Training method of energy consumption probability prediction model, energy consumption probability prediction method, computer device and readable storage medium

Abstract

The application provides a kind of energy consumption probability prediction model training method, energy consumption probability prediction method, computer equipment and readable storage medium, related to building energy management and intelligent prediction technical field.The method comprises: extracting features from historical building data of target building in a predetermined historical time period to obtain sample energy consumption related features;According to sample energy consumption related features and real energy consumption probability, a sample dataset is constructed;The sample dataset is divided into multiple sub-datasets;According to multiple sub-datasets, multiple initial energy consumption prediction models are trained respectively to obtain multiple trained energy consumption prediction sub-models;According to multiple trained energy consumption prediction sub-models, a target energy consumption prediction model is constructed.The application obtains a target energy consumption prediction model according to multiple sample energy consumption related features, so that the prediction interval coverage is high, and it is suitable for dynamic load change, and then the accurate prediction demand of building energy consumption is realized.

Description

Technical Field

[0001] This application relates to the field of building energy management and intelligent prediction technology, and more specifically, to a training method for an energy consumption probability prediction model, an energy consumption probability prediction method, a computer device, and a readable storage medium. Background Technology

[0002] With the continued deterioration of the global ecological environment, building energy consumption has gradually become a global focus. Among these, HVAC systems account for approximately 40% of total building energy consumption. Building carbon emissions account for 25%–33% of all carbon emissions. Meanwhile, frequent high-temperature weather events lead to a surge in electricity demand for building HVAC systems, exacerbating the load on the power grid. Therefore, accurate prediction of building HVAC energy consumption is of great significance for building energy conservation and the stable operation of the power grid.

[0003] Currently, most building energy simulation software relies on fixed schedules defined by the American Society of Heating, Refrigerating and Air-Conditioning Engineers (ASHEE) to simply define building energy usage.

[0004] However, this approach ignores the randomness of weather changes and fluctuations in pedestrian traffic, resulting in low coverage of the prediction range and an error range of 50% to 150% between predicted and actual energy consumption. It cannot adapt to dynamic load changes and therefore cannot meet the need for accurate prediction of building energy consumption. Summary of the Invention

[0005] The purpose of this application is to provide a training method for an energy consumption probability prediction model, an energy consumption probability prediction method, a computer device, and a readable storage medium, so as to obtain a target energy consumption prediction model based on the energy consumption correlation characteristics of multiple samples, so as to achieve high coverage of the prediction interval and adapt to dynamic load changes, thereby realizing the need for accurate prediction of building energy consumption.

[0006] To achieve the above objectives, the technical solutions adopted in the embodiments of this application are as follows: In a first aspect, embodiments of this application provide a training method for an energy consumption probability prediction model, comprising: Feature extraction is performed on the historical building data of the target building within a preset historical time period to obtain the sample energy consumption related features of the preset historical time period; A sample dataset is constructed based on the energy consumption characteristics of the samples during the preset historical time period and the actual energy consumption probability of the HVAC system in the target building during the preset historical time period. The sample dataset is divided into multiple sub-datasets, and the multiple sub-datasets correspond to multiple sub-time periods within the preset historical time period; Based on the multiple subset datasets, multiple initial energy consumption prediction models are trained to obtain multiple trained energy consumption prediction sub-models. Based on the multiple trained energy consumption prediction sub-models, a target energy consumption prediction model is constructed.

[0007] Optionally, the sample energy consumption-related features include: sample linear correlation features and sample nonlinear correlation features; the step of extracting features from historical building data of the target building within a preset historical time period to obtain the sample energy consumption-related features for the preset historical time period includes: Based on the actual energy consumption probability, a preset linear correlation analysis method is used to extract features from the historical building data to obtain sample linear correlation features; Based on the actual energy consumption probability, a preset nonlinear correlation analysis method is used to extract features from the historical building data to obtain sample nonlinear correlation features; The sample energy consumption correlation features are obtained by fusing the linear correlation features and the nonlinear correlation features of the sample.

[0008] Optionally, the step of extracting features from the historical building data based on the actual energy consumption probability using a preset linear correlation analysis method to obtain sample linear correlation features includes: Using the aforementioned preset linear correlation analysis method, the Pearson correlation coefficients of various feature data in the historical building data in the actual energy consumption probability are calculated respectively; Based on the Pearson correlation coefficient, linear correlation features of the samples are selected from various feature data in the historical building data.

[0009] Optionally, the step of extracting features from the historical building data based on the actual energy consumption probability using a preset nonlinear correlation analysis method to obtain sample nonlinear correlation features includes: Using the aforementioned preset nonlinear correlation analysis method, the Spearman rank correlation coefficients of various feature data in the historical building data in the true energy consumption probability are calculated respectively; Based on the Spearman rank correlation coefficient, the nonlinear correlation features of the samples are selected from various feature data in the historical building data.

[0010] Optionally, each initial energy consumption prediction model integrates: an initial long short-term memory (LSTM) network and an initial conformal quantile regression (CQR) prediction model; The step involves training multiple initial energy consumption prediction models based on the multiple subset datasets to obtain multiple trained energy consumption prediction sub-models, including: The initial LSTM network in each initial energy consumption prediction model is used to perform time-series processing on the energy consumption-related features of samples in each subset to obtain the time-series feature information corresponding to each subset. Using the initial CQR prediction model in each initial energy consumption prediction model, energy consumption prediction is performed on the time-series feature information corresponding to each subset of data to obtain the predicted energy consumption probability interval corresponding to each subset of data. Based on the actual energy consumption probability of each sub-dataset corresponding to the sub-time period and the corresponding predicted energy consumption probability interval, the initial LSTM network and the initial CQR prediction model in each initial energy consumption prediction model are trained to obtain a trained energy consumption prediction sub-model. The trained energy consumption prediction sub-model includes a trained LSTM network and a trained CQR prediction network.

[0011] Optionally, constructing the target energy consumption prediction model based on the plurality of trained energy consumption prediction sub-models includes: The target energy consumption prediction model is obtained by performing ensemble learning training on the multiple trained energy consumption prediction sub-models.

[0012] Optionally, the step of ensemble learning and training the plurality of trained energy consumption prediction sub-models to obtain the target energy consumption prediction model includes: Each trained energy consumption prediction sub-model is used to predict the energy consumption-related features of the samples in each sub-time period, thereby obtaining the predicted energy consumption probability interval for each trained energy consumption prediction sub-model for each sub-time period. Based on the predicted energy consumption probability range of each post-trained energy consumption prediction sub-model for each sub-time period and the actual energy consumption probability of each sub-time period, the prediction deviation of each post-trained energy consumption prediction sub-model for each sub-time period is obtained. Based on the prediction deviations of the multiple trained energy consumption prediction sub-models for each sub-time period, a prediction residual set is constructed; Based on the predicted residual set, the target energy consumption prediction model is constructed.

[0013] Secondly, embodiments of this application provide an energy consumption probability prediction method, including: Feature extraction is performed on the predicted building data of the building to be predicted in a preset future time period to obtain the predicted energy consumption related features for the preset future time period. A pre-trained target energy consumption prediction model is used to predict energy consumption based on the predicted energy consumption related characteristics of the preset future time period, thereby obtaining the predicted energy consumption probability range of the building to be predicted for heating, ventilation and air conditioning within the preset future time period.

[0014] Thirdly, embodiments of this application provide a computer device, including: a processor; a memory for storing processor-executable instructions; wherein the processor executes the executable instructions to implement the training method for the energy consumption probability prediction model as described in any one of the first aspects and the energy consumption probability prediction method as described in the second aspect.

[0015] Fourthly, embodiments of this application provide a computer-readable storage medium storing a computer program, which, when executed by a processor, performs the steps of the energy consumption probability prediction model training method as described in any one of the first aspects and the energy consumption probability prediction method as described in the second aspect.

[0016] Compared with existing technologies, the training method, energy consumption probability prediction method, computer equipment, and readable storage medium of the energy consumption probability prediction model provided in this application have the following technical effects: This application provides a training method for an energy consumption probability prediction model, an energy consumption probability prediction method, a computer device, and a readable storage medium, relating to the field of building energy management and intelligent prediction technology. The method includes: extracting features from historical building data of a target building within a preset historical time period to obtain sample energy consumption-related features for that period. This helps reduce the dimensionality of subsequent sample data, avoids interference from irrelevant data on the subsequent target energy consumption prediction model, and improves the training efficiency and accuracy of the target energy consumption prediction model. Based on the sample energy consumption-related features of the preset historical time period and the actual energy consumption probability of the HVAC system in the target building within that preset historical time period, a sample dataset is constructed to ensure that all data in the dataset is relevant to energy consumption prediction. The sample dataset is divided into multiple subsets, each corresponding to a sub-time period within the preset historical time period, giving each subset the energy consumption characteristics of a specific time period. Multiple initial energy consumption prediction models are trained on these subsets to obtain multiple trained energy consumption prediction sub-models, each exhibiting good predictive ability within its corresponding sub-time period. Integrating these energy consumption prediction sub-models to construct the target energy consumption prediction model combines the advantages of each model and adapts to a wider range of energy consumption scenarios. Meanwhile, a target energy consumption prediction model is obtained based on the energy consumption correlation characteristics of multiple samples, which makes the prediction interval highly covered and adaptable to dynamic load changes, thereby achieving the need for accurate prediction of building energy consumption. Attached Figure Description

[0017] To more clearly illustrate the technical solutions of the embodiments of this application, the accompanying drawings used in the embodiments of this application will be briefly introduced below. It should be understood that the following drawings only show some embodiments of this application and should not be regarded as a limitation of the scope. For those skilled in the art, other related drawings can be obtained based on these drawings without creative effort.

[0018] Figure 1 A schematic diagram of the structure of a training system for an energy consumption probability prediction model provided in an embodiment of this application; Figure 2 A schematic diagram of the structure of a computer device provided in an embodiment of this application; Figure 3 A flowchart illustrating a training method for an energy consumption probability prediction model provided in this application embodiment. Figure 1 ; Figure 4 A flowchart illustrating a training method for an energy consumption probability prediction model provided in this application embodiment. Figure 2 ; Figure 5 A flowchart illustrating a training method for an energy consumption probability prediction model provided in this application embodiment. Figure 3 ; Figure 6 A flowchart illustrating a training method for an energy consumption probability prediction model provided in this application embodiment. Figure 4 ; Figure 7 A flowchart illustrating a training method for an energy consumption probability prediction model provided in this application embodiment. Figure 5 ; Figure 8 A flowchart illustrating a training method for an energy consumption probability prediction model provided in this application embodiment. Figure 6 ; Figure 9 A flowchart illustrating a training method for an energy consumption probability prediction model provided in this application embodiment. Figure 7 ; Figure 10 A flowchart illustrating an energy consumption probability prediction method provided in an embodiment of this application; Figure 11 A schematic diagram of the structure of a training device for an energy consumption probability prediction model provided in an embodiment of this application; Figure 12 This is a schematic diagram of the structure of an energy consumption probability prediction device provided in an embodiment of this application. Detailed Implementation

[0019] The technical solutions of the embodiments of this application will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of this application, and not all embodiments. Some implementation methods of this application will be described in detail below with reference to the accompanying drawings. In the absence of conflict, the following embodiments and features in the embodiments can be combined with each other.

[0020] Figure 1 This is a schematic diagram of the structure of a training system for an energy consumption probability prediction model provided in an embodiment of this application. Figure 1 As shown, the training system 100 for the energy consumption probability prediction model may include: a feature extraction and sample dataset construction unit 110, an energy consumption prediction sub-model construction unit 120, and a target energy consumption prediction model construction unit 130.

[0021] The feature extraction and sample dataset construction unit 110 is used to extract features from historical building data, construct a sample dataset, and obtain multiple sub-datasets Ie. The energy consumption prediction sub-model construction unit 120 is used to construct multiple initial energy consumption prediction models based on the multiple sub-datasets Ie. Training is performed to obtain multiple trained energy consumption prediction sub-models LCQR; the target energy consumption prediction model building unit 130 is used to build the target energy consumption prediction model based on the multiple trained energy consumption prediction sub-models. We constructed the target energy consumption prediction model LECQR.

[0022] It should be noted that the training system 100 for the above-mentioned energy consumption probability prediction model can be integrated into the computer device 200. The computer device 200 can be selected according to the actual situation. For example, the computer device 200 can be a computer, etc.

[0023] Furthermore, Figure 2 This is a schematic diagram of the structure of a computer device provided in an embodiment of this application. Figure 2 As shown, the computer device 200 may include a processor 210 and a memory 220.

[0024] The memory 220 stores machine-executable instructions that can be executed by the processor 210. When the computer device 200 is running, these machine-readable instructions are executed. The processor 210 and the memory 220 communicate via a bus. The processor 210 can execute the machine-executable instructions to implement the training method for the energy consumption probability prediction model and the energy consumption probability prediction method.

[0025] The memory 220, processor 210, and bus components are electrically connected directly or indirectly to enable data transmission or interaction. For example, these components can be electrically connected via one or more communication buses or signal lines. The mobile storage device includes at least one software function module that can be stored in the memory 220 or embedded in the operating system (OS) of the microgrid controller in the form of software or firmware. The processor 210 is used to execute executable modules stored in the memory 220, such as the training method for the energy consumption probability prediction model of the mobile storage medium and the software function modules and computer programs included in the energy consumption probability prediction method.

[0026] The memory 220 may be, but is not limited to, random access memory (RAM), read-only memory (ROM), programmable read-only memory (PROM), erasable programmable read-only memory (EPROM), electrically erasable programmable read-only memory (EEPROM), etc.

[0027] The training method and energy consumption probability prediction method of the energy consumption probability prediction model provided in the embodiments of this application can be executed by the processor 210 in the computer device 200. The training method and energy consumption probability prediction method of the energy consumption probability prediction model provided in the above embodiments of this application will be explained and described in detail below with reference to the accompanying drawings. Figure 3 A flowchart illustrating a training method for an energy consumption probability prediction model provided in this application embodiment. Figure 1 .like Figure 3 As shown, the method may include: S301. Extract features from the historical building data of the target building within a preset historical time period to obtain sample energy consumption-related features for the preset historical time period.

[0028] The preset historical time period can be selected according to the actual situation. For example, the preset historical time period can be the past three months.

[0029] In one possible implementation, refer to the above. Figure 1The system extracts features from the temporal and environmental data of the target building within a preset historical time period to obtain sample energy consumption-related features for that period. Environmental data may include ambient temperature, humidity, wind speed, wind direction, and pedestrian traffic. Temporal data may include weekdays, hours, holidays, and weekdays.

[0030] S302. Construct a sample dataset based on the energy consumption characteristics of samples during a preset historical time period and the actual energy consumption probability of HVAC in the target building during the preset historical time period.

[0031] In one possible approach, the energy consumption-related features of samples within a preset historical time period are combined with the actual energy consumption probability of the HVAC system in the target building within the preset historical time period to construct a sample dataset. This enables the subsequent target energy consumption prediction model to consider both energy consumption-related factors and energy consumption uncertainties during the training process, thereby improving the target energy consumption prediction model's ability to predict energy consumption probabilities.

[0032] It should be noted that the actual energy consumption probability of the preset historical time period is used to illustrate that the energy consumption-related features of the sample obtained within the preset historical time period are features associated with energy consumption prediction.

[0033] S303. Divide the sample dataset into multiple sub-datasets, each sub-dataset corresponding to a sub-time period within a preset historical time period.

[0034] In one possible implementation, the sample dataset is divided into multiple subsets based on a preset historical time period. These subsets may include: test set samples. Training set samples ~e and validation set samples The multiple subsets correspond to multiple sub-time periods within a preset historical time period.

[0035] Among them, the test set samples Data used to evaluate the performance of the energy consumption prediction sub-model. Test set samples. In the test set samples Used after training and validation to evaluate test set samples. Its generalization ability. Training set samples ~e is the data used to train the initial energy consumption prediction model. This initial energy consumption prediction model learns from the training set samples. The internal parameters are constructed by relating the features and labels in ~e. Validation set samples. Data used to adjust the hyperparameters of the energy consumption prediction sub-model and evaluate the performance of the energy consumption prediction model. This validation set sample... It is typically used during training to prevent overfitting.

[0036] S304. Based on multiple subset datasets, train multiple initial energy consumption prediction models to obtain multiple trained energy consumption prediction sub-models.

[0037] In one possible implementation, multiple initial energy consumption prediction models are applied separately based on multiple subsets of data sets. Train each initial energy consumption prediction model to make it more efficient. (like Figure 1 In Focusing on the energy consumption characteristics of different sub-time periods helps improve the initial energy consumption prediction model. The adaptability to energy consumption changes over different time periods leads to the development of multiple trained energy consumption prediction sub-models, LCQR.

[0038] S305. Construct a target energy consumption prediction model based on multiple trained energy consumption prediction sub-models.

[0039] In one possible implementation, based on test set samples Input the energy consumption prediction sub-model LCQR, obtain the outputs of multiple trained energy consumption prediction sub-models LCQR, and then obtain the target energy consumption prediction model LECQR.

[0040] This application provides a training method for an energy consumption probability prediction model. Feature extraction is performed on historical building data of the target building within a preset historical time period to obtain sample energy consumption-related features for that period. This helps reduce the dimensionality of subsequent sample data, avoids interference from irrelevant data on the subsequent target energy consumption prediction model, and improves the training efficiency and accuracy of the target energy consumption prediction model. Based on the sample energy consumption-related features of the preset historical time period and the actual energy consumption probability of the HVAC system in the target building within that preset historical time period, a sample dataset is constructed to ensure that all data in the dataset is relevant to energy consumption prediction. The sample dataset is divided into multiple subsets, each corresponding to a sub-time period within the preset historical time period, giving each subset specific energy consumption characteristics for that time period. Multiple initial energy consumption prediction models are trained on these subsets to obtain multiple trained energy consumption prediction sub-models, each exhibiting good predictive ability within its corresponding sub-time period. Integrating these energy consumption prediction sub-models to construct the target energy consumption prediction model combines the advantages of each model and adapts to a wider range of energy consumption scenarios. Meanwhile, a target energy consumption prediction model is obtained based on the energy consumption correlation characteristics of multiple samples, which makes the prediction interval highly covered and adaptable to dynamic load changes, thereby achieving the need for accurate prediction of building energy consumption.

[0041] Optionally, the above-mentioned sample energy consumption related features include: sample linear correlation features and sample nonlinear correlation features. The sample energy consumption related features may include: ambient temperature, daily maximum temperature, daily minimum temperature, perceived temperature, relative humidity, dew point temperature, daily total rainfall, 10-minute average wind speed, gust wind speed, wind direction, wind direction degree, number of hours in a day, number of days in a week, number of days in a month, number of weeks, holiday characteristics, lighting and socket energy consumption, pedestrian traffic, and historical energy consumption of HVAC systems, etc. The sample linear correlation features are the linear sample features among the above-mentioned sample energy consumption related features; for example, running time is a sample linear correlation feature. The sample nonlinear correlation features are the nonlinear sample features among the above-mentioned sample energy consumption related features; for example, ambient temperature is a sample nonlinear correlation feature.

[0042] Figure 4 A flowchart illustrating a training method for an energy consumption probability prediction model provided in this application embodiment. Figure 2 .like Figure 4 As shown, the above method extracts features from historical building data of the target building within a preset historical time period to obtain sample energy consumption-related features for the preset historical time period, which may include: S401. Based on the actual energy consumption probability, a preset linear correlation analysis method is used to extract features from historical building data to obtain sample linear correlation features.

[0043] The preset linear correlation analysis method can be selected according to the actual situation. For example, the preset linear correlation analysis method can be selected as the Pearson correlation coefficient analysis method.

[0044] In one possible approach, based on the actual energy consumption probability, a pre-defined linear correlation analysis method (such as Pearson correlation coefficient analysis) is used to extract features from the time and environmental data in historical building data to obtain sample linear correlation features. This allows for the elimination of interference from some accidental factors based on the actual energy consumption probability, making the extracted sample linear correlation features more representative and reliable, thereby providing an accurate basis for subsequent energy consumption prediction.

[0045] S402. Based on the actual energy consumption probability, a preset nonlinear correlation analysis method is used to extract features from historical building data to obtain sample nonlinear correlation features.

[0046] The preset nonlinear correlation analysis method can be selected according to the actual situation. For example, the preset nonlinear correlation analysis method can be selected as the Spearman rank correlation coefficient analysis method.

[0047] In one possible approach, to comprehensively assess the correlation between features, a pre-defined nonlinear correlation analysis method (such as Spearman's rank correlation coefficient analysis method) can be used to extract features from the time and environmental data in historical building data based on the actual energy consumption probability. This yields sample nonlinear correlation features, which can better adapt to complex and ever-changing realities, thereby improving the model's adaptability to energy consumption prediction under different operating conditions.

[0048] S403. The linear correlation features and nonlinear correlation features of the samples are fused to obtain the energy consumption correlation features of the samples.

[0049] One possible approach involves fusing sample linear correlation features and sample nonlinear correlation features. The linear correlation features provide a stable and easily understood foundation for energy consumption relationships, while the nonlinear correlation features supplement the description of complex energy consumption relationships. The combined sample energy consumption correlation features provide richer information for subsequent target energy consumption prediction models, helping to improve the prediction accuracy and reduce prediction errors. Whether under normal operating conditions or special conditions such as extreme weather, the fused sample energy consumption correlation features enable the target energy consumption prediction model to make more accurate energy consumption predictions.

[0050] The training method for the energy consumption probability prediction model provided in this application includes a method where sample energy consumption-related features can be composed of sample linear correlation features and sample nonlinear correlation features. Based on the actual energy consumption probability, a pre-defined linear correlation analysis method is used to extract features from historical building data to obtain sample linear correlation features. Similarly, based on the actual energy consumption probability, a pre-defined nonlinear correlation analysis method is used to extract features from historical building data to obtain sample nonlinear correlation features. The sample linear correlation features and sample nonlinear correlation features are then fused to obtain the sample energy consumption-related features. Therefore, the sample energy consumption-related features obtained in this application allow the target energy consumption prediction model to no longer be limited to dependence on specific linear or nonlinear relationships, enabling it to better adapt to different types of buildings, different usage scenarios, and different environmental conditions. This enhances the generalization ability of the target energy consumption prediction model, enabling it to effectively predict energy consumption across a wider range.

[0051] Figure 5 A flowchart illustrating a training method for an energy consumption probability prediction model provided in this application embodiment. Figure 3 .like Figure 5 As shown, the above method uses a preset linear correlation analysis method to extract features from historical building data based on the actual energy consumption probability, obtaining sample linear correlation features, which may include: S501. Using a preset linear correlation analysis method, calculate the Pearson correlation coefficient between the actual energy consumption probability and various characteristic data in the historical building data.

[0052] In one possible approach, a pre-defined linear correlation analysis method (such as Pearson correlation coefficient analysis) is used to calculate the Pearson correlation coefficients between the actual energy consumption probability and various characteristic data in historical building data. .

[0053] The Pearson correlation coefficient can be expressed by the following formula (1): Formula (1) In the above formula (1), i and j are both positive integers from 1 to 19, i.e. . and These represent the values ​​of different energy consumption-related features of the samples. Specifically, These represent ambient temperature, daily maximum temperature, daily minimum temperature, perceived temperature, relative humidity, dew point temperature, daily total rainfall, 10-minute average wind speed, gust wind speed, wind direction, wind direction degree, number of hours in a day, number of days in a week, number of days in a month, week number, holiday characteristics, lighting and socket energy consumption, pedestrian traffic, and historical energy consumption of HVAC systems.

[0054] In the above formula (1), n ​​represents the number of samples (i.e., the number of observations) of energy consumption-related features for each sample, while and They represent and The average value.

[0055] S502. Based on the Pearson correlation coefficient, select samples with linear correlation features from various feature data in historical building data.

[0056] In one possible implementation, linear correlation features of samples are selected from various feature data in historical building data based on the Pearson correlation coefficient expressed by formula (1) above. The Pearson correlation coefficient is a commonly used method for measuring the strength of linear relationships between variables, but it is sensitive to outliers.

[0057] The training method for the energy consumption probability prediction model provided in this application employs a pre-defined linear correlation analysis method to calculate the Pearson correlation coefficients between the actual energy consumption probability and various feature data in historical building data. Based on the Pearson correlation coefficients, linearly correlated features are selected from various feature data in the historical building data. By setting a certain Pearson correlation coefficient threshold, features linearly correlated with energy consumption can be effectively selected from numerous historical building data features, avoiding interference from irrelevant features on the subsequent target energy consumption prediction model. This allows the target energy consumption prediction model to focus on the factors that truly affect energy consumption. Therefore, the target energy consumption prediction model constructed based on the selected linearly correlated features can more accurately capture changes in the linear relationship between energy consumption and related features.

[0058] Figure 6 A flowchart illustrating a training method for an energy consumption probability prediction model provided in this application embodiment. Figure 4 .like Figure 6 As shown, the above method uses a preset nonlinear correlation analysis method to extract features from historical building data based on the actual energy consumption probability, obtaining sample nonlinear correlation features, which may include: S601. Using a pre-defined nonlinear correlation analysis method, calculate the Spearman rank correlation coefficients between the actual energy consumption probability and various characteristic data in the historical building data.

[0059] In one possible approach, a pre-defined nonlinear correlation analysis method (such as Spearman's rank correlation coefficient analysis) is used to calculate the Spearman's rank correlation coefficients between the actual energy consumption probability and various characteristic data in the historical building data. .

[0060] The Spearman rank correlation coefficient can be expressed by the following formula (2): Formula (2) in, This represents the grade difference between each pair of observations. It can be expressed by the following formula (3): Formula (3) The result of formula (3) can be input into formula (2) to obtain the Spearman rank correlation coefficient. .

[0061] S602. Based on the Spearman rank correlation coefficient, select samples with nonlinear correlation features from various feature data in historical building data.

[0062] In one possible implementation, the Spearman rank correlation coefficient, as expressed in formula (2) above, is used to select samples with nonlinear correlation characteristics from various feature data in historical building data. The Spearman rank correlation coefficient primarily measures the monotonic relationship between two variables and does not require the data to satisfy a linear relationship or a specific distribution pattern.

[0063] The training method for the energy consumption probability prediction model provided in this application employs a pre-defined nonlinear correlation analysis method to calculate the Spearman rank correlation coefficients between the actual energy consumption probability and various feature data in historical building data. Based on the Spearman rank correlation coefficients, nonlinear correlation features are selected from various feature data in the historical building data. By selecting samples with nonlinear correlation features, richer input information can be provided to the target energy consumption prediction model. Therefore, this application considers both linear and nonlinear correlation sample features when constructing the target energy consumption prediction model, which can further optimize the model. This target energy consumption prediction model can handle linear and nonlinear relationships according to the characteristics of different features, thereby improving the prediction accuracy and stability of the target energy consumption prediction model.

[0064] Optionally, each of the above initial energy consumption prediction models integrates: an initial long short-term memory (LSTM) network and an initial conformal quantile regression (CQR) prediction model.

[0065] The Initial Long Short-Term Memory (LSTM) network is a novel architecture of recurrent neural networks designed to overcome the vanishing or exploding gradient problems. The internal structure of the LSTM network is based on a series of interconnected units.

[0066] Figure 7 A flowchart illustrating a training method for an energy consumption probability prediction model provided in this application embodiment. Figure 5 .like Figure 7 As shown, the above method trains multiple initial energy consumption prediction models based on multiple subsets of data, resulting in multiple trained energy consumption prediction sub-models, which may include: S701. Use the initial LSTM network in each initial energy consumption prediction model to perform time-series processing on the energy consumption-related features of the samples in each subset to obtain the time-series feature information corresponding to each subset.

[0067] In one possible implementation, the initial LSTM network in each initial energy consumption prediction model is used to perform time-series processing on the energy consumption-related features of samples in each subset, thereby obtaining the time-series feature information corresponding to each subset.

[0068] The Initial Long Short-Term Memory (LSTM) network is equipped with a cell state feature to store information, which enables it to learn temporal information across multiple time scales. Each cell in the LSTM network contains three main gates: a forget gate, an input gate, and an output gate. The forget gate controls the previous time step. The system retains pre-defined historical time-series features. The input gate identifies historical features crucial for predicting HVAC energy consumption at the current time step t. The output gate controls the current hidden state. The propagation of the initial Long Short-Term Memory (LSTM) network can be represented by the following formula (4): Formula (4) in, This represents the input information at time step t; , and These represent the states of the input gate, forget gate, and output gate at time step t, respectively; the activation function of the fully connected layer is the sigmoid function. ; represents the hidden state from the previous time step t-1; w represents the weights of different gates; b represents the bias terms of different gates.

[0069] The initial Long Short-Term Memory (LSTM) network is also equipped with candidate units, which provide new information to the memory units. This information may replace or supplement the existing information in the memory units. The calculation of the candidate unit is similar to the gate calculation steps in the above formula (4), but the tangent function is used. At time step t, it can be expressed by the following formulas (5), (6) and (7): Formula (5) Formula (6) Formula (7) in, This indicates the generation of candidate information in the candidate unit; Indicates the current state of the cell; This indicates the hidden state at the current moment.

[0070] S702. Using the initial CQR prediction model in each initial energy consumption prediction model, energy consumption is predicted for the time-series feature information corresponding to each subset of data, and the predicted energy consumption probability interval corresponding to each subset of data is obtained.

[0071] In one possible implementation, in the training set samples Based on ~e, each initial energy consumption prediction model is adopted. The initial CQR (Conformalized Quantile Regression) prediction model in the dataset predicts the training set samples in each subset. Energy consumption prediction is performed using the corresponding time-series feature information to obtain the training set samples in each subset of the dataset. The corresponding predicted energy consumption probability interval. Among them, the initial CQR prediction model is a probability prediction model based on statistical theory, which combines the advantages of conformal prediction and quantile regression (QR) models. It can construct an effective prediction interval for heteroscedastic data without assuming data distribution. In addition, it can adapt to local changes in data. Therefore, it is suitable for HVAC energy consumption probability prediction with complex data distribution. Among them, the initial CQR prediction model can be expressed by the following formula (8) using conformal prediction and quantile regression: Formula (8) Among them, i ; and These represent the upper and lower quantiles, respectively; f represents the energy consumption prediction sub-model LCQR; Represents input features, Represents the true value; This represents the lower quantile regression function; This represents the upper quantile regression function.

[0072] in, Where 'a' represents the confidence interval.

[0073] Based on the above formula (8), energy consumption prediction is performed on the time-series feature information corresponding to each subset of data to obtain the predicted energy consumption probability interval for each subset of data. It can be expressed by the following formula (9).

[0074] Formula (9) Among them, in the above formula (9) It can be expressed by the following formula (10), where the prediction bias in the above formula (9) is... It can be expressed by the following formula (11): Formula (10) Formula (11) S703. Based on the actual energy consumption probability of each sub-dataset corresponding to the sub-time period and the corresponding predicted energy consumption probability interval, train the initial LSTM network and the initial CQR prediction model in each initial energy consumption prediction model to obtain a trained energy consumption prediction sub-model.

[0075] The post-training energy consumption prediction sub-model includes a post-training LSTM network and a post-training CQR prediction network.

[0076] In one possible implementation, each initial energy consumption prediction model is configured based on the actual energy consumption probability for each sub-time period corresponding to each subset of data and the corresponding predicted energy consumption probability interval. The initial LSTM network and the initial CQR prediction model are trained to obtain a trained energy consumption prediction sub-model LCQR.

[0077] For example, see reference Figure 1 , training set samples The features in ~e are input into the initial Long Short-Term Memory (LSTM) network to obtain the temporal feature information corresponding to each subset. Then, the temporal feature information corresponding to each subset is input into conformal prediction and quantile regression to obtain the interval corresponding to the above formula (8). Then, energy consumption prediction is performed on the temporal feature information corresponding to each subset according to the above formula (8) to obtain the predicted energy consumption probability interval corresponding to each subset. This leads to the formation of a post-trained energy consumption prediction sub-model, LCQR.

[0078] The training method for the energy consumption probability prediction model provided in this application integrates an initial Long Short-Term Memory (LSTM) network and an initial Shape-Conserving Quantile Regression (CQR) prediction model into each initial energy consumption prediction model. The initial LSTM network excels at processing time-series data and can capture time-series information from the energy consumption-related features of the samples. The initial CQR prediction model can perform probabilistic predictions of energy consumption, providing a prediction interval. These initial energy consumption prediction models enable the model to handle both time-series data and perform probabilistic energy consumption predictions, thus improving the overall prediction capability. First, the initial LSTM network in each initial energy consumption prediction model is used to perform time-series processing on the energy consumption-related features of samples in each subset, obtaining the time-series feature information corresponding to each subset. Then, the initial CQR prediction model in each initial energy consumption prediction model is used to predict energy consumption based on the time-series feature information corresponding to each subset, obtaining the predicted energy consumption probability interval for each subset, thus making the prediction results more reliable. Finally, based on the actual energy consumption probability of each subset's corresponding sub-time period and the corresponding predicted energy consumption probability interval, the initial LSTM network and initial CQR prediction model in each initial energy consumption prediction model are trained to obtain a trained energy consumption prediction sub-model. This trained sub-model is continuously optimized and can be composed of a trained LSTM network and a trained CQR prediction network. Since each energy consumption prediction sub-model is trained based on different subsets, they can learn the energy consumption characteristics under different time periods and operating conditions, further improving the predictive ability of the energy consumption prediction sub-model for building energy consumption in different scenarios and adapting to the dynamic changes in building energy consumption.

[0079] Optionally, the method described above, which constructs a target energy consumption prediction model based on multiple trained energy consumption prediction sub-models, may include: The target energy consumption prediction model is obtained by ensemble learning training of multiple trained energy consumption prediction sub-models.

[0080] In one possible implementation, the predicted energy consumption probability interval is obtained by applying the above formula (9) to multiple trained energy consumption prediction sub-models LCQR. Then for each training set sample Corresponding predicted energy consumption probability interval Perform ensemble learning training, for example, for each training set sample Corresponding predicted energy consumption probability interval Calculate the mean to obtain multiple mean-based energy consumption probability intervals. Then, based on multiple means, predict the energy consumption probability interval. Generate the target energy consumption prediction model LECQR.

[0081] The energy consumption probability prediction model training method provided in this application involves ensemble learning training of multiple trained energy consumption prediction sub-models to obtain a target energy consumption prediction model. Therefore, this application allows different trained energy consumption prediction sub-models to be trained based on different feature subsets, data subsets, or using different algorithm parameters. Through ensemble learning, these sub-models can complement each other, thereby improving the overall accuracy of energy consumption prediction in different scenarios.

[0082] Figure 8 A flowchart illustrating a training method for an energy consumption probability prediction model provided in this application embodiment. Figure 6 .like Figure 8 As shown, the above method integrates multiple trained energy consumption prediction sub-models to obtain the target energy consumption prediction model, which may include: S801. Each trained energy consumption prediction sub-model is used to predict the energy consumption-related features of the samples in each sub-time period, so as to obtain the predicted energy consumption probability interval of each trained energy consumption prediction sub-model for each sub-time period.

[0083] In one possible implementation, based on the validation set samples Each post-trained energy consumption prediction sub-model (LCQR) is used to predict the energy consumption-related features of samples in each sub-time period, thus obtaining the validation set samples for each post-trained energy consumption prediction sub-model. Predicted energy consumption probability intervals for each sub-time period .

[0084] S802. Based on the predicted energy consumption probability range of each trained energy consumption prediction sub-model for each sub-time period and the actual energy consumption probability of each sub-time period, obtain the prediction deviation of each trained energy consumption prediction sub-model for each sub-time period.

[0085] In one possible implementation, the predicted energy consumption probability interval for each sub-time period is determined based on the LCQR of each trained energy consumption prediction sub-model. The actual energy consumption probability of each sub-time period can be obtained by using the following formulas (12) and (13) to determine the prediction bias of each trained energy consumption prediction sub-model LCQR for each sub-time period. .

[0086] Formula (12) Formula (13) in, and .

[0087] S803. Based on the prediction deviations of multiple trained energy consumption prediction sub-models for each sub-time period, construct a prediction residual set.

[0088] S804. Based on the predicted residual set, construct the target energy consumption prediction model.

[0089] In one possible implementation, the prediction bias of multiple trained energy consumption prediction sub-models for each sub-time period is taken into account. Construct the prediction residual set [ , Furthermore, since the prediction interval generated by the target energy consumption prediction model LECQR is constructed by LECQR using the estimated quantile function, absolute consistency score, and asymmetric consistency score, then according to the prediction residual set [ , The energy consumption prediction interval of the target energy consumption prediction model LECQR is obtained using the following formula (14). .

[0090]

[0091] Formula (14) in, It is an aggregate function; Represents the residual set.

[0092] The residual set The T predicted values ​​can be used to update the corresponding T original values ​​in the residual set, i.e. , .

[0093] The training method for the energy consumption probability prediction model provided in this application involves using each trained energy consumption prediction sub-model to predict the energy consumption-related features of samples in each sub-time period, obtaining the predicted energy consumption probability interval for each sub-time period. Based on the predicted energy consumption probability interval for each sub-time period and the actual energy consumption probability for each sub-time period, the prediction deviation of each sub-model is obtained. By comparing the predicted energy consumption probability interval and the actual energy consumption probability of each sub-model, the prediction deviation of each sub-model in different time periods can be accurately obtained. This comparison method based on probability intervals more comprehensively reflects the accuracy of model prediction than the comparison of a single value because it considers the uncertainty of energy consumption. Then, based on the prediction deviations of multiple trained energy consumption prediction sub-models for each sub-time period, a prediction residual set is constructed. This residual set contains the error information of each sub-model in different time periods, reflecting the advantages and disadvantages of different energy consumption prediction sub-models in specific time periods. Based on the prediction residual set, a target energy consumption prediction model is constructed to adjust and optimize the model using information from the residual set. Therefore, the target energy consumption prediction model provided in this application integrates the advantages of various energy consumption prediction sub-models, reducing prediction errors and improving overall prediction accuracy. In practical applications, more accurate energy consumption prediction helps building energy management systems to rationally allocate energy supply and equipment operation, thereby reducing energy costs.

[0094] To facilitate understanding of the training method for the energy consumption probability prediction model described above, this application also provides an example of the flowchart for the training method of the energy consumption probability prediction model, which will be further described below with reference to the accompanying drawings. Figure 9 A flowchart illustrating a training method for an energy consumption probability prediction model provided in this application embodiment. Figure 7 .like Figure 9 As shown in the illustration, the embodiments provided in this application provide... Figure 7 It may include: S901. Extract features from the historical building data of the target building within a preset historical time period to obtain sample energy consumption-related features for the preset historical time period.

[0095] Specifically, refer to Figure 1 Based on the Pearson correlation coefficient analysis method, and referring to the above formula (1), linear correlation features of samples were selected from various feature data of historical buildings in the preset historical time period for the target building. At the same time, based on the Spearman rank correlation coefficient analysis method, and referring to the above formula (2), nonlinear correlation features of samples were selected from various feature data of historical buildings in the preset historical time period for the target building. Then, the linear correlation features and nonlinear correlation features of the samples were sorted and fused according to the preset time period or preset time point to obtain the energy consumption correlation features of the samples.

[0096] S902. Based on the energy consumption characteristics of samples in a preset historical time period and the actual energy consumption probability of HVAC in the target building in the preset historical time period, construct a sample dataset and divide the sample dataset into multiple sub-datasets.

[0097] Specifically, a sample dataset is constructed by combining sample energy consumption-related features from a preset historical time period with the actual energy consumption probability. This sample dataset is then divided into multiple subsets based on the preset historical time period. These subsets may include: test set samples. Training set samples ~e and validation set samples .

[0098] S903. Based on multiple subset datasets, train multiple initial energy consumption prediction models to obtain multiple trained energy consumption prediction sub-models.

[0099] Specifically, based on training set samples from multiple subsets of the dataset ~e is input into the initial LSTM network to process the training set samples. The energy consumption-related features of the samples in ~e are processed in time series to obtain the time series feature information corresponding to each subset (refer to the above formula (4)). Then, the time series feature information corresponding to each subset is used in the conformal prediction and quantile regression models to obtain multiple initial CQR prediction models. Then multiple initial CQR prediction models The output results are stored in the validation set samples. Based on the validation set samples Multiple trained energy consumption prediction sub-models LCQR were obtained.

[0100] S904. Construct a target energy consumption prediction model based on multiple trained energy consumption prediction sub-models.

[0101] Specifically, the training set samples ~e is input into multiple trained energy consumption prediction sub-models LCQR. According to the above formula (9), the output results of multiple trained energy consumption prediction sub-models LCQR are obtained, that is, the training set samples. ~e corresponds to the predicted energy consumption probability interval Then for each training set sample Corresponding predicted energy consumption probability interval Perform mean processing (refer to formulas (12) and (13) above) to obtain the predicted residual set [ , Then, based on the predicted residual set [ , The target energy consumption prediction model LECQR and its output are obtained through the above formula (14), which is the energy consumption prediction interval. .

[0102] The training method for the energy consumption probability prediction model provided in this application involves extracting features from historical building data of the target building over a preset historical time period to obtain sample energy consumption-related features for that period. Based on these features and the actual energy consumption probability of the HVAC system within the target building during that period, a sample dataset is constructed and divided into multiple subsets. Multiple initial energy consumption prediction models are then trained on these subsets to obtain multiple trained energy consumption prediction sub-models. Finally, a target energy consumption prediction model is constructed based on these sub-models. Therefore, this application comprehensively considers the impact of different factors on HVAC energy consumption, employing a fusion of linear and nonlinear correlation analysis methods to identify features with significant impact on HVAC energy consumption. Furthermore, when making probabilistic predictions of HVAC energy consumption, this application introduces an ensemble learning training strategy and improves the residual set of the prediction interval to construct a LECQR (Local Energy Consumption Probabilistic Prediction Model) that adapts to different time series lengths, thereby obtaining a high-precision prediction interval and achieving the goal of accurate building energy consumption prediction.

[0103] Optionally, Figure 10 This is a flowchart illustrating an energy consumption probability prediction method provided in an embodiment of this application. Figure 10 As shown, the energy consumption probability prediction method may include: S1001. Extract features from the predicted building data of the building to be predicted in a preset future time period to obtain the predicted energy consumption related features for the preset future time period.

[0104] S1002. Using a pre-trained target energy consumption prediction model, energy consumption is predicted based on the predicted energy consumption related characteristics of a preset future time period, and the predicted energy consumption probability range of the building to be predicted in the preset future time period is obtained.

[0105] In one possible approach, feature extraction is first performed on the predicted building data for the building to be predicted over a preset future time period. This process filters out key energy-related factors from a large amount of building data, obtaining predicted energy consumption-related features for the preset future time period. Since these features are extracted for a preset future time period, they can reflect the possible future building operation and environmental conditions. Then, a pre-trained target energy consumption prediction model, LECQR, is used to predict energy consumption based on the extracted predicted energy consumption-related features. This yields the predicted energy consumption probability range for the building's HVAC system within the preset future time period. Because the target energy consumption prediction model LECQR has undergone multiple rounds of training and optimization, it integrates the advantages of multiple trained energy consumption prediction sub-models and can fully consider the complex relationships between various influencing factors. Therefore, it can give a relatively accurate range of possible energy consumption, providing a scientific basis for energy management.

[0106] For example, continue to refer to Figure 1 Test set samples The energy consumption-related features of the samples are input into the trained energy consumption prediction sub-model LCQR. Then, the output of the energy consumption prediction sub-model LCQR is input into the target energy consumption prediction model LECQR to obtain the output of the target energy consumption prediction model LECQR, i.e., the energy consumption prediction interval. .

[0107] Furthermore, based on this predicted energy consumption probability interval The Prediction Intervals Coverage Probability (PICP), Prediction Intervals Normalized Average Width (PINAW), and Coverage Width-based Criterion (CWC) are selected to evaluate the predicted energy consumption probability interval. Verification is required.

[0108] Among them, the interval prediction coverage rate measures the coverage of the prediction interval, ensuring that the true value has a sufficient probability of falling within the prediction interval, and can be expressed by the following formula (15): Formula (15) In formula (15) above, K can be expressed by formula (16) below: Formula (16) Among them, in the above formula (16) It is the predicted mean. It is the prediction interval.

[0109] In evaluating the coverage of an interval forecast, the coverage of the interval forecast should be as close as possible to the preset confidence level (e.g., 95%), indicating that the forecast interval has a high coverage.

[0110] The normalized average width of the prediction interval is used to calculate the relative average width, which measures the width of the prediction interval and ensures that the prediction interval is not too wide, thus losing its practical significance. It can be expressed by the following formula (17): Formula (17) In evaluating the normalized mean width of the prediction interval, PINAW should be as small as possible, indicating a narrower prediction interval and higher accuracy.

[0111] The coverage width standard, by integrating PICP and PINAW, balances coverage and width, providing a more comprehensive performance verification of the prediction interval. It can be expressed by the following formula (18): Formula (18) In evaluating coverage width criteria, CWC should be as small as possible, meaning that the prediction interval width should be as narrow as possible while meeting the preset coverage rate. It should be noted that the output of the target energy consumption prediction model LECQR in this application, i.e., the energy consumption prediction interval... In the evaluation, the CWC needs to be as large as possible to make the energy consumption prediction range... The wider.

[0112] For example, this application also provides a specific application embodiment, where the historical building data of the target building within a preset historical time period is the actual data of a public building in Macau from March 1, 2021 to September 30, 2021. The historical building data is used to construct a sample dataset, which is then divided into multiple subsets as follows: the training set samples contain data from March to August, using... ~e indicates that the validation set sample contains data from the first half of the month, from September 1st to 15th. This indicates that the test set sample contains data from the second half of the month, from September 16th to 30th. express.

[0113] Furthermore, the training set samples ~e is divided into three subsets, each denoted by subset count b, corresponding to the following time periods: March to April, May to June, and July to August. Using the above formulas (1) and (2), the main correlation features are lighting and socket energy consumption. Following closely are the correlations with ambient temperature, daily maximum temperature, daily minimum temperature, perceived temperature, and dew point temperature. However, the correlations with relative humidity, total daily rainfall, 10-minute average wind speed, gust wind speed, wind direction, and week number are relatively weak, and their impact on energy consumption is negligible. Therefore, ambient temperature, daily maximum temperature, daily minimum temperature, perceived temperature, dew point temperature, number of hours in a day, number of days in a week, number of days in a month, whether it is a holiday, lighting, socket energy consumption, HVAC energy consumption history, and pedestrian traffic are selected as the main features affecting energy consumption.

[0114] Furthermore, this application also provides a table illustrating the preset effects of each energy consumption prediction sub-model LCQR compared to the target energy consumption prediction model LECQR proposed in this invention. Table 1 shows the prediction results of each energy consumption prediction sub-model and the target energy consumption prediction model provided in the embodiments of this application. As shown in Table 1.

[0115]

[0116] As shown in Table 1, in the evaluation of the Interval Prediction Coverage Proportion (PICP), the target energy consumption prediction model LECQR has a higher interval prediction coverage than the other energy consumption prediction sub-models LCQR. In the evaluation of the Normalized Average Width of the Prediction Interval (PINAW), the prediction intervals of energy consumption prediction sub-models LCQR1 and LCQR2 are narrower, while the prediction interval of the target energy consumption prediction model LECQR is relatively wider. In the evaluation of the Coverage Width Criterion (CWC), the interval prediction coverage of the target energy consumption prediction model LECQR is better than that of the other energy consumption prediction sub-models LCQR.

[0117] The energy consumption probability prediction method provided in this application extracts features from the predicted building data of the building to be predicted over a preset future time period. This allows for the precise identification of key energy consumption-related features from complex data, resulting in predicted energy consumption-related features for the preset future time period. These features fully consider various possible scenarios for future building operation. A pre-trained target energy consumption prediction model is then used to predict energy consumption based on these predicted energy consumption-related features for the preset future time period. This yields the predicted energy consumption probability range for the building's HVAC system within the preset future time period, effectively quantifying the uncertainty in energy consumption prediction. Therefore, the target energy consumption prediction model LECQR of this application can effectively assess the building's energy consumption probability, reduce the impact of energy supply issues on normal building operation, and enhance the building's stability and reliability.

[0118] Based on the same inventive concept, this application also provides a training device for an energy consumption probability prediction model and an energy consumption probability prediction device. Since the principle of the device in this application is similar to the training method and energy consumption probability prediction method of the energy consumption probability prediction model described above in this application, the implementation of the device can refer to the implementation of the method, and the repeated parts will not be described again.

[0119] Figure 11 This is a schematic diagram of the structure of a training device for an energy consumption probability prediction model provided in an embodiment of this application. Figure 11 As shown, the training device 1100 for the energy consumption probability prediction model may include: The first extraction module 1101 is used to extract features from the historical building data of the target building in a preset historical time period to obtain sample energy consumption related features in the preset historical time period. The first construction module 1102 is used to construct a sample dataset based on the sample energy consumption related features of a preset historical time period and the actual energy consumption probability of the HVAC system in the target building during the preset historical time period. The partitioning module 1103 is used to divide the sample dataset into multiple sub-datasets, each sub-dataset corresponding to a sub-time period within a preset historical time period. Training module 1104 is used to train multiple initial energy consumption prediction models based on multiple subset datasets to obtain multiple trained energy consumption prediction sub-models. The second building module 1105 is used to build a target energy consumption prediction model based on multiple trained energy consumption prediction sub-models.

[0120] In one optional implementation, the sample energy consumption related features include: sample linear correlation features and sample nonlinear correlation features; the first extraction module 1101 is specifically used to: extract features from historical building data using a preset linear correlation analysis method based on the actual energy consumption probability to obtain sample linear correlation features; extract features from historical building data using a preset nonlinear correlation analysis method based on the actual energy consumption probability to obtain sample nonlinear correlation features; and fuse the sample linear correlation features and sample nonlinear correlation features to obtain sample energy consumption related features.

[0121] In one optional implementation, the first extraction module 1101 is specifically used to: calculate the Pearson correlation coefficient between the actual energy consumption probability and various feature data in the historical building data using a preset linear correlation analysis method; and select sample linear correlation features from various feature data in the historical building data based on the Pearson correlation coefficient.

[0122] In one optional implementation, the first extraction module 1101 is specifically used to: calculate the Spearman rank correlation coefficient between the actual energy consumption probability and various feature data in the historical building data using a preset nonlinear correlation analysis method; and select sample nonlinear correlation features from various feature data in the historical building data based on the Spearman rank correlation coefficient.

[0123] In one optional implementation, each initial energy consumption prediction model integrates: an initial long short-term memory (LSTM) network and an initial conformal quantile regression (CQR) prediction model; the training module 1104 is specifically used to: use the initial LSTM network in each initial energy consumption prediction model to perform time-series processing on the energy consumption-related features of samples in each subset, to obtain the time-series feature information corresponding to each subset; use the initial CQR prediction model in each initial energy consumption prediction model to predict energy consumption on the time-series feature information corresponding to each subset, to obtain the predicted energy consumption probability interval corresponding to each subset; and train the initial LSTM network and the initial CQR prediction model in each initial energy consumption prediction model according to the actual energy consumption probability of the corresponding sub-time period of each subset and the corresponding predicted energy consumption probability interval, to obtain a trained energy consumption prediction sub-model, which includes: a trained LSTM network and a trained CQR prediction network.

[0124] In one optional implementation, the second building module 1105 is specifically used to: perform ensemble learning training on multiple trained energy consumption prediction sub-models to obtain a target energy consumption prediction model.

[0125] In one optional implementation, the second construction module 1105 is specifically used for: using each post-trained energy consumption prediction sub-model to predict the sample energy consumption-related features of each sub-time period, obtaining the predicted energy consumption probability interval of each post-trained energy consumption prediction sub-model for each sub-time period; obtaining the prediction deviation of each post-trained energy consumption prediction sub-model for each sub-time period based on the predicted energy consumption probability interval of each post-trained energy consumption prediction sub-model for each sub-time period and the actual energy consumption probability of each sub-time period; constructing a prediction residual set based on the prediction residual set; and constructing a target energy consumption prediction model based on the prediction residual set.

[0126] Figure 12 This is a schematic diagram of an energy consumption probability prediction device provided in an embodiment of this application. Figure 12 As shown, the energy consumption probability prediction device 1200 may include: The second extraction module 1201 is used to extract features from the predicted building data of the building to be predicted in a preset future time period to obtain the predicted energy consumption related features for the preset future time period. The prediction module 1202 is used to use a pre-trained target energy consumption prediction model to predict energy consumption based on the predicted energy consumption related features of a preset future time period, and to obtain the predicted energy consumption probability range of the building to be predicted in the preset future time period.

[0127] It should be noted that for details not disclosed in the training device and energy consumption probability prediction device of the energy consumption probability prediction model in the embodiments of this application, please refer to the details disclosed in the training method and energy consumption probability prediction method of the energy consumption probability prediction model in the embodiments of this application, which will not be repeated here.

[0128] These modules can be one or more integrated circuits configured to implement the above methods, such as one or more Application Specific Integrated Circuits (ASICs), one or more microprocessors, or one or more Field Programmable Gate Arrays (FPGAs). Alternatively, when a module is implemented using processing element scheduler code, the processing element can be a general-purpose processor, such as a Central Processing Unit (CPU) or other processor capable of calling program code. Furthermore, these modules can be integrated together as a system-on-a-chip (SOC).

[0129] Optionally, embodiments of this application also provide a computer-readable storage medium storing a computer program. When the computer program is run by a processor, the processor executes the steps of the training method and the energy consumption probability prediction method for the removable storage medium described in the above embodiments. The specific implementation and technical effects are similar and will not be repeated here.

[0130] The above description is merely an embodiment of this application and is not intended to limit the scope of protection of this application. Various modifications and variations can be made to this application by those skilled in the art. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of this application should be included within the scope of protection of this application.

Claims

1. A training method for an energy consumption probability prediction model, characterized in that, include: Feature extraction is performed on the historical building data of the target building within a preset historical time period to obtain the sample energy consumption related features of the preset historical time period; A sample dataset is constructed based on the energy consumption characteristics of the samples during the preset historical time period and the actual energy consumption probability of the HVAC system in the target building during the preset historical time period. The sample dataset is divided into multiple sub-datasets, and the multiple sub-datasets correspond to multiple sub-time periods within the preset historical time period; Based on the multiple subset datasets, multiple initial energy consumption prediction models are trained to obtain multiple trained energy consumption prediction sub-models; each initial energy consumption prediction model integrates: an initial long short-term memory LSTM network and an initial conformal quantile regression CQR prediction model. The step involves training multiple initial energy consumption prediction models based on the multiple subset datasets to obtain multiple trained energy consumption prediction sub-models, including: The initial LSTM network in each initial energy consumption prediction model is used to perform time-series processing on the energy consumption-related features of samples in each subset to obtain the time-series feature information corresponding to each subset. Using the initial CQR prediction model in each initial energy consumption prediction model, energy consumption prediction is performed on the time-series feature information corresponding to each subset of data to obtain the predicted energy consumption probability interval corresponding to each subset of data. Based on the actual energy consumption probability of each sub-dataset corresponding to the sub-time period and the corresponding predicted energy consumption probability interval, the initial LSTM network and the initial CQR prediction model in each initial energy consumption prediction model are trained to obtain a trained energy consumption prediction sub-model. The trained energy consumption prediction sub-model includes: a trained LSTM network and a trained CQR prediction network. Based on the multiple trained energy consumption prediction sub-models, a target energy consumption prediction model is constructed.

2. The training method for the energy consumption probability prediction model according to claim 1, characterized in that, The sample energy consumption-related features include: sample linear correlation features and sample nonlinear correlation features; the step of extracting features from historical building data of the target building within a preset historical time period to obtain the sample energy consumption-related features for the preset historical time period includes: Based on the actual energy consumption probability, a preset linear correlation analysis method is used to extract features from the historical building data to obtain sample linear correlation features; Based on the actual energy consumption probability, a preset nonlinear correlation analysis method is used to extract features from the historical building data to obtain sample nonlinear correlation features; The sample energy consumption correlation features are obtained by fusing the linear correlation features and the nonlinear correlation features of the sample.

3. The training method for the energy consumption probability prediction model according to claim 2, characterized in that, The step involves extracting features from the historical building data based on the actual energy consumption probability using a preset linear correlation analysis method to obtain sample linear correlation features, including: Using the aforementioned preset linear correlation analysis method, the Pearson correlation coefficients between the actual energy consumption probability and various feature data in the historical building data are calculated respectively; Based on the Pearson correlation coefficient, linear correlation features of the samples are selected from various feature data in the historical building data.

4. The training method for the energy consumption probability prediction model according to claim 2, characterized in that, The step involves extracting features from the historical building data based on the actual energy consumption probability using a preset nonlinear correlation analysis method to obtain sample nonlinear correlation features, including: Using the aforementioned preset nonlinear correlation analysis method, the Spearman rank correlation coefficients between the actual energy consumption probability and various feature data in the historical building data are calculated respectively. Based on the Spearman rank correlation coefficient, the nonlinear correlation features of the samples are selected from various feature data in the historical building data.

5. The training method for the energy consumption probability prediction model according to claim 1, characterized in that, The step of constructing a target energy consumption prediction model based on the plurality of trained energy consumption prediction sub-models includes: The target energy consumption prediction model is obtained by performing ensemble learning training on the multiple trained energy consumption prediction sub-models.

6. The training method for the energy consumption probability prediction model according to claim 5, characterized in that, The step of ensemble learning and training the multiple trained energy consumption prediction sub-models to obtain the target energy consumption prediction model includes: Each trained energy consumption prediction sub-model is used to predict the energy consumption-related features of the samples in each sub-time period, thereby obtaining the predicted energy consumption probability interval for each trained energy consumption prediction sub-model for each sub-time period. Based on the predicted energy consumption probability range of each post-trained energy consumption prediction sub-model for each sub-time period and the actual energy consumption probability of each sub-time period, the prediction deviation of each post-trained energy consumption prediction sub-model for each sub-time period is obtained. Based on the prediction deviations of the multiple trained energy consumption prediction sub-models for each sub-time period, a prediction residual set is constructed; Based on the predicted residual set, the target energy consumption prediction model is constructed.

7. A method for predicting the probability of energy consumption, characterized in that, include: Feature extraction is performed on the predicted building data of the building to be predicted in a preset future time period to obtain the predicted energy consumption related features for the preset future time period. A pre-trained target energy consumption prediction model is used to predict energy consumption based on the predicted energy consumption related characteristics of the preset future time period, thereby obtaining the predicted energy consumption probability range of the building to be predicted for HVAC within the preset future time period. The target energy consumption prediction model is trained using the training method for the energy consumption probability prediction model as described in any one of claims 1 to 6.

8. A computer device, characterized in that, include: processor; A memory for storing processor-executable instructions; wherein the processor implements the training method of the energy consumption probability prediction model as described in any one of claims 1 to 6 and the energy consumption probability prediction method as described in claim 7 by running the executable instructions.

9. A computer-readable storage medium, characterized in that, The storage medium stores a computer program, which, when executed by a processor, performs the training method of the energy consumption probability prediction model as described in any one of claims 1 to 6 and the steps of the energy consumption probability prediction method as described in claim 7.

Images