as the unit, and this dataset includes building data, meteorological data, heat supply metering data, temperature measurement data, and heat exchange station unit operation data; the output is the prediction of the single-household indoor temperature;

[0099] S33: Define the input feature set in groups, perform tensor conversion on the data in the feature set, and normalize the parameters of the dataset after tensor conversion;

[0100] S34: On the basis of normalization, shuffle the dataset, and divide the shuffled dataset into a training set and a test set according to a certain proportion;

[0101] S35: Initialize and define the parameters of the MLP algorithm model, and define the loss function and optimization algorithm of the MLP algorithm model;

[0102] S36: Train according to the set MLP algorithm model parameters. When the termination condition is reached, obtain the trained MLP model network structure and parameters, and output the predicted value of the heat user's room temperature.

[0103] Figure 5 It is the MLP network architecture diagram; Furthermore: The structure of the MLP algorithm model is determined according to the task characteristics and data characteristics of room temperature prediction;

[0104] The MLP algorithm model includes the number of neurons in the input layer, hidden layer, and output layer, as well as the number of hidden layers and neurons;

[0105] The dimension of the input layer is equal to the number of data features related to room temperature collected; if 8 data features are collected, the dimension of the input layer is 8

[0106] Several hidden layers are set, and the number of hidden layers and the number of neurons in each layer are adjusted through experiments and experience. Generally speaking, for a more complex room temperature prediction task, 2 - 3 hidden layers can be set, and the number of neurons in each layer is between 32 - 128. The dimension of the output layer is 1, that is, the predicted room temperature value. The dimension of the output layer is 1, that is, the predicted room temperature value.

[0107] Activation function selection: In the hidden layer, the ReLU (Rectified Linear Unit) activation function is selected. The expression of the ReLU function is f(x) = max(0, x), which can effectively solve the problem of gradient disappearance and introduce non - linear factors into the model, enabling it to learn the complex relationship between the input data and room temperature. In the output layer, since the room temperature is a continuous value, no activation function is used, and the predicted value is directly output.

[0108] The weights and biases of the model are initialized using the method of random initialization. Specifically, the weight matrix is initialized using a Gaussian distribution with a mean of 0 and a small standard deviation (such as 0.01), and the bias vector is initialized to 0. This initialization method can break the symmetry between neurons and avoid the situation where all neurons learn the same features during the training process.

[0109] Furthermore, the loss function of the MLP algorithm model adopts the mean square error, and the optimization algorithm adopts the Adam optimization algorithm.

[0110] The task characteristics of the room temperature prediction include the prediction time scale (short - term, long - term), the accuracy requirement of the prediction, and the application scenarios of the prediction results (such as smart home control, building energy consumption management, etc.);

[0111] In terms of data features, it is determined according to the specific characteristics and features of heat metering - related data, unit - related data in the heat exchange station, meteorological data - related factors, building - related data, and temperature measurement data.

[0112] An intelligent indoor temperature prediction device based on digital - intelligence dual - drive includes:

[0113] Correlation analysis module: used to evaluate the correlation between indoor temperature and heat metering - related data, unit - related data in the heat exchange station, meteorological data - related factors, building - related data, and temperature measurement data by using the multi - variable correlation analysis method;

[0114] Clustering and classification module: used to sort according to the evaluated correlation results, select the three variables with the strongest indoor temperature correlation, and adopt the K-means clustering method to cluster and classify the heat users to identify user groups with similar attributes;

[0115] Prediction module: used to model each type of heat user separately using the MLP algorithm based on the clustering and classification results to achieve the indoor temperature prediction of this type of heat user.

[0116] A readable storage medium stores program modules, and the program modules can be implemented to realize the method described in any one of the above when running in a processor.

[0117] Finally, it should be noted that: the above embodiments are only used to illustrate the technical solutions of the present invention, rather than to limit them; although the present invention has been described in detail with reference to the foregoing embodiments, those of ordinary skill in the art should understand that: they can still modify the technical solutions described in the foregoing embodiments, or perform equivalent replacements on some or all of the technical features; and these modifications or replacements do not make the essence of the corresponding technical solutions deviate from the scope of the technical solutions of the embodiments of the present invention.

Claims

1. An intelligent indoor temperature prediction method based on digital and intelligent dual drive, characterized in that: It includes the following steps: Adopt the multi-variable correlation analysis method to evaluate the correlations among indoor temperature and heat metering related data, unit related data in the heat exchange station, meteorological data related factors, building related data, and temperature measurement data; Sort according to the evaluated correlation results, select the three variables with the strongest correlation with indoor temperature, and use the K-means clustering method to cluster and classify heat users to identify user groups with similar attributes; Based on the clustering and classification results, use the MLP algorithm to model each type of heat user respectively to realize the indoor temperature prediction of this type of heat user.

2. The intelligent indoor temperature prediction method based on digital and intelligent dual drive according to claim 1, wherein: The evaluation of the correlation is to evaluate the degree of linear correlation.

3. The intelligent indoor temperature prediction method based on digital and intelligent dual drive according to claim 1, characterized in that: The process of sorting according to the evaluated correlation results, selecting the three variables with the strongest correlation with indoor temperature, and using the K-means clustering algorithm to cluster and classify heat users to identify user groups with similar attributes is as follows: S21: Take the three variables with the strongest correlation with indoor temperature as the input samples of the K-means clustering algorithm; S22: Randomly select k samples as the initial clustering centers; S23: For each sample in the dataset, calculate the distances of this sample to the K clustering centers respectively, and assign it to the class corresponding to the clustering center with the minimum distance; S24: For each category, recalculate the clustering center of each; S25: Repeat the above two steps of S23 and S24 until the clustering centers no longer change, and output the clustering result.

4. The intelligent indoor temperature prediction method based on digital and intelligent dual drive according to claim 1, characterized in that: The process of using the MLP algorithm to model each type of heat user respectively based on the clustering and classification results to realize the indoor temperature prediction of this type of heat user is as follows: S31: According to the clustering results, establish the MLP algorithm model for each category; S32: Define the input and output of the MLP algorithm model. The input is the single-household original dataset in units of heat exchange stations, and this dataset includes building data, meteorological data, heat metering data, temperature measurement data, and heat exchange station unit operation data; the output is the single-household indoor temperature prediction; S33: Define the input feature set in groups, perform tensor conversion on the data in the feature set, and perform parameter normalization on the dataset after tensor conversion; S34: On the basis of normalization, shuffle the dataset, and divide the shuffled dataset into a training set and a test set according to a ratio; S35: Initialize and define the parameters of the MLP algorithm model, and define the loss function and optimization algorithm of the MLP algorithm model; S36: Train according to the set MLP algorithm model parameters. When the termination condition is reached, obtain the trained MLP model network structure and parameters, and output the predicted value of the indoor temperature of heat users.

5. A smart indoor temperature prediction method based on digital and intelligent dual drive according to claim 1, characterized in that: The structure of the MLP algorithm model is determined according to the task characteristics and data characteristics of indoor temperature prediction; The MLP algorithm model includes the number of neurons in the input layer, hidden layer, output layer, the number of hidden layers, and the number of neurons; The dimension of the input layer is equal to the number of data features related to room temperature collected; The number of hidden layers is set to 2 - 3 layers, and the number of neurons in each layer is between 32 and 128; The dimension of the output layer is 1, that is, the predicted room temperature value.

6. The intelligent indoor temperature prediction method based on digital and intelligent dual drive according to claim 1, characterized in that: The loss function of the MLP algorithm model uses mean squared error, and the optimization algorithm uses the Adam optimization algorithm.

7. A method for predicting intelligent indoor temperature based on digital and intelligent dual drive according to claim 1, characterized in that: The task characteristics of the room temperature prediction include the prediction time scale, the prediction accuracy requirement, and the application scenario of the prediction result. In terms of data features, it is determined according to the specific characteristics and features of heat metering related data, unit related data in the heat exchange station, meteorological data related factors, building related data, and temperature measurement data.

8. An intelligent indoor temperature prediction device based on digital and intelligent dual drive, characterized in that: It includes: Correlation analysis module: used to evaluate the correlation between indoor temperature and heat metering related data, unit related data in the heat exchange station, meteorological data related factors, building related data, and temperature measurement data by using the multivariate correlation analysis method. Clustering and classification module: used to sort according to the evaluated correlation results, select the three variables with the strongest correlation with indoor temperature, and use the K-means clustering method to cluster and classify heat users to identify user groups with similar attributes. Prediction module: used to build models for each type of heat user respectively using the MLP algorithm based on the clustering and classification results to achieve the prediction of the indoor temperature of this type of heat user.

9. A readable storage medium stores program modules, characterized in that, The running of the program module in the processor can implement the method according to any one of claims 1-7.

Images