Heating load prediction method and system based on big data

Abstract

The present invention discloses a heating load prediction method and system based on big data, which belongs to the field of building energy management technology. The method includes data preparation, data processing, relevant factor analysis, heating load prediction and building energy regulation. The present invention combines the results of building thermal insulation performance evaluation and weather condition prediction to further predict the heating load, improve the prediction accuracy and real-time performance, enhance the practicality and stability of the system, and help to adjust the heating system in time to cope with extreme weather changes; the use of a bidirectional time-gated transformer model for heating load prediction can effectively capture the long-term and short-term dependencies of the time series, integrate multiple feature information, improve the comprehensive prediction ability and accuracy of the model, enhance the system reliability, and help optimize the energy management strategy of the heating system.

Description

Technical Field

[0001] The present invention belongs to the technical field of building energy management, and specifically refers to a heating load prediction method and system based on big data. Background Art

[0002] Heating load forecasting uses big data technology to obtain a variety of relevant data that affect the heating load, and uses machine learning technology to predict heating demand within a certain period of time in the future, thereby assisting energy managers to effectively adjust the operating strategy of the heating system, improve energy utilization efficiency and economic benefits, and promote the intelligent development of energy management. Accurately predicting heating load helps ensure appropriate indoor temperature and improve user comfort.

[0003] However, in the existing heating load forecasting process, since the heating load is easily affected by two objective factors, building characteristics and weather conditions, prediction based solely on building characteristics is prone to miss the impact of external factors on heating demand, and only considering weather conditions cannot combine the differences between specific building insulation performance and structure. Therefore, the traditional method lacks consideration of the combination of building characteristics and weather conditions, which affects the comprehensiveness and accuracy of heating load prediction. There is a technical problem that traditional models are difficult to simultaneously apply to the prediction needs of long-term and short-term factors after the combination of weather conditions and building characteristics. Since it is necessary to consider both the long-term time series dependency of long-term prediction and the instantaneous demand prediction of instantaneous bursts, the prediction accuracy of traditional models needs to be improved. Summary of the invention

[0004] In view of the above situation, in order to overcome the defects of the prior art, the present invention provides a heating load prediction method and system based on big data. In the existing heating load prediction process, since the heating load is easily affected by two objective factors, building characteristics and weather conditions, prediction based solely on building characteristics is prone to miss the impact of external factors on heating demand, and only considering weather conditions cannot combine the differences between specific building insulation performance and structure. Therefore, the traditional method lacks consideration of the combination of building characteristics and weather conditions, which affects the comprehensiveness and accuracy of heating load prediction. The present solution creatively combines the results of building insulation performance evaluation and weather condition prediction to further predict the heating load, thereby improving the prediction accuracy and real-time It enhances the practicality and stability of the system, helps to adjust the heating system in time to cope with extreme weather changes; in the existing heating load forecasting process, the traditional model is difficult to simultaneously apply to the forecasting needs of long-term and short-term factors after the weather conditions and building characteristics are combined. Since it is necessary to consider both the long-term dependency of long-term forecasting and the instantaneous demand forecasting of instantaneous bursts, the forecasting accuracy of the traditional model needs to be improved. This solution creatively adopts a bidirectional time-gated transformer model for heating load forecasting, which can effectively capture the long-term and short-term dependencies of the time series, integrate a variety of feature information, improve the comprehensive forecasting ability and accuracy of the model, enhance the system reliability, and help optimize the energy management strategy of the heating system.

[0005] The technical solution adopted by the present invention is as follows: The present invention provides a heating load prediction method based on big data, which comprises the following steps:

[0006] Step S1: data preparation;

[0007] Step S2: data processing;

[0008] Step S3: Correlation factor analysis;

[0009] Step S4: heating load prediction;

[0010] Step S5: Building energy regulation.

[0011] Furthermore, in step S1, the data preparation is used to collect raw data required for heating load prediction, specifically, by collecting historical heating load data, weather data, building information, indoor environment data and calendar tags to obtain the raw data for heating load prediction.

[0012] Further, in step S2, the data processing is specifically to obtain high-quality data for heating load prediction by performing data preprocessing on the original data for heating load prediction, and to obtain important features for heating load prediction by performing feature selection on the high-quality data for heating load prediction, including the following steps:

[0013] Step S21: data preprocessing, specifically filling missing values ​​in the original data of heating load prediction by linear interpolation, and performing outlier filtering and data normalization to obtain high-quality data for heating load prediction;

[0014] Step S22: feature selection, specifically, by calculating the Pearson correlation coefficient, filtering out features that are highly correlated with the heating load from the high-quality data for heating load prediction, and obtaining important features for heating load prediction.

[0015] Further, in step S3, the related factor analysis is used to evaluate the thermal insulation performance of the building and predict the weather conditions, specifically using a support vector regression model to evaluate the thermal insulation performance of the building to obtain the thermal insulation performance information of the building, and using a long-term and short-term network model based on an attention mechanism to predict the weather conditions to obtain the weather condition prediction information, including the following steps:

[0016] Step S31: constructing a support vector regression model, including the following steps:

[0017] Step S311: construct a kernel function, specifically selecting a Gaussian kernel function as the kernel function of the support vector regression model, and the calculation formula is:

[0018] ;

[0019] Where, K(·) is the kernel function, i is the first index of the input sample, j is the second index of the input sample, the first index of the input sample is not equal to the second index of the input sample, and the input sample specifically refers to the high-quality data sample for heating load prediction, x i is the i-th input sample, x j is the jth input sample, exp(·) is the exponential function, is the kernel function parameter, which is used to control the width of the kernel function, ||x i -x j || is the Euclidean distance between the i-th input sample and the j-th input sample;

[0020] Step S312: construct a regression function, the calculation formula is:

[0021] ;

[0022] Where f(·) is the regression function, w f is the regression weight, b f is the regression bias term;

[0023] Step S313: constructing an objective function, specifically introducing a slack variable to optimize the objective function, constructing the objective function, and obtaining an optimized objective function. The calculation formula is:

[0024] ;

[0025] In the formula, is the optimization objective function, is the square norm of the regression weight, S is the penalty parameter used to control the penalty intensity of the model, n is the number of input samples, is the relaxed first variable corresponding to the i-th input sample, is the slack second variable corresponding to the i-th input sample, and the slack first variable and the slack second variable are used to measure the degree to which the prediction error exceeds the tolerance;

[0026] Step S314: constructing constraint conditions, specifically introducing slack variables to optimize constraint conditions, constructing constraint conditions, and obtaining optimized constraint conditions. The calculation formula is:

[0027] ;

[0028] In the formula, y i is the predicted value of the ith input sample, is the tolerance, which is used to indicate the acceptable range of prediction error;

[0029] Step S315: constructing a Lagrangian function to solve the optimal parameters of the model;

[0030] Step S316: constructing a support vector regression model by constructing the kernel function, the regression equation, the objective function, the constraint conditions and the Lagrangian function to obtain a support vector regression model;

[0031] Step S32: constructing a building thermal insulation performance evaluation model, specifically, obtaining a building thermal insulation performance evaluation model by training a support vector regression model;

[0032] Step S33: generating a building thermal insulation performance evaluation result, specifically using the high-quality data of heating load prediction, evaluating the building thermal insulation performance through a building thermal insulation performance evaluation model, and obtaining building thermal insulation performance information;

[0033] Step S34: weather condition prediction, specifically, performing weather condition prediction through a long-term and short-term network model based on an attention mechanism to obtain weather condition prediction information;

[0034] The long-short-term network model based on the attention mechanism specifically includes an input layer, a long-short-term network layer, a sparse attention module and an output layer.

[0035] Further, in step S4, the heating load prediction is specifically to combine the building thermal insulation performance information, the weather condition prediction information and the important characteristics of the heating load prediction, and use the bidirectional sequential gated transformer model to perform heating load prediction to obtain heating load prediction information;

[0036] The bidirectional temporal gated transformer model specifically includes a data input layer, a bidirectional long-term and short-term encoder, a gated residual network layer, a bidirectional long-term and short-term decoder, a masked multi-head self-attention module and an output prediction layer;

[0037] The data input layer is used to convert the format of input data, and specifically includes a position encoding layer and an embedding layer;

[0038] The bidirectional long-short term encoder is used to capture temporal dependencies;

[0039] The gated residual network layer is used to fuse time series features and static features;

[0040] The bidirectional long-term and short-term decoder is used to capture future timing information;

[0041] The masked multi-head self-attention module is used to capture the dependencies between different time steps;

[0042] The output prediction layer is used to generate prediction results, and specifically includes a position feedforward layer and a fully connected layer;

[0043] The heating load prediction comprises the following steps:

[0044] Step S41: Designing a bidirectional sequential gated transformer model, including the following steps:

[0045] Step S411: construct a data input layer, specifically, take the building thermal insulation performance information, weather condition forecast information and important features of heating load forecast as input data, set a position encoding layer and an embedding layer in the data input layer, and convert the input data into a high-dimensional embedding vector through the data input layer to obtain a building feature vector, a weather forecast feature vector and a heating load feature vector;

[0046] Step S412: construct a bidirectional long-term and short-term encoder, specifically, use a bidirectional long-term and short-term network as an encoder to obtain a bidirectional long-term and short-term encoder, encode the heating load feature vector through the bidirectional long-term and short-term encoder, and obtain a context feature vector. The calculation formula is:

[0047] ;

[0048] Where fea1 is the context feature vector, Encoder BiLSTM (·) is a bidirectional long-term and short-term encoder, fea hot is the heating load characteristic vector;

[0049] Step S413: construct a gated residual network layer, specifically, fuse the context feature vector and the building feature vector through the gated residual network layer to obtain a time-static fusion feature, and the calculation formula is:

[0050] ;

[0051] Where fea2 is the time-static fusion feature, GRN(·) is the gated residual network layer, and fea build is the building feature vector;

[0052] Step S414: construct a bidirectional long-term and short-term decoder, specifically, use the bidirectional long-term and short-term network as a decoder to obtain a bidirectional long-term and short-term decoder, decode the time-static fusion feature and the weather forecast feature vector through the bidirectional long-term and short-term decoder, and obtain a decoded feature vector. The calculation formula is:

[0053] ;

[0054] Where fea3 is the decoded feature vector, Decoder BiLSTM (·) is a bidirectional long-term and short-term decoder, fea weather is the weather forecast feature vector;

[0055] Step S415: constructing a masked multi-head self-attention module, specifically calculating the temporal static fusion feature and the decoding feature vector through the masked multi-head self-attention module to obtain the comprehensive feature of the heating load;

[0056] Step S416: constructing a prediction output layer, specifically setting a position feedforward layer and a fully connected layer in the prediction output layer, calculating the comprehensive characteristics of the heating load through the prediction output layer to obtain a prediction result;

[0057] Step S42: constructing a heating load prediction model, specifically, training a bidirectional sequential gated transformer model to obtain a heating load prediction model;

[0058] Step S43: generating a heating load prediction result, specifically, performing heating load prediction through a heating load prediction model to obtain heating load prediction information.

[0059] Furthermore, in step S5, the building energy regulation is used to regulate the heating system, specifically to dynamically adjust the operating parameters of the heating system according to the heating load prediction information to achieve building energy regulation.

[0060] The heating load prediction system based on big data provided by the present invention comprises: a data preparation module, a data preprocessing module, a related factor analysis module, a heating load prediction module and a building energy regulation module;

[0061] The data preparation module is used for data preparation, and obtains the original data of heating load prediction through data preparation, and sends the original data of heating load prediction to the data processing module;

[0062] The data processing module is used for data processing, and obtains high-quality data for heating load prediction and important features for heating load prediction through data processing, and sends the high-quality data for heating load prediction to the relevant factor analysis module, and sends the important features for heating load prediction to the heating load prediction module;

[0063] The relevant factor analysis module is used for relevant factor analysis, and obtains building thermal insulation performance information and weather condition forecast information through relevant factor analysis, and sends the building thermal insulation performance information and the weather condition forecast information to the heating load forecast module;

[0064] The heating load prediction module is used for heating load prediction, obtains heating load prediction information through heating load prediction, and sends the heating load prediction information to the energy regulation module;

[0065] The building energy regulation module is used for building energy regulation. According to the heating load prediction information, the operating parameters of the heating system are dynamically adjusted to achieve building energy regulation.

[0066] The beneficial effects achieved by the present invention using the above scheme are as follows:

[0067] (1) In the existing heating load forecasting process, since the heating load is easily affected by two objective factors, building characteristics and weather conditions, the prediction based solely on building characteristics is prone to miss the impact of external factors on heating demand, and only considering weather conditions cannot combine the differences between specific building insulation performance and structure. Therefore, the traditional method lacks the combined consideration of building characteristics and weather conditions, which affects the comprehensiveness and accuracy of heating load forecasting. This solution creatively combines the results of building insulation performance evaluation and weather condition forecasting to further forecast heating load, improves the forecast accuracy and real-time performance, enhances the practicality and stability of the system, and helps to adjust the heating system in time to cope with extreme weather changes;

[0068] (2) In the existing heating load forecasting process, the traditional model is difficult to simultaneously apply to the forecasting needs of long-term and short-term factors that are a combination of weather conditions and building characteristics. Since both the long-term time series dependency of long-term forecasting and the instantaneous demand forecast of instantaneous bursts must be considered, the prediction accuracy of the traditional model needs to be improved. This solution creatively adopts a bidirectional time-gated transformer model for heating load forecasting, which can effectively capture the long-term and short-term dependencies of the time series, integrate multiple feature information, improve the comprehensive forecasting ability and accuracy of the model, enhance system reliability, and help optimize the energy management strategy of the heating system. BRIEF DESCRIPTION OF THE DRAWINGS

[0069] Figure 1 A schematic diagram of the flow chart of the heating load prediction method based on big data provided by the present invention;

[0070] Figure 2 A schematic diagram of a heating load prediction system based on big data provided by the present invention;

[0071] Figure 3 is a schematic flow chart of step S3;

[0072] Figure 4 is a schematic flow chart of step S4;

[0073] Figure 5 It is a flowchart diagram of step S41.

[0074] The accompanying drawings are used to provide further understanding of the present invention and constitute a part of the specification. They are used to explain the present invention together with the embodiments of the present invention and do not constitute a limitation of the present invention. DETAILED DESCRIPTION

[0075] The technical solutions in the embodiments of the present invention will be clearly and completely described below in conjunction with the drawings in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, rather than all the embodiments; based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without making creative work are within the scope of protection of the present invention.

[0076] In the description of the present invention, it should be understood that terms such as “upper”, “lower”, “front”, “back”, “left”, “right”, “top”, “bottom”, “inside” and “outside” indicating directions or positional relationships are based on the directions or positional relationships shown in the accompanying drawings, and are only for the convenience of describing the present invention and simplifying the description, rather than indicating or implying that the device or element referred to must have a specific direction, be constructed and operated in a specific direction, and therefore should not be understood as limiting the present invention.

[0077] Example 1, see Figure 1The present invention provides a method for predicting heating load based on big data, which comprises the following steps:

[0078] Step S1: data preparation;

[0079] Step S2: data processing;

[0080] Step S3: Correlation factor analysis;

[0081] Step S4: heating load prediction;

[0082] Step S5: Building energy regulation.

[0083] Example 2, see Figure 1 , this embodiment is based on the above embodiment. In step S1, the data preparation is used to collect the original data required for heating load prediction, specifically, by collecting historical heating load data, weather data, building information, indoor environment data and calendar tags, to obtain the original data for heating load prediction;

[0084] The historical heating load data includes instantaneous power, instantaneous flow and accumulated heat;

[0085] The weather data includes dry bulb temperature, wind speed, wind direction, relative humidity and atmospheric pressure;

[0086] The building information includes building type, material, area and geographical location;

[0087] The indoor environment data includes indoor temperature, indoor humidity and heating system set temperature;

[0088] The calendar tag includes working day and holiday information.

[0089] Example 3, see Figure 1 This embodiment is based on the above embodiment. In step S2, the data processing is specifically to obtain high-quality data for heating load prediction by performing data preprocessing on the original data for heating load prediction, and to obtain important features for heating load prediction by performing feature selection on the high-quality data for heating load prediction, including the following steps:

[0090] Step S21: data preprocessing, specifically filling missing values ​​in the original data of heating load prediction by linear interpolation, and performing outlier filtering and data normalization to obtain high-quality data for heating load prediction;

[0091] Step S22: feature selection, specifically, by calculating the Pearson correlation coefficient, filtering out features that are highly correlated with the heating load from the high-quality data for heating load prediction, and obtaining important features for heating load prediction.

[0092] Example 4, see Figure 1 and Figure 3 This embodiment is based on the above embodiment. In step S3, the related factor analysis is used to evaluate the thermal insulation performance of the building and predict the weather conditions. Specifically, the support vector regression model is used to evaluate the thermal insulation performance of the building to obtain the thermal insulation performance information of the building. The weather conditions are predicted by the long-term and short-term network model based on the attention mechanism to obtain the weather condition prediction information, including the following steps:

[0093] Step S31: constructing a support vector regression model, including the following steps:

[0094] Step S311: construct a kernel function, specifically selecting a Gaussian kernel function as the kernel function of the support vector regression model, and the calculation formula is:

[0095] ;

[0096] Where, K(·) is the kernel function, i is the first index of the input sample, j is the second index of the input sample, the first index of the input sample is not equal to the second index of the input sample, and the input sample specifically refers to the high-quality data sample for heating load prediction, x i is the i-th input sample, x j is the jth input sample, exp(·) is the exponential function, is the kernel function parameter, which is used to control the width of the kernel function, ||x i -x j || is the Euclidean distance between the i-th input sample and the j-th input sample;

[0097] Step S312: construct a regression function, the calculation formula is:

[0098] ;

[0099] Where f(·) is the regression function, w f is the regression weight, b f is the regression bias term;

[0100] Step S313: constructing an objective function, specifically introducing a slack variable to optimize the objective function, constructing the objective function, and obtaining an optimized objective function. The calculation formula is:

[0101] ;

[0102] In the formula, is the optimization objective function, is the square norm of the regression weight, S is the penalty parameter used to control the penalty intensity of the model, n is the number of input samples, is the relaxed first variable corresponding to the i-th input sample, is the slack second variable corresponding to the i-th input sample, and the slack first variable and the slack second variable are used to measure the degree to which the prediction error exceeds the tolerance;

[0103] Step S314: constructing constraint conditions, specifically introducing slack variables to optimize constraint conditions, constructing constraint conditions, and obtaining optimized constraint conditions. The calculation formula is:

[0104] ;

[0105] In the formula, y i is the predicted value of the i-th input sample, is the tolerance, which is used to indicate the acceptable range of prediction error;

[0106] Step S315: constructing a Lagrangian function to solve the optimal parameters of the model;

[0107] Step S316: constructing a support vector regression model by constructing the kernel function, the regression equation, the objective function, the constraint conditions and the Lagrangian function to obtain a support vector regression model;

[0108] Step S32: constructing a building thermal insulation performance evaluation model, specifically, obtaining a building thermal insulation performance evaluation model by training a support vector regression model;

[0109] Step S33: generating an evaluation result, specifically using the high-quality data of heating load prediction to evaluate the building thermal insulation performance through a building thermal insulation performance evaluation model to obtain building thermal insulation performance information;

[0110] Step S34: weather condition prediction, specifically, performing weather condition prediction through a long-term and short-term network model based on an attention mechanism to obtain weather condition prediction information;

[0111] The long-short-term network model based on the attention mechanism specifically includes an input layer, a long-short-term network layer, a sparse attention module and an output layer;

[0112] The weather condition prediction comprises the following steps:

[0113] Step S341: construct an input layer, and input the weather data in the heating load forecast high-quality data into the model through the input layer to obtain model input data;

[0114] Step S342: construct a long-term and short-term network layer, calculate the model input data through the long-term and short-term network layer, and obtain weather time series characteristics;

[0115] Step S343: construct a sparse attention module, and calculate the weather time series features through the sparse attention module to obtain an attention context vector;

[0116] Step S344: construct an output layer, specifically using a fully connected layer as the output layer, and calculating the attention context vector through the output layer to obtain the model prediction result. The calculation formula is:

[0117] ;

[0118] Where t is the time step index, y t is the output result of the output layer at the tth time step, W y is the output layer weight, a t is the attention context vector at the tth time step, b y is the output layer bias term;

[0119] Step S345: constructing a long-short-term network model based on an attention mechanism, specifically by constructing an input layer, constructing a long-short-term network layer, constructing a sparse attention module, and constructing an output layer, to construct a long-short-term network model based on an attention mechanism;

[0120] Step S346: constructing a weather condition prediction model, and obtaining a weather condition prediction model by training a long-term and short-term network model based on an attention mechanism;

[0121] Step S347: Generate weather condition prediction results, perform weather condition prediction through a weather condition prediction model, and obtain weather condition prediction information.

[0122] Example 5, see Figure 1 , Figure 4 and Figure 5 , this embodiment is based on the above embodiment. In step S4, the heating load prediction is specifically to combine the building thermal insulation performance information, weather condition prediction information and important characteristics of heating load prediction, and use a bidirectional sequential gated transformer model to perform heating load prediction to obtain heating load prediction information;

[0123] The bidirectional temporal gated transformer model specifically includes a data input layer, a bidirectional long-term and short-term encoder, a gated residual network layer, a bidirectional long-term and short-term decoder, a masked multi-head self-attention module and an output prediction layer;

[0124] The data input layer is used to convert the format of input data, and specifically includes a position encoding layer and an embedding layer;

[0125] The bidirectional long-short term encoder is used to capture temporal dependencies;

[0126] The gated residual network layer is used to fuse time series features and static features;

[0127] The bidirectional long-term and short-term decoder is used to capture future timing information;

[0128] The masked multi-head self-attention module is used to capture the dependencies between different time steps;

[0129] The output prediction layer is used to generate prediction results, and specifically includes a position feedforward layer and a fully connected layer;

[0130] The heating load prediction comprises the following steps:

[0131] Step S41: Designing a bidirectional sequential gated transformer model, including the following steps:

[0132] Step S411: construct a data input layer, specifically, take the building thermal insulation performance information, weather condition forecast information and important features of heating load forecast as input data, set a position encoding layer and an embedding layer in the data input layer, and convert the input data into a high-dimensional embedding vector through the data input layer to obtain a building feature vector, a weather forecast feature vector and a heating load feature vector;

[0133] Step S412: construct a bidirectional long-term and short-term encoder, specifically, use a bidirectional long-term and short-term network as an encoder to obtain a bidirectional long-term and short-term encoder, encode the heating load feature vector through the bidirectional long-term and short-term encoder, and obtain a context feature vector. The calculation formula is:

[0134] ;

[0135] Where fea1 is the context feature vector, Encoder BiLSTM (·) is a bidirectional long-term and short-term encoder, fea hot is the heating load characteristic vector;

[0136] Step S413: construct a gated residual network layer, specifically, fuse the context feature vector and the building feature vector through the gated residual network layer to obtain a time-static fusion feature, and the calculation formula is:

[0137] ;

[0138] Where fea2 is the time-static fusion feature, GRN(·) is the gated residual network layer, and fea build is the building feature vector;

[0139] Step S414: construct a bidirectional long-term and short-term decoder, specifically, use the bidirectional long-term and short-term network as a decoder to obtain a bidirectional long-term and short-term decoder, decode the time-static fusion feature and the weather forecast feature vector through the bidirectional long-term and short-term decoder, and obtain a decoded feature vector. The calculation formula is:

[0140] ;

[0141] Where fea3 is the decoded feature vector, Decoder BiLSTM (·) is a bidirectional long-term and short-term decoder, fea weather is the weather forecast feature vector;

[0142] Step S415: constructing a masked multi-head self-attention module, specifically calculating the temporal static fusion feature and the decoding feature vector through the masked multi-head self-attention module to obtain the comprehensive feature of the heating load;

[0143] Step S416: constructing a prediction output layer, specifically setting a position feedforward layer and a fully connected layer in the prediction output layer, calculating the comprehensive characteristics of the heating load through the prediction output layer to obtain a prediction result;

[0144] Step S42: constructing a heating load prediction model, specifically, training a bidirectional sequential gated transformer model to obtain a heating load prediction model;

[0145] Step S43: generating a prediction result, specifically, performing a heating load prediction through a heating load prediction model to obtain heating load prediction information;

[0146] By performing the above operations, in view of the fact that in the existing heating load prediction process, since the heating load is easily affected by two objective factors, building characteristics and weather conditions, the prediction based solely on building characteristics is prone to miss the impact of external factors on heating demand, and only considering weather conditions cannot combine the differences between specific building insulation performance and structure. Therefore, the traditional method lacks the combined consideration of building characteristics and weather conditions, which affects the comprehensiveness and accuracy of heating load prediction. This solution creatively combines the results of building insulation performance evaluation and weather condition prediction to further predict the heating load, improves the prediction accuracy and real-time performance, enhances the practicality and stability of the system, and helps to adjust the heating system in time to cope with extreme weather changes;

[0147] In the existing heating load forecasting process, the traditional model is difficult to be applied to the prediction needs of long-term and short-term factors after the weather conditions and building characteristics are combined. Since it is necessary to consider both the long-term time series dependency of long-term prediction and the instantaneous demand prediction of instantaneous bursts, the prediction accuracy of the traditional model needs to be improved. This scheme creatively adopts a bidirectional time-gated transformer model for heating load forecasting, which can effectively capture the long-term and short-term dependencies of the time series, integrate multiple feature information, improve the comprehensive prediction ability and accuracy of the model, enhance system reliability, and help optimize the energy management strategy of the heating system.

[0148] Example 6, see Figure 1This embodiment is based on the above embodiment. In step S5, the building energy regulation is used to adjust the heating system, specifically to dynamically adjust the operating parameters of the heating system according to the heating load prediction information to achieve building energy regulation.

[0149] Embodiment 7, see Figure 2 , this embodiment is based on the above embodiment, and the heating load prediction system based on big data provided by the present invention includes: a data preparation module, a data preprocessing module, a related factor analysis module, a heating load prediction module and a building energy regulation module;

[0150] The data preparation module is used for data preparation, and obtains the original data of heating load prediction through data preparation, and sends the original data of heating load prediction to the data processing module;

[0151] The data processing module is used for data processing, and obtains high-quality data for heating load prediction and important features for heating load prediction through data processing, and sends the high-quality data for heating load prediction to the relevant factor analysis module, and sends the important features for heating load prediction to the heating load prediction module;

[0152] The relevant factor analysis module is used for relevant factor analysis, and obtains building thermal insulation performance information and weather condition forecast information through relevant factor analysis, and sends the building thermal insulation performance information and the weather condition forecast information to the heating load forecast module;

[0153] The heating load prediction module is used for heating load prediction, obtains heating load prediction information through heating load prediction, and sends the heating load prediction information to the energy regulation module;

[0154] The building energy regulation module is used for building energy regulation. According to the heating load prediction information, the operating parameters of the heating system are dynamically adjusted to achieve building energy regulation.

[0155] It should be noted that, in this article, relational terms such as first and second, etc. are only used to distinguish one entity or operation from another entity or operation, and do not necessarily require or imply any such actual relationship or order between these entities or operations. Moreover, the terms "include", "comprise" or any other variants thereof are intended to cover non-exclusive inclusion, so that a process, method, article or device including a series of elements includes not only those elements, but also other elements not explicitly listed, or also includes elements inherent to such process, method, article or device.

[0156] While the embodiments of the present invention have been shown and described, it will be appreciated by those skilled in the art that many changes, modifications, substitutions and variations can be made to the embodiments without departing from the principles and spirit of the invention.

[0157] The present invention and its embodiments are described above, and such description is not restrictive. The drawings show only one embodiment of the present invention, and the actual structure is not limited thereto. In short, if ordinary technicians in the field are inspired by it, without departing from the purpose of the invention, they can design a structure and embodiment similar to the technical solution without creativity, which should belong to the protection scope of the present invention.

Claims

1. A heating load prediction method based on big data, characterized by: The method comprises the following steps: Step S1: data preparation, used to collect the original data required for heating load forecasting; Step S2: data processing, specifically, performing data preprocessing on the original data of the heating load prediction to obtain high-quality data for the heating load prediction, and performing feature selection on the high-quality data for the heating load prediction to obtain important features for the heating load prediction; Step S3: Correlation factor analysis is used to evaluate the thermal insulation performance of the building and predict the weather conditions. Specifically, the support vector regression model is used to evaluate the thermal insulation performance of the building to obtain the thermal insulation performance information of the building, and the long-term and short-term network model based on the attention mechanism is used to predict the weather conditions to obtain the weather condition prediction information, including the following steps: Step S31: constructing a support vector regression model, including the following steps: Step S311: construct a kernel function, specifically selecting a Gaussian kernel function as the kernel function of the support vector regression model, and the calculation formula is: ； Where, K(·) is the kernel function, i is the first index of the input sample, j is the second index of the input sample, the first index of the input sample is not equal to the second index of the input sample, and the input sample specifically refers to the high-quality data sample for heating load prediction, x i is the i-th input sample, x j is the jth input sample, exp(·) is the exponential function, is the kernel function parameter, which is used to control the width of the kernel function, ||x i -x j || is the Euclidean distance between the i-th input sample and the j-th input sample; Step S312: construct a regression function, the calculation formula is: ； Where f(·) is the regression function, w f is the regression weight, b f is the regression bias term; Step S313: constructing an objective function, specifically introducing a slack variable to optimize the objective function, constructing the objective function, and obtaining an optimized objective function. The calculation formula is: ； In the formula, is the optimization objective function, is the square norm of the regression weight, S is the penalty parameter used to control the penalty intensity of the model, n is the number of input samples, is the relaxed first variable corresponding to the i-th input sample, is the slack second variable corresponding to the i-th input sample, and the slack first variable and the slack second variable are used to measure the degree to which the prediction error exceeds the tolerance; Step S314: constructing constraint conditions, specifically introducing slack variables to optimize constraint conditions, constructing constraint conditions, and obtaining optimized constraint conditions. The calculation formula is: ； In the formula, y i is the predicted value of the ith input sample, is the tolerance, which is used to indicate the acceptable range of prediction error; Step S315: constructing a Lagrangian function to solve the optimal parameters of the model; Step S316: constructing a support vector regression model by constructing the kernel function, the regression equation, the objective function, the constraint conditions and the Lagrangian function to obtain a support vector regression model; Step S32: constructing a building thermal insulation performance evaluation model, specifically, obtaining a building thermal insulation performance evaluation model by training a support vector regression model; Step S33: generating a building thermal insulation performance evaluation result, specifically using the high-quality data of heating load prediction, evaluating the building thermal insulation performance through a building thermal insulation performance evaluation model, and obtaining building thermal insulation performance information; Step S34: weather condition prediction, specifically, performing weather condition prediction through a long-term and short-term network model based on an attention mechanism to obtain weather condition prediction information; The long-short-term network model based on the attention mechanism specifically includes an input layer, a long-short-term network layer, a sparse attention module and an output layer; Step S4: heating load prediction, specifically combining the building thermal insulation performance information, weather condition prediction information and important characteristics of heating load prediction, using a bidirectional sequential gated transformer model to perform heating load prediction, and obtaining heating load prediction information; The bidirectional temporal gated transformer model specifically includes a data input layer, a bidirectional long-term and short-term encoder, a gated residual network layer, a bidirectional long-term and short-term decoder, a masked multi-head self-attention module and an output prediction layer; The data input layer is used to convert the format of input data, and specifically includes a position encoding layer and an embedding layer; The bidirectional long-short term encoder is used to capture temporal dependencies; The gated residual network layer is used to fuse time series features and static features; The bidirectional long-term and short-term decoder is used to capture future timing information; The masked multi-head self-attention module is used to capture the dependencies between different time steps; The output prediction layer is used to generate prediction results, and specifically includes a position feedforward layer and a fully connected layer; The heating load prediction comprises the following steps: Step S41: Designing a bidirectional sequential gated transformer model, including the following steps: Step S411: construct a data input layer, specifically, take the building thermal insulation performance information, weather condition forecast information and important features of heating load forecast as input data, set a position encoding layer and an embedding layer in the data input layer, and convert the input data into a high-dimensional embedding vector through the data input layer to obtain a building feature vector, a weather forecast feature vector and a heating load feature vector; Step S412: construct a bidirectional long-term and short-term encoder, specifically, use a bidirectional long-term and short-term network as an encoder to obtain a bidirectional long-term and short-term encoder, encode the heating load feature vector through the bidirectional long-term and short-term encoder, and obtain a context feature vector. The calculation formula is: ； Where fea1 is the context feature vector, Encoder BiLSTM (·) is a bidirectional long-term and short-term encoder, fea hot is the heating load characteristic vector; Step S413: construct a gated residual network layer, specifically, fuse the context feature vector and the building feature vector through the gated residual network layer to obtain a time-static fusion feature, and the calculation formula is: ； Where fea2 is the time-static fusion feature, GRN(·) is the gated residual network layer, and fea build is the building feature vector; Step S414: construct a bidirectional long-term and short-term decoder, specifically, use the bidirectional long-term and short-term network as a decoder to obtain a bidirectional long-term and short-term decoder, decode the time-static fusion feature and the weather forecast feature vector through the bidirectional long-term and short-term decoder, and obtain a decoded feature vector. The calculation formula is: ； Where fea3 is the decoded feature vector, Decoder BiLSTM (·) is a bidirectional long-term and short-term decoder, fea weather is the weather forecast feature vector; Step S415: constructing a masked multi-head self-attention module, specifically calculating the temporal static fusion feature and the decoding feature vector through the masked multi-head self-attention module to obtain the comprehensive feature of the heating load; Step S416: constructing a prediction output layer, specifically setting a position feedforward layer and a fully connected layer in the prediction output layer, calculating the comprehensive characteristics of the heating load through the prediction output layer to obtain a prediction result; Step S42: constructing a heating load prediction model, specifically, training a bidirectional sequential gated transformer model to obtain a heating load prediction model; Step S43: generating a heating load prediction result, specifically, performing heating load prediction through a heating load prediction model to obtain heating load prediction information; Step S5: Building energy regulation.

2. The heating load prediction method based on big data according to claim 1 is characterized in that: In step S2, the data processing is specifically to obtain high-quality data for heating load prediction by performing data preprocessing on the original data for heating load prediction, and to obtain important features for heating load prediction by performing feature selection on the high-quality data for heating load prediction, including the following steps: Step S21: data preprocessing, specifically filling missing values ​​in the original data of heating load prediction by linear interpolation, and performing outlier filtering and data normalization to obtain high-quality data for heating load prediction; Step S22: feature selection, specifically, by calculating the Pearson correlation coefficient, filtering out features that are highly correlated with the heating load from the high-quality data for heating load prediction, and obtaining important features for heating load prediction.

3. The heating load prediction method based on big data according to claim 2 is characterized in that: In step S5, the building energy regulation is used to regulate the heating system, specifically, dynamically adjust the operating parameters of the heating system according to the heating load prediction information to achieve building energy regulation.

4. The heating load prediction method based on big data according to claim 3 is characterized in that: In step S1, the data preparation is used to collect the original data required for heating load prediction, specifically, by collecting historical heating load data, weather data, building information, indoor environment data and calendar tags to obtain the original data for heating load prediction.

5. A heating load prediction system based on big data, used to implement the heating load prediction method based on big data as described in any one of claims 1 to 4, characterized in that: It includes data preparation module, data preprocessing module, relevant factor analysis module, heating load prediction module and building energy regulation module.

6. The heating load prediction system based on big data according to claim 5 is characterized in that: The data preparation module is used for data preparation, and obtains the original data of heating load prediction through data preparation, and sends the original data of heating load prediction to the data processing module; The data processing module is used for data processing, and obtains high-quality data for heating load prediction and important features for heating load prediction through data processing, and sends the high-quality data for heating load prediction to the relevant factor analysis module, and sends the important features for heating load prediction to the heating load prediction module; The relevant factor analysis module is used for relevant factor analysis, and obtains building thermal insulation performance information and weather condition forecast information through relevant factor analysis, and sends the building thermal insulation performance information and the weather condition forecast information to the heating load forecast module; The heating load prediction module is used for heating load prediction, obtains heating load prediction information through heating load prediction, and sends the heating load prediction information to the energy regulation module; The building energy regulation module is used for building energy regulation. According to the heating load prediction information, the operating parameters of the heating system are dynamically adjusted to achieve building energy regulation.

Images