Heat supply system heat load prediction device, method and system and storage medium

Abstract

The invention discloses a heat supply system heat load prediction device, method and system and a storage medium, and the device comprises a data processing module which is used for preprocessing collected heat supply system data; the feature generation module is used for processing the preprocessed heat supply system data and generating new data features through variational mode decomposition; the multi-feature fusion data set construction module is used for fusing the original collected data features and the generated new data features and constructing a thermal load prediction data set; the AI model construction module is used for constructing an AI model adopting a stacked bidirectional gating circulation unit and an optimized multi-head attention mechanism module; and the heat load prediction module is used for performing heat supply system heat load prediction based on the constructed AI model and the heat load prediction data set to obtain a heat load prediction result. According to the method, short-term prediction of the thermal load of the heat supply system is realized, and the precision and accuracy of thermal load prediction are effectively improved.

Description

Technical Field

[0001] The present invention belongs to the technical field of heating systems, and in particular relates to a device and method, a system and a storage medium for predicting heat load of a heating system. Background Art

[0002] The stable operation of the heating system depends on the high accuracy and efficiency of short-term load forecasting. Accurate load forecasting is a key factor in ensuring that the heating system can operate efficiently and energy-saving. Therefore, in-depth understanding and analysis of the inherent characteristics of the heating system operation data, as well as improving the prediction accuracy, has become a challenging and extremely important topic in the field. In recent years, with the development of technology, researchers at home and abroad have explored and proposed a series of prediction methods, which can be roughly summarized into the following three categories: traditional prediction methods, machine learning methods, and deep learning methods. Traditional prediction methods include but are not limited to regression analysis, autoregressive integrated moving average model (ARIMA) and Kalman filtering. This type of method is based on statistical principles and performs well in processing data sets with simple structures and clear patterns, while also having high computational efficiency. However, when faced with nonlinear or highly complex data samples, the prediction accuracy of traditional methods is not high. Summary of the invention

[0003] The technical problem to be solved by the present invention is to provide a device and method, system and storage medium for predicting heat load of a heating system, so as to solve the problem of insufficient accuracy in heat load prediction of existing actual heating systems using traditional methods.

[0004] To achieve the above object, the present invention adopts the following technical solution:

[0005] A heat load prediction device for a heating system, comprising:

[0006] A data processing module, used for preprocessing the collected heating system data;

[0007] A feature generation module is used to process the preprocessed heating system data and generate new data features through variational mode decomposition;

[0008] The multi-feature fusion data set construction module is used to fuse the original collected data features and the generated new data features to construct a heat load prediction data set;

[0009] AI model building module, used to build AI models using stacked bidirectional gated recurrent units and optimized multi-head attention mechanism modules;

[0010] The heat load prediction module is used to predict the heat load of the heating system based on the constructed AI model and the heat load prediction data set to obtain the heat load prediction results.

[0011] Preferably, the collected heating system data include: primary transient power, primary transient flow, primary water supply temperature, primary return water temperature, primary water supply pressure, primary return water pressure, secondary water supply temperature, secondary return water temperature, secondary water supply pressure, secondary return water pressure, indoor temperature, outdoor temperature, outdoor humidity, outdoor wind speed and solar radiation value per minute in each heating system.

[0012] Preferably, the feature generation module comprises:

[0013] The minute-to-hour data conversion unit is used to convert the primary transient power, primary transient flow, primary water supply temperature, primary return water temperature, primary water supply pressure, primary return water pressure, secondary water supply temperature, secondary return water temperature, secondary water supply pressure, secondary return water pressure, indoor temperature, outdoor temperature, outdoor humidity, outdoor wind speed and solar radiation value per minute into the corresponding primary transient power TP per hour by adding and averaging them. o , primary transient flow rate TFR, primary water supply temperature ST s , primary return water temperature RT s , primary water supply pressure PSP p , primary return water pressure PRP p , Secondary water supply temperature ST r , Secondary return water temperature RT r , Secondary water supply pressure PSP s , Secondary return water pressure PRP s , Indoor temperature T i , outdoor temperature T o , Outdoor humidity H o , outdoor wind speed W o And solar radiation data SR o ;

[0014] Generate a new feature unit to convert the transient power TP once per hour o The data features are processed using variational mode decomposition to obtain a series of intrinsic mode functions with specific center frequencies and bandwidths. The number of IMFs k is obtained by an adaptive calculation method based on the Bayesian information criterion;

[0015] The heating system feature generation module is used to convert all minute data into hourly data and obtain k intrinsic mode functions with specific center frequency and bandwidth based on variational mode decomposition as new data features to describe the heating system.

[0016] Preferably, the stacked bidirectional gated recurrent unit is formed by stacking two layers of bidirectional gated recurrent units. For each training sample, the data dimension is 48*(15+k)+4. The training sample is input into the SBiGRU for calculation and the calculation result is output.

[0017] The optimized multi-head attention mechanism module will use the output results of the SBiGRU layer to calculate the query, key, and value of the multi-head attention mechanism layer, perform linear transformation on the calculated query, key, and value, and split them into 8 multi-heads, and perform scaled dot product attention calculation at the same time.

[0018] The present invention also provides a method for predicting heat load of a heating system, comprising:

[0019] Step S1, preprocessing the collected heating system data;

[0020] Step S2, processing the preprocessed heating system data, and generating new data features through variational mode decomposition;

[0021] Step S3, integrating the original collected data features and the generated new data features to construct a heat load prediction data set;

[0022] Step S4, constructing an AI model using stacked bidirectional gated recurrent units and optimized multi-head attention mechanism modules;

[0023] Step S5: Based on the constructed AI model and the heat load prediction data set, the heat load of the heating system is predicted to obtain the heat load prediction result.

[0024] Preferably, the collected heating system data include: primary transient power, primary transient flow, primary water supply temperature, primary return water temperature, primary water supply pressure, primary return water pressure, secondary water supply temperature, secondary return water temperature, secondary water supply pressure, secondary return water pressure, indoor temperature, outdoor temperature, outdoor humidity, outdoor wind speed and solar radiation value per minute in each heating system.

[0025] Preferably, step S2 comprises:

[0026] Step S21, convert the primary transient power, primary transient flow, primary water supply temperature, primary return water temperature, primary water supply pressure, primary return water pressure, secondary water supply temperature, secondary return water temperature, secondary water supply pressure, secondary return water pressure, indoor temperature, outdoor temperature, outdoor humidity, outdoor wind speed and solar radiation value per minute into the corresponding primary transient power TP per hour by adding and averaging them. o , primary transient flow rate TFR, primary water supply temperature ST s , primary return water temperature RT s , primary water supply pressure PSP p , primary return water pressure PRP p , Secondary water supply temperature ST r , Secondary return water temperature RT r , Secondary water supply pressure PSPs , Secondary return water pressure PRP s , Indoor temperature T i , outdoor temperature T o , Outdoor humidity H o , outdoor wind speed W o And solar radiation data SR o ;

[0027] Step S22: convert the hourly transient power TP o The data features are processed using variational mode decomposition to obtain a series of intrinsic mode functions with specific center frequencies and bandwidths. The number of IMFs k is obtained by an adaptive calculation method based on the Bayesian information criterion;

[0028] Step S23: convert all minute data into hourly data, and use k natural mode functions with specific center frequencies and bandwidths obtained based on variational mode decomposition as new data features to describe the heating system.

[0029] Preferably, the stacked bidirectional gated recurrent unit is formed by stacking two layers of bidirectional gated recurrent units. For each training sample, the data dimension is 48*(15+k)+4. The training sample is input into the SBiGRU for calculation and the calculation result is output.

[0030] The optimized multi-head attention mechanism module will use the output results of the SBiGRU layer to calculate the query, key, and value of the multi-head attention mechanism layer, perform linear transformation on the calculated query, key, and value, and split them into 8 multi-heads, and perform scaled dot product attention calculation at the same time.

[0031] An embodiment of the present invention further provides a heat load prediction system for a heating system, comprising: a memory and a processor, wherein the memory stores a computer program executed by the processor, and the computer program executes a heat load prediction method for a heating system when executed by the processor.

[0032] An embodiment of the present invention further provides a storage medium, on which a computer program is stored, and the computer program executes a method for predicting heat load of a heating system when running.

[0033] The present invention constructs a training data set by fusing the original collected data and the new data generated by variational mode decomposition, and then constructs an AI model based on stacked bidirectional gated recurrent units and optimized multi-head attention mechanism to improve the prediction accuracy of the thermal load of the heating system. BRIEF DESCRIPTION OF THE DRAWINGS

[0034] In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the drawings required for use in the embodiments or the description of the prior art will be briefly introduced below. Obviously, the drawings described below are only embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on the provided drawings without paying creative work.

[0035] Figure 1 It is a structural schematic diagram of a heat load prediction device for a heating system based on multi-feature fusion according to an embodiment of the present invention;

[0036] Figure 2 Schematic diagram of the AI ​​model structure of an embodiment of the present invention. DETAILED DESCRIPTION

[0037] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0038] In order to make the above-mentioned objects, features and advantages of the present invention more obvious and easy to understand, the present invention is further described in detail below with reference to the accompanying drawings and specific embodiments.

[0039] Embodiment 1:

[0040] like Figure 1 As shown, an embodiment of the present invention provides a heat load prediction device for a heating system based on multi-feature fusion, comprising: a data processing module, a feature generation module, a multi-feature fusion data set construction module, an AI model construction module, and a heat load prediction module;

[0041] The data processing module is used to store the collected heating system data into the corresponding database and pre-process the heating system data.

[0042] As an implementation method of the embodiment of the present invention, the data processing module includes a data acquisition unit for collecting heating system data;

[0043] The heating system data include primary transient power, primary transient flow, primary water supply temperature, primary return water temperature, primary water supply pressure, primary return water pressure, secondary water supply temperature, secondary return water temperature, secondary water supply pressure, secondary return water pressure, indoor temperature, outdoor temperature, outdoor humidity, outdoor wind speed and solar radiation value per minute;

[0044] As an implementation method of the embodiment of the present invention, the data processing module further includes a data preprocessing unit for preprocessing the data collected by the heating system; wherein the data preprocessing unit includes:

[0045] The deletion subunit is used to delete duplicate data in the heating system data; in this embodiment, the data stored in the database is extracted, and the duplicate data collected multiple times at the same time are deleted.

[0046] The filling subunit is used to fill the missing data in the heating system data by using a linear interpolation filling method; in this embodiment, for missing data, the missing values ​​are filled in by using a linear interpolation filling method in chronological order. The linear interpolation filling method is to record the two values ​​and their time points before and after the missing value, and interpolate between the two time points at a frequency of minutes. The inserted value is filled by subtracting the previous value from the next value, and then evenly distributing it according to the number of inserted values ​​plus 1.

[0047] The outlier detection subunit is used to judge and process the outliers of the heating system data using the Z-score method. In this embodiment, the Z-score method is used to judge the outliers of each collected data. If it is an outlier, the nearest neighbor value replaces the outlier.

[0048] A feature generation module is used to process the preprocessed heating system data and generate new data features through variational mode decomposition;

[0049] As an implementation method of an embodiment of the present invention, the feature generation module includes:

[0050] The minute-to-hour data conversion unit is used to convert the primary transient power, primary transient flow, primary water supply temperature, primary return water temperature, primary water supply pressure, primary return water pressure, secondary water supply temperature, secondary return water temperature, secondary water supply pressure, secondary return water pressure, indoor temperature, outdoor temperature, outdoor humidity, outdoor wind speed and solar radiation value data of each minute into the corresponding primary transient power TP per hour by adding and averaging them. o , primary transient flow rate TFR, primary water supply temperature ST s , primary return water temperature RT s , primary water supply pressure PSP p , primary return water pressure PRP p , Secondary water supply temperature ST r , Secondary return water temperature RT r , Secondary water supply pressure PSP s , Secondary return water pressure PRP s , Indoor temperature T i , outdoor temperature To , Outdoor humidity H o , outdoor wind speed W o And solar radiation data SR o ;

[0051] Generate a new feature unit and convert the transient power TP once per hour o The data features are processed using variational mode decomposition (VMD) to obtain a series of intrinsic mode functions (IMFs) with specific center frequencies and bandwidths. The number of IMFs k is obtained by an adaptive calculation method based on the Bayesian information criterion (BIC).

[0052] Variational mode decomposition is a non-recursive method for signal processing, which decomposes complex multi-component signals into a series of intrinsic mode functions (IMFs) with specific center frequencies and bandwidths. Each IMF has a specific center frequency and bandwidth. VMD non-recursively decomposes the historical transient power data to obtain characteristic trends of k different frequencies.

[0053] The VMD method uses Wiener filtering for noise reduction and optimizes related parameters to obtain K estimated center frequencies. Then, each modal function and its center frequency are updated using alternating direction multiplication to minimize the total estimated bandwidth. The calculation formula is as follows:

[0054]

[0055] The above formula is subject to the constraint that k μ k =f. Among them, μ k ={μ1, μ2, ..., μ k} is the decomposed modal components of k different frequencies, ω k ={ω1, ω2, ..., ω k} is the center frequency of each mode, δ(t) is the unit pulse, and * is the convolution operation.

[0056] Initialize k to a value range of [2, 10], and then use the adaptive calculation method based on the Bayesian Information Criterion (BIC) to find the optimal k with the smallest BIC value. The calculation formula for the BIC value is as follows:

[0057] BIC=-sln(L)+kln(n)

[0058] Among them, L is the maximum likelihood value of the model, k is the number of parameters of the model, and n is the number of samples. The smaller the BIC value, the better the model performs in balancing goodness of fit and complexity.

[0059] The BIC calculation process is as follows:

[0060] The original transient power time series data is x t , with a length of N. For different numbers of IMFs k, from 2 to 10, VMD is used to decompose the data and obtain a set of k corresponding IMFs. The likelihood value L is then calculated. The likelihood value L represents the degree of fit of the force model to the data. Here, the reconstruction error is used to approximate the likelihood value. The calculation formula for the reconstruction error is as follows:

[0061]

[0062] The likelihood value formula is as follows:

[0063]

[0064] Among them, σ 2 is the variance of the noise.

[0065] Then the BIC value is calculated for each k, and the k with the smallest BIC value is selected as the optimal number of IMFs.

[0066] The multi-feature fusion data set construction module fuses the original collected data features and the generated new data features to construct a heat load prediction data set;

[0067] In this embodiment, the transient power (TP m ), primary transient flow rate (TFR), primary water supply temperature (ST s ), primary return water temperature (RT s ), primary water supply pressure (PSP p ), primary return water pressure (PRP p ), secondary water supply temperature (ST r ), secondary return water temperature (RT r ), secondary water supply pressure (PSP s ), secondary return water pressure (PRP s ), indoor temperature (T i ), outdoor temperature (T o ), outdoor humidity (H o ), outdoor wind speed (W o ), solar radiation value data (SR o ), and k IMFs, a multi-feature fusion heat load prediction dataset is constructed.

[0068] As an implementation of an embodiment of the present invention, the data set construction module includes:

[0069] Input feature building unit, used to convert the transient power TP of the previous 48 hours into o , the transient flow rate TFR in the first 48 hours, the water supply temperature ST in the first 48 hours s, the return water temperature RT in the first 48 hours s , the water supply pressure PSP in the first 48 hours p , the return water pressure PRP in the first 48 hours p , Secondary water supply temperature ST in the first 48 hours r , Secondary return water temperature RT in the first 48 hours r , Secondary water supply pressure PSP in the first 48 hours s , Secondary return water pressure PRP in the first 48 hours s 、Indoor temperature T in the first 48 hours i 、Outdoor temperature T in the first 48 hours o 、Outdoor humidity in the first 48 hours H o , outdoor wind speed W in the first 48 hours o 、Solar radiation data for the first 48 hours SR o and IMF1, IMF2...IMF calculated by VMD in the first 48 hours k , and the outdoor temperature T predicted for the next hour o , the outdoor humidity H predicted for the next hour o , the outdoor wind speed W predicted for the next hour o And the solar radiation data SR predicted for the next hour o As the input features of the dataset.

[0070] The prediction feature building unit is used to calculate the transient power TP in the next hour. o-1 As the prediction feature of the data set; the data set construction unit is used to integrate the input features and the prediction features to construct the heat load prediction data set; and the heat load prediction data set is divided into a training set, a validation set and a test set.

[0071] The constructed multi-feature fused heat load dataset of the heating system is randomly shuffled, 80% of the data is used as the training and validation sets, and the remaining 20% ​​of the data is used as the test dataset.

[0072] like Figure 2 As shown, the AI ​​model building module builds an AI model using stacked bidirectional gated recurrent units and optimized multi-head attention mechanism modules;

[0073] In this embodiment, a stacked bidirectional gated recurrent unit (SBiGRU) module is formed by stacking two layers of bidirectional gated recurrent units (BiGRU). For each training sample, the data dimension is 48*(15+k)+4. The training sample is input into the SBiGRU for calculation and the calculation result is output.

[0074] In this embodiment, the optimized multi-head attention mechanism module uses the output of the SBiGRU layer to calculate the query (Q), key (K), and value (V) of the multi-head attention mechanism layer. After that, the calculated query, key, and value are linearly transformed and divided into multiple heads, where the number of multiple heads is selected as 8, and then the scaled dot product attention calculation is performed.

[0075] The calculation formula of the 8-head attention mechanism is as follows:

[0076] MultiHead(Q,K,V)=Concat(head1,head2,...head8)W O

[0077]

[0078]

[0079] The calculation formula for zoom click attention is as follows:

[0080]

[0081] Attention weight calculation formula:

[0082]

[0083] Among them, the shapes of Q and K are (N, T, d k ), d k is the dimension of the key, T is the sequence length, and N is the batch size.

[0084] In addition, in order to solve the problems of low efficiency, long time and local information loss when SBiGRU learns sequence data, the residual mechanism and layer normalization are used to optimize the multi-head attention mechanism. That is, the output of the SBiGRU layer is concatenated with the output of the multi-head attention layer, and then layer normalization is performed.

[0085] The output module inputs the matrix trained by the optimized multi-head attention mechanism into a fully connected layer, and finally outputs the transient power prediction value of the heating system in the next hour.

[0086] The heat load prediction module is used to predict the heat load of the heating system based on the constructed AI model and the heat load prediction data set to obtain the heat load prediction results.

[0087] As an implementation of an embodiment of the present invention, the heat load prediction module includes:

[0088] The heat load prediction unit is used to input the test set into the trained AI model to predict the heat load of the heating system, obtain the heat load prediction results, and realize the heat load prediction for each test sample;

[0089] The model performance evaluation unit uses the mean absolute percentage error to evaluate the performance of the integrated learning model in the training phase and the test phase to obtain the evaluation results.

[0090] In this embodiment, MAPE (Mean Absolute Percentage Error) is used to evaluate the performance of the model during both the AI ​​model training and testing phases. The calculation formula for MAPE is as follows:

[0091]

[0092] Among them, y i is the actual value, is the predicted value, and n is the number of samples.

[0093] MAPE is an intuitive measure of model accuracy, ranging from 0 to 1. The closer the value is to 0, the higher the prediction accuracy of the model.

[0094] In this embodiment, the AI ​​model is trained with a training data set, and the performance of the AI ​​model is tested with a test set, and finally the MAPE value of the AI ​​model under the test set is output.

[0095] As an implementation of the embodiment of the present invention, a heating system control module is further included, which is used to control the heating system based on the result of the heat load prediction. In this embodiment, the thermal power per square meter obtained by dividing the instantaneous power predicted in the next hour by the total heating area can be used as the predicted heating amount of the heat source, thereby effectively controlling the heat source part of the heating system and reducing the energy consumption of the heat source.

[0096] The embodiment of the present invention performs variational mode decomposition on the heating load data and generates new data features. Compared with the traditional method of using directly collected data for heat load prediction, the present invention combines the directly collected data and the newly generated data to construct a new data set, thereby improving the quality of the heat load prediction data. Compared with simple machine learning and deep learning models, the present invention proposes an AI model based on stacked bidirectional gated recurrent units and optimized multi-head attention mechanism modules. The stacked bidirectional gated recurrent unit simultaneously considers information from the forward and backward sequences to learn the complex relationships in the time series, can effectively process time series data, and improves the comprehensive performance of the model. The optimized multi-head attention mechanism assigns different weights to the hidden states of the stacked bidirectional gated recurrent units, thereby enhancing the influence of different heating-related information on the heat load demand. Based on the fused multi-feature heat load prediction data set and the AI ​​model, the heat load prediction of the heating system is performed to obtain the heat load prediction result. The present invention can improve the accuracy of the short-term heat load prediction of the heating system and achieve the goal of reducing energy waste. The heating system controls the corresponding hardware through the heat load value output by the AI ​​model to achieve efficient energy management

[0097] Embodiment 2:

[0098] The embodiment of the present invention further provides a method for predicting heat load of a heating system, comprising:

[0099] Step S1, preprocessing the collected heating system data;

[0100] Step S2, processing the preprocessed heating system data, and generating new data features through variational mode decomposition;

[0101] Step S3, integrating the original collected data features and the generated new data features to construct a heat load prediction data set;

[0102] Step S4, constructing an AI model using stacked bidirectional gated recurrent units and optimized multi-head attention mechanism modules;

[0103] Step S5: Based on the constructed AI model and the heat load prediction data set, the heat load of the heating system is predicted to obtain the heat load prediction result.

[0104] As an implementation mode of an embodiment of the present invention, the collected heating system data include: primary transient power, primary transient flow, primary water supply temperature, primary return water temperature, primary water supply pressure, primary return water pressure, secondary water supply temperature, secondary return water temperature, secondary water supply pressure, secondary return water pressure, indoor temperature, outdoor temperature, outdoor humidity, outdoor wind speed and solar radiation value per minute in each heating system.

[0105] As an implementation of the embodiment of the present invention, step S2 includes:

[0106] Step S21, convert the primary transient power, primary transient flow, primary water supply temperature, primary return water temperature, primary water supply pressure, primary return water pressure, secondary water supply temperature, secondary return water temperature, secondary water supply pressure, secondary return water pressure, indoor temperature, outdoor temperature, outdoor humidity, outdoor wind speed and solar radiation value per minute into the corresponding primary transient power TP per hour by adding and averaging them. o , primary transient flow rate TFR, primary water supply temperature ST s , primary return water temperature RT s , primary water supply pressure PSP p , primary return water pressure PRP p , Secondary water supply temperature ST r , Secondary return water temperature RT r , Secondary water supply pressure PSP s , Secondary return water pressure PRP s , Indoor temperature T i , outdoor temperature T o , Outdoor humidity H o , outdoor wind speed W o And solar radiation data SR o ;

[0107] Step S22: convert the hourly transient power TP o The data features are processed using variational mode decomposition to obtain a series of intrinsic mode functions with specific center frequencies and bandwidths. The number of IMFs k is obtained by an adaptive calculation method based on the Bayesian information criterion;

[0108] Step S23: convert all minute data into hourly data, and use k natural mode functions with specific center frequencies and bandwidths obtained based on variational mode decomposition as new data features to describe the heating system.

[0109] As an implementation method of an embodiment of the present invention, a stacked bidirectional gated recurrent unit is composed of two layers of stacked bidirectional gated recurrent units. For each training sample, the data dimension is 48*(15+k)+4. The training sample is input into SBiGRU for calculation and the calculation result is output; the optimized multi-head attention mechanism module uses the output result after calculation of the SBiGRU layer to calculate the query, key, and value of the multi-head attention mechanism layer, performs linear transformation on the calculated query, key, and value, and divides it into 8 multi-heads, and performs scaled dot product attention calculation at the same time.

[0110] Embodiment 3:

[0111] An embodiment of the present invention further provides a heat load prediction system for a heating system, comprising: a memory and a processor, wherein the memory stores a computer program executed by the processor, and the computer program executes a heat load prediction method for a heating system when executed by the processor.

[0112] Embodiment 4:

[0113] An embodiment of the present invention further provides a storage medium, on which a computer program is stored, and the computer program executes a method for predicting heat load of a heating system when running.

[0114] The embodiments described above are only descriptions of the preferred embodiments of the present invention and are not intended to limit the scope of the present invention. Without departing from the design spirit of the present invention, various modifications and improvements made to the technical solutions of the present invention by ordinary technicians in this field should all fall within the protection scope determined by the claims of the present invention.

Claims

1. A heat load prediction device for a heating system, characterized in that: include: A data processing module, used for preprocessing the collected heating system data; A feature generation module is used to process the preprocessed heating system data and generate new data features through variational mode decomposition; The multi-feature fusion data set construction module is used to fuse the original collected data features and the generated new data features to construct a heat load prediction data set; AI model building module, used to build AI models using stacked bidirectional gated recurrent units and optimized multi-head attention mechanism modules; The heat load prediction module is used to predict the heat load of the heating system based on the constructed AI model and the heat load prediction data set to obtain the heat load prediction results.

2. The heat load prediction device for a heating system according to claim 1, characterized in that: The collected heating system data include: primary transient power, primary transient flow, primary water supply temperature, primary return water temperature, primary water supply pressure, primary return water pressure, secondary water supply temperature, secondary return water temperature, secondary water supply pressure, secondary return water pressure, indoor temperature, outdoor temperature, outdoor humidity, outdoor wind speed and solar radiation value in each heating system every minute.

3. The heat load prediction device for a heating system according to claim 2, characterized in that: The feature generation module comprises: The minute-to-hour data conversion unit is used to convert the primary transient power, primary transient flow, primary water supply temperature, primary return water temperature, primary water supply pressure, primary return water pressure, secondary water supply temperature, secondary return water temperature, secondary water supply pressure, secondary return water pressure, indoor temperature, outdoor temperature, outdoor humidity, outdoor wind speed and solar radiation value per minute into the corresponding primary transient power TP per hour by adding and averaging them. o , primary transient flow rate TFR, primary water supply temperature ST s , primary return water temperature RT s , primary water supply pressure PSP p , primary return water pressure PRP p , Secondary water supply temperature ST r , Secondary return water temperature RT r , Secondary water supply pressure PSP s , Secondary return water pressure PRP s , Indoor temperature T i , outdoor temperature T o , Outdoor humidity H o , outdoor wind speed W o And solar radiation data SR o ; Generate a new feature unit to convert the transient power TP once per hour o The data features are processed using variational mode decomposition to obtain a series of intrinsic mode functions with specific center frequencies and bandwidths. The number of IMFs k is obtained by an adaptive calculation method based on the Bayesian information criterion; The heating system feature generation module is used to convert all minute data into hourly data and obtain k intrinsic mode functions with specific center frequency and bandwidth based on variational mode decomposition as new data features to describe the heating system.

4. The heat load prediction device for a heating system according to claim 3, characterized in that: The stacked bidirectional gated recurrent unit is composed of two layers of bidirectional gated recurrent units. For each training sample, its data dimension is 48*(15+k)+4. The training sample is input into SBiGRU for calculation and outputs the calculation result. The optimized multi-head attention mechanism module will use the output results of the SBiGRU layer to calculate the query, key, and value of the multi-head attention mechanism layer, perform linear transformation on the calculated query, key, and value, and split them into 8 multi-heads, and perform scaled dot product attention calculation at the same time.

5. A method for predicting heat load of a heating system, characterized in that: include: Step S1, preprocessing the collected heating system data; Step S2, processing the preprocessed heating system data, and generating new data features through variational mode decomposition; Step S3, integrating the original collected data features and the generated new data features to construct a heat load prediction data set; Step S4, constructing an AI model using stacked bidirectional gated recurrent units and optimized multi-head attention mechanism modules; Step S5: Based on the constructed AI model and the heat load prediction data set, the heat load of the heating system is predicted to obtain the heat load prediction result.

6. The method for predicting heat load of a heating system according to claim 4, characterized in that: The collected heating system data include: primary transient power, primary transient flow, primary water supply temperature, primary return water temperature, primary water supply pressure, primary return water pressure, secondary water supply temperature, secondary return water temperature, secondary water supply pressure, secondary return water pressure, indoor temperature, outdoor temperature, outdoor humidity, outdoor wind speed and solar radiation value in each heating system every minute.

7. The method for predicting heat load of a heating system according to claim 5, characterized in that: Step S2 includes: Step S21, convert the primary transient power, primary transient flow, primary water supply temperature, primary return water temperature, primary water supply pressure, primary return water pressure, secondary water supply temperature, secondary return water temperature, secondary water supply pressure, secondary return water pressure, indoor temperature, outdoor temperature, outdoor humidity, outdoor wind speed and solar radiation value per minute into the corresponding primary transient power TP per hour by adding and averaging them. o , primary transient flow rate TFR, primary water supply temperature ST s , primary return water temperature RT s , primary water supply pressure PSP p , primary return water pressure PRP p , Secondary water supply temperature ST r , Secondary return water temperature RT r , Secondary water supply pressure PSP s , Secondary return water pressure PSP s , Indoor temperature T i , outdoor temperature T o , Outdoor humidity H o , outdoor wind speed W o And solar radiation data SR o ; Step S22: convert the hourly transient power TP o The data features are processed using variational mode decomposition to obtain a series of intrinsic mode functions with specific center frequencies and bandwidths. The number of IMFs k is obtained by an adaptive calculation method based on the Bayesian information criterion; Step S23: convert all minute data into hourly data, and use k natural mode functions with specific center frequencies and bandwidths obtained based on variational mode decomposition as new data features to describe the heating system.

8. The method for predicting heat load of a heating system according to claim 7, characterized in that: The stacked bidirectional gated recurrent unit is composed of two layers of bidirectional gated recurrent units. For each training sample, its data dimension is 48*(15+k)+4. The training sample is input into SBiGRU for calculation and outputs the calculation result. The optimized multi-head attention mechanism module will use the output results of the SBiGRU layer to calculate the query, key, and value of the multi-head attention mechanism layer, perform linear transformation on the calculated query, key, and value, and split them into 8 multi-heads, and perform scaled dot product attention calculation at the same time.

9. A heat load prediction system for a heating system, characterized in that: include: A memory and a processor, wherein the memory stores a computer program executed by the processor, and when the computer program is executed by the processor, the method for predicting the heat load of a heating system as described in any one of claims 5 to 8 is executed.

10. A storage medium, characterized in that: The storage medium stores a computer program, which, when running, executes the method for predicting the heat load of a heating system according to any one of claims 5 to 8.

Images