Energy consumption prediction monitoring system based on BIM

Abstract

The invention discloses an energy consumption prediction monitoring system based on BIM, and relates to the technical field of intelligent building energy management, which comprises the steps of defining an optimization objective function based on modal weight, performing iterative updating by using an OIALM algorithm, obtaining a low-rank matrix, outputting initial prediction energy consumption through a prediction model, and optimizing model parameters by using a golden vehicle optimization algorithm. Generating final predicted energy consumption; through a three-mode alignment and sparse suppression tensor modeling method based on BIM and Internet of Things data, building physical semantics and dynamic environment features are effectively fused, the accuracy and robustness of energy consumption feature expression are improved, a prediction framework fusing key time node rearrangement, attention enhancement Tucker decomposition and OIALM low-rank optimization is constructed, and the prediction efficiency is improved. And the modeling capability and the energy consumption prediction precision of the unsteady-state thermal process are improved.

Description

Technical Field

[0001] This invention relates to the field of intelligent building energy management technology, and in particular to a BIM-based energy consumption prediction and monitoring system. Background Technology

[0002] BIM technology can not only provide geometric and topological information of building components, but also provide thermal performance, equipment parameters and operating conditions. It is an important data source for realizing refined analysis of building energy consumption. With the development of Internet of Things technology, external meteorological and environmental data can be collected through high-frequency acquisition by multiple types of sensors and real-time data platforms, forming a dynamic environmental input sequence. This enables building energy consumption modeling to gradually evolve from static experience models to data-driven models, improving the timeliness and granularity of building energy consumption simulation.

[0003] However, existing technologies still have shortcomings. Current methods directly input BIM geometry, thermal parameters and environmental data into the model, lacking a time-space alignment mechanism. This results in the physical properties of components not being able to accurately correspond to the time dimension with real-time environmental data, causing intermodal asynchrony in predicted features. Existing BIM energy consumption prediction lacks the mining of low-rank structures and does not combine OIALM optimization algorithms for optimization, making the model susceptible to noise and with weak generalization ability. It also fails to adequately characterize key dynamic phenomena such as peak electricity consumption and wall heat storage. Summary of the Invention

[0004] In view of the aforementioned existing problems, the present invention is proposed.

[0005] Therefore, this invention provides a BIM-based energy consumption prediction and monitoring system to address the shortcomings of existing technologies. Existing methods directly input BIM geometry, thermal parameters, and environmental data into the model, lacking a time-space alignment mechanism. This results in the physical properties of components failing to accurately correspond to the time dimension with real-time environmental data, leading to intermodal asynchrony in prediction features. Existing BIM energy consumption prediction lacks the ability to mine low-rank structures and does not incorporate the OIALM optimization algorithm for optimization, making the model susceptible to noise and exhibiting weak generalization ability. It also fails to adequately characterize key dynamic phenomena such as peak electricity consumption and wall heat storage.

[0006] To solve the above-mentioned technical problems, the present invention provides the following technical solution: This invention provides a BIM-based energy consumption prediction and monitoring system, which includes: The unified spatiotemporal matrix and dynamic energy consumption driving sequence construction module is used to obtain BIM models through API interfaces and real-time meteorological data through IoT platforms. The BIM models and IoT platforms are preprocessed to form a unified spatiotemporal matrix. Based on the unified spatiotemporal matrix, the transient heat flux energy of components, transient load of rooms and HVAC equivalent load are calculated to obtain the dynamic energy consumption driving sequence. The sparse suppression and smoothing tensor generation module is used to construct a three-modal feature tensor based on the structured energy consumption parameter matrix, extended environment sequence, and dynamic energy consumption driving sequence. Based on the three-modal feature tensor, the energy value is calculated for each feature dimension. The sparse suppression tensor is obtained by filtering. The time modes in the sparse suppression tensor are linearly interpolated to obtain the local gradient. The optimal node sequence is constructed. The time order is rearranged based on the optimal node sequence to obtain the rearranged sparse suppression tensor. Random perturbation is added to each time point to generate a smooth signal and obtain the smoothed sparse suppression tensor. The attention-enhanced feature reconstruction module is used to calculate the first-order gradient of the building physics feature mode in the three-modal feature tensor based on the smooth sparse suppression tensor, generate a column-uniform matrix, and perform SVD decomposition to obtain the factor matrix. The core tensor is obtained through Tucker inverse mapping. Row vectors are extracted based on the factor matrix and attention weights are calculated. A diagonal weighted matrix is ​​constructed to reconstruct the core tensor along the time direction. The final enhanced features are obtained through Tucker reconstruction, and modal weights are constructed. The low-rank optimization and energy consumption prediction output module is used to define the optimization objective function based on modal weights, use the OIALM algorithm for iterative updates to obtain the low-rank matrix, output the initial predicted energy consumption through the prediction model, and use the Golden Jackal optimization algorithm to optimize the model parameters to generate the final predicted energy consumption.

[0007] As a preferred embodiment of the BIM-based energy consumption prediction and monitoring system of the present invention, wherein: the rearrangement of the time sequence based on the optimal node sequence to obtain the rearranged sparse suppression tensor includes: The structured energy consumption parameter matrix, extended environment sequence, and dynamic energy consumption driving sequence are horizontally concatenated along the feature dimension to obtain a unified concatenation tensor. The feature dimensions of the unified concatenation tensor are then divided to obtain a three-modal feature tensor. The energy value of each feature dimension of the three-modal feature tensor is calculated. If the energy value is less than the energy value threshold, the corresponding feature dimension is deleted to obtain a sparse suppression tensor. The local gradient is calculated and sorted in descending order. The time point corresponding to the maximum local gradient is selected to obtain the optimal node sequence. The temporal order in the sparse suppression tensor is rearranged based on the optimal node sequence to obtain the rearranged sparse suppression tensor. Random perturbation is added to each time point in the rearranged sparse suppression tensor, and a smoothing signal is calculated. The smoothing signal is then used to replace the rows in the expanded matrix to obtain the smoothed sparse suppression tensor.

[0008] As a preferred embodiment of the BIM-based energy consumption prediction and monitoring system of the present invention, the step of generating a column-consistent matrix and performing SVD decomposition to obtain a factor matrix includes: Based on the smooth sparse suppression tensor, the first-order gradient of the building physical feature modes is calculated to generate a column-uniform matrix. The uniform matrix is ​​then decomposed by SVD to obtain the time mode factor matrix. The three-mode tensor is expanded by mode-2 and mode-3 to obtain the building physical feature matrix and the environmental and operational feature matrix. The two matrices are then decomposed by SVD to obtain the building physical feature factor matrix and the environmental and operational feature factor matrix.

[0009] As a preferred embodiment of the BIM-based energy consumption prediction and monitoring system of the present invention, the construction of a diagonal weighted matrix to reconstruct the core tensor along the time direction, and obtaining the final enhanced features through Tucker reconstruction, includes: The core tensor is obtained by performing Tucker inverse mapping on the time modality factor matrix, the building physics feature factor matrix, and the environmental and operational feature factor matrix. The row vectors corresponding to each time point are extracted from the temporal modality factor matrix, the attention weights are calculated, a diagonal weighted matrix is ​​constructed by constructing a diagonal matrix, the core tensor is reconstructed in the temporal direction to obtain the attention-enhanced core tensor, and the attention-enhanced core tensor is reconstructed by Tucker to obtain the final enhanced features.

[0010] As a preferred embodiment of the BIM-based energy consumption prediction and monitoring system of the present invention, the step of constructing modal weights based on the final enhanced features includes: Based on the time mode, building physics mode, and environmental and operational mode in the final enhanced features, matrix expansion is performed to obtain the mode matrix. The energy intensity is calculated for each mode matrix, and the mode weights are constructed.

[0011] As a preferred embodiment of the BIM-based energy consumption prediction and monitoring system of the present invention, wherein: the defined optimization objective function is iteratively updated using the OIALM algorithm to obtain a low-rank matrix, including: Based on modal weights, optimization is defined as the objective function. The modal matrix is ​​folded to obtain an initialization matrix. The initialization matrix is ​​then iteratively updated using the OIALM algorithm to obtain a low-rank matrix.

[0012] As a preferred embodiment of the BIM-based energy consumption prediction and monitoring system of the present invention, the step of outputting the initial predicted energy consumption through the prediction model includes: A mode folding operation is performed on the low-rank matrix to obtain a stable time series. Row right padding is then performed on the series to obtain the TCN input sequence. This sequence is input into the TCN network to obtain a deep feature vector. The deep feature vector is then input into a bidirectional GRU to obtain the hidden state. An importance score is obtained at each time point through an attention mechanism. The attention weights of the importance scores are calculated using the Softmax function. The importance scores and attention weights are then weighted and summed to obtain a context vector. This context vector is then input into the fully connected output layer to obtain the initial prediction energy consumption.

[0013] As a preferred embodiment of the BIM-based energy consumption prediction and monitoring system of the present invention, the step of optimizing the model parameters using the Golden Jackal optimization algorithm to generate the final predicted energy consumption includes: Initialize the hyperparameter space to be optimized, including the kernel size, dilation rate and regularization coefficient of TCN, the number of hidden units and layers of bidirectional GRU, and the key vector dimension of the attention mechanism. Use the Golden Jackal optimization algorithm to iteratively update the hyperparameter space to be optimized to obtain the optimal hyperparameter space and generate the final predicted energy consumption.

[0014] As a preferred embodiment of the BIM-based energy consumption prediction and monitoring system of the present invention, the step of obtaining a BIM model through an API interface and acquiring real-time meteorological data through an IoT platform, and preprocessing the BIM model and the IoT platform respectively, includes: By acquiring BIM models through API interfaces and real-time meteorological data through IoT platforms, the BIM models and IoT platforms are preprocessed to obtain structured energy consumption parameter matrices and extended environmental sequences.

[0015] As a preferred embodiment of the BIM-based energy consumption prediction and monitoring system of the present invention, wherein: the calculation of transient heat flux energy of components, transient room load, and HVAC equivalent load yields a dynamic energy consumption driving sequence, including: The structured energy consumption parameter matrix and the extended environment sequence are concatenated to obtain an alignment matrix. All alignment matrices within time t are then sorted according to time. The dimensions are stacked to obtain a unified spatiotemporal matrix, and a dynamic energy consumption driving sequence is constructed. This includes calculating the transient heat flux energy of components, the transient load of the room, and the equivalent load of HVAC. Combined with the solar heat gain, the normalization process is performed and a vector horizontal splicing operation is carried out to obtain the dynamic energy consumption driving sequence.

[0016] The beneficial effects of this invention are as follows: This invention defines an optimization objective function based on modal weights, uses the OIALM algorithm for iterative updates to obtain a low-rank matrix, outputs the initial predicted energy consumption through the prediction model, and uses the Golden Jackal optimization algorithm to optimize the model parameters to generate the final predicted energy consumption; it effectively integrates building physical semantics and dynamic environmental features, improves the accuracy and robustness of energy consumption feature expression, and enhances the modeling ability and energy consumption prediction accuracy for unsteady thermal processes. Attached Figure Description

[0017] To more clearly illustrate the technical solutions of the embodiments of the present invention, the drawings used in the following description of the embodiments will be briefly introduced. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0018] Figure 1 This is a flowchart illustrating the BIM-based energy consumption prediction and monitoring system in Example 1.

[0019] Figure 2 This is a schematic diagram of the BIM-based energy consumption prediction and monitoring system in Example 1. Detailed Implementation

[0020] To make the above-mentioned objects, features and advantages of the present invention more apparent and understandable, the specific embodiments of the present invention will be described in detail below with reference to the accompanying drawings.

[0021] Many specific details are set forth in the following description in order to provide a full understanding of the invention. However, the invention may also be practiced in other ways different from those described herein, and those skilled in the art can make similar extensions without departing from the spirit of the invention. Therefore, the invention is not limited to the specific embodiments disclosed below.

[0022] Secondly, the term "one embodiment" or "embodiment" as used herein refers to a specific feature, structure, or characteristic that may be included in at least one implementation of the present invention. The phrase "in one embodiment" appearing in different places in this specification does not necessarily refer to the same embodiment, nor is it a single or selective embodiment that is mutually exclusive with other embodiments.

[0023] Example 1, referring to Figure 1 and Figure 2 This is the first embodiment of the present invention, which provides a BIM-based energy consumption prediction and monitoring system, including: S1, the unified spatiotemporal matrix and dynamic energy consumption driving sequence construction module, is used to obtain BIM model through API interface and real-time meteorological data through IoT platform, preprocess BIM model and IoT platform respectively to form unified spatiotemporal matrix, and calculate transient heat flux energy of components, transient load of room and HVAC equivalent load based on unified spatiotemporal matrix to obtain dynamic energy consumption driving sequence. Specifically, the BIM model is obtained through an API interface, and real-time meteorological data is acquired through an IoT platform. Preprocessing is then performed on both the BIM model and the IoT platform, including: The BIM model is obtained through the API interface and real-time meteorological data is obtained through the IoT platform. The BIM model and the IoT platform are preprocessed respectively to obtain the structured energy consumption parameter matrix and the extended environmental sequence. The preprocessing of the BIM model includes extracting spatial classifications from the BIM model, such as the spaces to which buildings, floors, rooms, and components (such as walls, roofs, floors, windows, and doors) belong, and arranging them horizontally to obtain a hierarchical sequence. Based on the hierarchical sequence, the geometric parameters of each component are extracted (related data is obtained synchronously when acquiring BIM model data through API), such as area, thickness, orientation, and tilt angle. The thermal parameters of each component are also extracted, such as thermal conductivity, specific heat capacity, total thermal resistance, density, and solar heat gain. The HVAC (such as VRV, fan coil units, and fresh air units) equipment parameters of the space to which each component belongs are also extracted, such as rated power, air volume, supply air temperature, and COP value. A structured energy consumption parameter matrix is ​​constructed from geometric parameters, thermal parameters, and equipment parameters, using the following formula: , in, , , , as well as For the first Data from the geometric parameters, thermal parameters, and equipment parameters of each component. The number of components; The real-time meteorological data includes, for example, temperature, humidity, wind speed, radiation, and cloud cover. The preprocessing of real-time meteorological data includes filling missing values ​​in the real-time meteorological data using bidirectional spline interpolation, and extracting meteorological variables such as radiation and wind speed from the filled real-time meteorological data to normalize the meteorological components, thereby obtaining preprocessed real-time meteorological data. Based on preprocessed real-time meteorological data, the temperature difference between time t and time t-1 is calculated to obtain the temperature change rate. The radiation difference between time t and time t-1 is calculated to obtain the radiation increment. The radiation increment within time t is accumulated to obtain the cumulative sunshine amount. By horizontally arranging the cumulative solar radiation, temperature change rate, radiation variables, and preprocessed real-time meteorological data, an extended environmental sequence is obtained.

[0024] By performing independent preprocessing on the BIM model and real-time meteorological data, and constructing a structured energy consumption parameter matrix and extended environmental sequence, a deep coupling between the static attributes of the building and the external dynamic environment is achieved. The extraction of spatial hierarchy, component geometry, thermal properties, and HVAC equipment performance in the preprocessing process significantly enhances the model's ability to express the multi-source heterogeneity of the building energy system, enabling energy consumption characteristics to present clear physical meaning at the input stage. By employing bidirectional interpolation, normalization, and incremental calculation on meteorological data, abnormal fluctuations caused by sensor noise can be significantly reduced. Furthermore, the construction of cumulative solar radiation and temperature change rate makes the environmental input more time-sensitive. The resulting structured parameter matrix and extended environmental sequence lay a stable and highly interpretable data foundation for subsequent sequential energy consumption extrapolation, achieving a unified expression of the building's static attributes and external dynamic factors, and significantly improving the discrimination accuracy of subsequent prediction models.

[0025] Furthermore, the transient thermal energy of the components, the transient load of the room, and the equivalent HVAC load are calculated to obtain the dynamic energy consumption drive sequence, including: The structured energy consumption parameter matrix and the extended environment sequence are concatenated to obtain the alignment matrix, as shown in the formula: , in, For the alignment matrix, To extend the environment sequence, This is a structured energy consumption parameter matrix; All alignment matrices within time t are sorted by time. By stacking dimensions, a unified spatiotemporal matrix is ​​obtained; Based on a unified spatiotemporal matrix, a dynamic energy consumption driving sequence is constructed, including calculating the transient heat flux energy of components, the transient load of the room, and the equivalent load of HVAC. Combined with solar heat gain, normalization is performed, and a vector horizontal concatenation operation is conducted to obtain the dynamic energy consumption driving sequence, as shown in the formula: , in, For the transient thermal energy of the component, For the first The transient heat transfer energy of a building component (e.g., a wall). For time, outdoor temperature For component temperature, The total thermal resistance of the component, For the first The amount of solar heat received by a light-transmitting component (such as a window). For the room transient load, For HVAC equivalent load, The energy efficiency ratio (EER) is obtained from the parameters of the HVAC equipment. and This refers to the number of building components and the number of light-transmitting components.

[0026] By constructing a dynamic energy consumption driving sequence, the transient thermal behavior of building components is integrated with room heat load and HVAC operation response in a unified spatiotemporal matrix, which has significant systemic value. On the one hand, the structured energy consumption parameter matrix and the extended environmental sequence are spliced ​​together through an alignment matrix, achieving precise alignment of building physical characteristics and external disturbances in the time dimension, effectively eliminating the time offset problem between modes. On the other hand, by dynamically deriving transient heat flux energy, transient load, and equivalent HVAC load, the energy consumption driving sequence has a clear physical causal logic. Especially under conditions such as sudden temperature rise, sudden wind speed change, and strong radiation, it can present the real thermal inertia effect and equipment response law. The unified driving sequence after vector splicing has both continuity and maintains physical consistency, providing a profound dynamic driving force expression for energy consumption prediction, and significantly improving the model's sensitivity and generalizability to fluctuations in heating and cooling loads.

[0027] S2, the sparse suppression and smoothing tensor generation module, is used to construct a three-modal feature tensor based on the structured energy consumption parameter matrix, the extended environment sequence, and the dynamic energy consumption driving sequence. Based on the three-modal feature tensor, the energy value is calculated for each feature dimension. The sparse suppression tensor is obtained by filtering. The time modes in the sparse suppression tensor are linearly interpolated to obtain the local gradient. The optimal node sequence is constructed. The time order is rearranged based on the optimal node sequence to obtain the rearranged sparse suppression tensor. Random perturbation is added to each time point to generate a smooth signal and obtain the smoothed sparse suppression tensor. Specifically, the temporal order is rearranged based on the optimal node sequence to obtain the rearranged sparse suppression tensor, including: The structured energy consumption parameter matrix (with time-broadcast extension), extended environmental sequence, and dynamic energy consumption driving sequence are normalized respectively. The normalized data are then horizontally concatenated along the feature dimensions to obtain a unified concatenated tensor. The feature dimensions of the unified concatenated tensor are then divided to obtain a three-modal feature tensor, including time mode, building physical feature mode (e.g., wall thermal resistance, window-to-wall ratio, equipment rated power, etc.), and environmental and operational feature mode (e.g., external temperature, humidity, solar radiation, operating conditions). Calculate the energy value for each feature dimension of the trimodal feature tensor. If the energy value is less than the energy value threshold (set based on empirical rules), delete the corresponding feature dimension to obtain the sparse suppression tensor, as shown in the formula: , in, For the first Energy values ​​for each feature dimension For the number of time points, For the first Each feature dimension; Linear interpolation is performed on the time modes in the sparse suppression tensor to obtain the time mode interpolation. The difference between the time mode interpolation at time t and time t-1 is calculated to obtain the local gradient. Divide the timestamps into equal intervals to obtain R equal intervals. Sort the local gradients of each equal interval in descending order, select the time point corresponding to the maximum local gradient, set it as the optimal node, and arrange the optimal nodes horizontally to obtain the optimal node sequence. The temporal order in the sparse suppression tensor is rearranged based on the optimal node sequence (the sparse suppression tensor is arranged according to time t, and the rearranged sparse suppression tensor is arranged according to the time point corresponding to the optimal node sequence), resulting in the rearranged sparse suppression tensor. To calculate the smoothed signal, a random perturbation is added to each time point in the rearranged sparse suppression tensor, and the formula is: , in, For smoothing signals, As the expected value, To rearrange the sparse suppression tensor, For random perturbations; By replacing the rows in the expansion matrix with the smoothed signal, we obtain the smoothed expansion matrix, which in turn yields the smoothed sparse suppression tensor. The expansion matrix refers to the expansion matrix obtained by performing a mode-1 expansion (meaning expansion along the time modes) on the rearranged sparse suppression tensor.

[0028] By constructing a sparse suppression tensor and an optimal node sequence, key challenges such as dimensional redundancy of multimodal features, temporal information noise, and difficulty in capturing dynamic mutation points are addressed. Dimensional sparsity is achieved through energy value screening, which effectively eliminates low-contribution features, enabling the model to maintain expressiveness while achieving higher computational efficiency. Optimal node screening using local gradients can accurately identify key perturbation points in long time series, such as moments of rapid increases in heat load, sudden changes in wind speed, or sharp increases in radiation, resulting in a more structured dynamic distribution in the time-rearranged sequence. Generating smooth signals through random perturbations can suppress environmental noise while preserving dynamic features, thereby constructing a stable sparse suppression tensor. This process effectively improves the structural clarity of temporal representation, allowing subsequent tensor decomposition to operate in a cleaner space and significantly enhancing the model's ability to capture mutation behavior.

[0029] S3, Attention Enhancement Feature Reconstruction Module, is used to calculate the first-order gradient of the building physics feature mode in the three-modal feature tensor based on the smooth sparse suppression tensor, generate a column-consistent matrix, and perform SVD decomposition to obtain the factor matrix. The core tensor is obtained through Tucker inverse mapping. Row vectors are extracted based on the factor matrix and attention weights are calculated. A diagonal weighted matrix is ​​constructed to reconstruct the core tensor along the time direction. The final enhanced features are obtained through Tucker reconstruction, and modal weights are constructed. Specifically, a column-consistent matrix is ​​generated, and SVD decomposition is performed to obtain the factor matrix, including: Based on the smooth sparse suppression tensor, the first-order gradient of the building physical feature modes is calculated, generating a column-uniform matrix, as shown in the formula: , in, For the first-order gradient, To smooth the sparse suppression tensor, Elements of the column-consistent matrix, To smooth the sparse suppression tensor, For time indexing; The uniformized matrix is ​​decomposed by SVD to obtain the time modality factor matrix. The three-modal tensor (building physical feature mode and environmental and operational feature mode) is expanded by mode-2 and mode-3 to obtain the building physical feature matrix and the environmental and operational feature matrix. The two matrices are then decomposed by SVD to obtain the building physical feature factor matrix and the environmental and operational feature factor matrix.

[0030] By generating a column-consistent matrix and performing singular value decomposition, highly representative latent factor structures can be extracted from the multimodal tensor, enabling deep decoupling of building physical features, environmental operational features, and temporal features. By unifying the first-order gradients of the building physical feature modes, the differences in feature variations among different components are mapped uniformly, thereby reducing the distribution bias problem of structural features. By performing singular value decomposition on the three modes respectively, a factor matrix with a low-rank structure can be obtained, allowing building attributes, temporal variations, and environmental disturbances to be represented in a compact manner. This decomposition mechanism not only improves the stability of data representation but also provides a standardized basis for subsequent Tucker reconstruction, enabling building operational features to be expressed modally independently while maintaining overall correlation, significantly improving the structured interpretability of the overall model.

[0031] Furthermore, a diagonal weighted matrix is ​​constructed to reconstruct the core tensor along the time direction, and the final enhanced features are obtained through Tucker reconstruction, including: The core tensor is obtained by performing Tucker inverse mapping on the time modality factor matrix, the building physics feature factor matrix, and the environmental and operational feature factor matrix. Extract the row vector corresponding to each time point from the time modality factor matrix, and calculate the attention weight using the following formula: , in, For time The significance score, For attention weight vectors, For transpose, For time The row vector, For time The significance score, For time Attention weights; Based on attention weights, a diagonal weighted matrix is ​​constructed through the operation of constructing a diagonal matrix. This matrix is ​​then used to reconstruct the core tensor in the time direction, resulting in the attention-enhanced core tensor, as shown in the formula: , in, To enhance the attention core tensor, For the core tensor, This indicates matrix-tensor multiplication along the direction of the time mode (first mode). It is a diagonal weighted matrix; The attention-enhancing core tensor is reconstructed using Tucker to obtain the final enhanced features.

[0032] By constructing a diagonal weighted matrix and performing Tucker reconstruction, the saliency of temporal features is emphasized, enabling the core tensor to achieve dynamic attention enhancement. By extracting saliency scores from the time factor, the contribution of key moments in energy consumption changes can be effectively highlighted, giving the model a stronger perception capability in handling critical stages such as peak load, switching between hot and cold sources, and strong radiation disturbances. The weighted core tensor has higher discriminative power in the time dimension, making the final enhanced features after Tucker reconstruction exhibit significant temporal sensitivity and more complete modal expression capability. This process strengthens the dynamic coupling between the three modes, giving the enhanced features a more distinct pattern structure and temporal discriminability, providing high-quality input for subsequent optimization and prediction, and effectively improving the robustness of energy consumption prediction.

[0033] Furthermore, based on the final enhanced features, modal weights are constructed, including: The temporal mode, building physics mode, and environmental and operational mode are extracted from the final enhanced features. Each mode is then expanded to obtain a mode matrix. The energy intensity of each mode matrix is ​​calculated, and mode weights are constructed using the following formula: , in, For modality energy intensity, For modality The modal matrix, It is the Frobenius norm. For modality Modal weights, For modality energy value This represents the number of modes.

[0034] By calculating modal weights for the final enhanced features, the importance of each modality is quantitatively expressed, providing a goal-oriented weighted basis for subsequent low-rank optimization. By calculating the energy intensity of the time mode, building physics mode, and environmental operation mode respectively, the contribution of each mode to energy consumption changes can be accurately reflected. The introduction of modal weights can strengthen key modes, making the low-rank decomposition process more focused on decisive features under the guidance of weights, while weakly contributing modes are naturally weakened, thereby improving the efficiency and accuracy of the overall decomposition. This mechanism can significantly improve the structural stability of the model, making the further OIALM optimization steps more targeted and robust.

[0035] S4, the low-rank optimization and energy consumption prediction output module, is used to define the optimization objective function based on modal weights, use the OIALM algorithm for iterative updates to obtain the low-rank matrix, output the initial predicted energy consumption through the prediction model, and use the Golden Jackal optimization algorithm to optimize the model parameters to generate the final predicted energy consumption. Specifically, an optimization objective function is defined, and the OIALM algorithm is used for iterative updates to obtain a low-rank matrix, including: Based on modal weights, optimization is defined as the objective function. The modal matrix undergoes a mode-folding operation to obtain an initialization matrix. This initialization matrix is ​​then iteratively updated using the OIALM (Orthogonal Inexact Augmented Lagrange Multiplier Method) algorithm. This includes singular value thresholding of low-rank terms, soft thresholding of sparse outliers, and adjusting the kernel norm weights of the modal matrix based on the modal weights using a weighted kernel norm. The resulting low-rank matrix is ​​given by the following formula: , in, To optimize the objective function, For modality The lower-rank term, For sparse regularization parameters, It is a sparse outlier. For modality The nuclear norm of the low-rank term.

[0036] By constructing an optimization objective and employing the OIALM algorithm for weighted low-rank decomposition, we can effectively suppress sparse outliers while maintaining the global structure of the data. This gives the low-rank matrix high stability and noise resistance. The iterative optimization framework of OIALM allows the low-rank terms and sparse representation to be updated synchronously, achieving global smoothing of the energy consumption sequence and repair of local anomalies. Guided by modal weights, the decomposition process can be made closer to the true contribution structure of building energy consumption, thus obtaining a low-rank matrix with higher interpretability. This matrix effectively removes invalid disturbances and has good robustness to local anomalies caused by extreme weather and abnormal equipment behavior, providing an extremely stable input foundation for the prediction model.

[0037] Furthermore, the initial predicted energy consumption is output through the prediction model, including: Based on a low-rank matrix, the initial predicted energy consumption is output through the prediction model, including performing mode folding operation on the low-rank matrix to obtain a stable time series, performing row right padding on the series to obtain the TCN input sequence, and inputting it into the TCN network to obtain a deep feature vector. The deep feature vector is input into the bidirectional GRU to obtain the hidden state. The importance score at each time point is obtained through the attention mechanism. The importance score is used to calculate the attention weight through the Softmax function. The importance score and the attention weight are weighted and summed to obtain the context vector. The context vector is input into the fully connected output layer to obtain the initial prediction energy consumption.

[0038] By utilizing a low-rank matrix as high-quality input and implementing initial energy consumption prediction through a multi-layer time series model, the prediction accuracy can be significantly improved. TCN can capture local patterns in long sequences, bidirectional GRU can learn contextual relationships, and the attention mechanism further identifies key time windows in energy consumption changes, thereby achieving differentiated modeling of energy consumption contribution at different stages. Initial predicted energy consumption is obtained through context vectors and fully connected layers, enabling the model to exhibit stronger generalization ability after multimodal deep fusion, and it has a stable ability to capture both peak energy consumption and low-frequency trends.

[0039] Furthermore, the model parameters are optimized using the Golden Jackal optimization algorithm to generate the final predicted energy consumption, including: Initialize the hyperparameter space to be optimized, including the kernel size, dilation rate and regularization coefficient of TCN, the number of hidden units and layers of bidirectional GRU, and the key vector dimension of the attention mechanism. Use the Golden Jackal optimization algorithm to iteratively update the hyperparameter space to be optimized (minimize the root mean square error) to obtain the optimal hyperparameter space and generate the final predicted energy consumption.

[0040] The Golden Jackal optimization algorithm is used to globally optimize the hyperparameters of the prediction model, so that the final prediction results maintain high accuracy under different environmental conditions. The algorithm has strong global search capabilities and a high ability to escape local optima. It can quickly find the optimal combination in complex hyperparameter space, so that the model structure is optimally configured. By minimizing the error to drive the optimization process, the final prediction results have lower bias and higher stability, providing strong engineering practical value for building energy consumption prediction.

[0041] In summary, this invention defines an optimization objective function based on modal weights, uses the OIALM algorithm for iterative updates to obtain a low-rank matrix, outputs initial predicted energy consumption through a prediction model, and uses the Golden Jackal optimization algorithm to optimize the model parameters to generate the final predicted energy consumption. It effectively integrates building physical semantics and dynamic environmental features, improves the accuracy and robustness of energy consumption feature expression, and enhances the modeling ability and energy consumption prediction accuracy for unsteady thermal processes.

[0042] It should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and are not intended to limit it. Although the present invention has been described in detail with reference to preferred embodiments, those skilled in the art should understand that modifications or equivalent substitutions can be made to the technical solutions of the present invention without departing from the spirit and scope of the technical solutions of the present invention, and all such modifications or substitutions should be covered within the scope of the claims of the present invention.

Claims

1. A BIM-based energy consumption prediction and monitoring system, characterized in that: include, The unified spatiotemporal matrix and dynamic energy consumption driving sequence construction module is used to obtain BIM models through API interfaces and real-time meteorological data through IoT platforms. The BIM models and IoT platforms are preprocessed to form a unified spatiotemporal matrix. Based on the unified spatiotemporal matrix, the transient heat flux energy of components, transient load of rooms and HVAC equivalent load are calculated to obtain the dynamic energy consumption driving sequence. The sparse suppression and smoothing tensor generation module is used to construct a three-modal feature tensor based on the structured energy consumption parameter matrix, extended environment sequence, and dynamic energy consumption driving sequence. Based on the three-modal feature tensor, the energy value is calculated for each feature dimension. The sparse suppression tensor is obtained by filtering. The time modes in the sparse suppression tensor are linearly interpolated to obtain the local gradient. The optimal node sequence is constructed. The time order is rearranged based on the optimal node sequence to obtain the rearranged sparse suppression tensor. Random perturbation is added to each time point to generate a smooth signal and obtain the smoothed sparse suppression tensor. The attention-enhanced feature reconstruction module is used to calculate the first-order gradient of the building physics feature mode in the three-modal feature tensor based on the smooth sparse suppression tensor, generate a column-uniform matrix, and perform SVD decomposition to obtain the factor matrix. The core tensor is obtained through Tucker inverse mapping. Row vectors are extracted based on the factor matrix and attention weights are calculated. A diagonal weighted matrix is ​​constructed to reconstruct the core tensor along the time direction. The final enhanced features are obtained through Tucker reconstruction, and modal weights are constructed. The low-rank optimization and energy consumption prediction output module is used to define the optimization objective function based on modal weights, use the OIALM algorithm for iterative updates to obtain the low-rank matrix, output the initial predicted energy consumption through the prediction model, and use the Golden Jackal optimization algorithm to optimize the model parameters to generate the final predicted energy consumption.

2. The BIM-based energy consumption prediction and monitoring system as described in claim 1, characterized in that: The process of rearranging the temporal order based on the optimal node sequence to obtain the rearranged sparse suppression tensor includes: The structured energy consumption parameter matrix, extended environment sequence, and dynamic energy consumption driving sequence are horizontally concatenated along the feature dimension to obtain a unified concatenation tensor. The feature dimensions of the unified concatenation tensor are then divided to obtain a three-modal feature tensor. The energy value of each feature dimension of the three-modal feature tensor is calculated. If the energy value is less than the energy value threshold, the corresponding feature dimension is deleted to obtain a sparse suppression tensor. The local gradient is calculated and sorted in descending order. The time point corresponding to the maximum local gradient is selected to obtain the optimal node sequence. The temporal order in the sparse suppression tensor is rearranged based on the optimal node sequence to obtain the rearranged sparse suppression tensor. Random perturbation is added to each time point in the rearranged sparse suppression tensor, and a smoothing signal is calculated. The smoothing signal is then used to replace the rows in the expanded matrix to obtain the smoothed sparse suppression tensor.

3. The BIM-based energy consumption prediction and monitoring system as described in claim 2, characterized in that: The generation of the column-consistent matrix and the subsequent SVD decomposition to obtain the factor matrix include: Based on the smooth sparse suppression tensor, the first-order gradient of the building physical feature modes is calculated to generate a column-uniform matrix. The uniform matrix is ​​then decomposed by SVD to obtain the time mode factor matrix. The three-mode tensor is expanded by mode-2 and mode-3 to obtain the building physical feature matrix and the environmental and operational feature matrix. The two matrices are then decomposed by SVD to obtain the building physical feature factor matrix and the environmental and operational feature factor matrix.

4. The BIM-based energy consumption prediction and monitoring system as described in claim 3, characterized in that: The constructed diagonal weighted matrix is ​​used to reconstruct the core tensor along the time direction, and the final enhanced features are obtained through Tucker reconstruction, including: The core tensor is obtained by performing Tucker inverse mapping on the time modality factor matrix, the building physics feature factor matrix, and the environmental and operational feature factor matrix. The row vectors corresponding to each time point are extracted from the temporal modality factor matrix, the attention weights are calculated, a diagonal weighted matrix is ​​constructed by constructing a diagonal matrix, the core tensor is reconstructed in the temporal direction to obtain the attention-enhanced core tensor, and the attention-enhanced core tensor is reconstructed by Tucker to obtain the final enhanced features.

5. The BIM-based energy consumption prediction and monitoring system as described in claim 4, characterized in that: The construction of modal weights includes: Based on the time mode, building physics mode, and environmental and operational mode in the final enhanced features, matrix expansion is performed to obtain the mode matrix. The energy intensity is calculated for each mode matrix, and the mode weights are constructed.

6. The BIM-based energy consumption prediction and monitoring system as described in claim 5, characterized in that: The defined objective function is optimized and iteratively updated using the OIALM algorithm to obtain a low-rank matrix, including: Based on modal weights, optimization is defined as the objective function. The modal matrix is ​​folded to obtain an initialization matrix. The initialization matrix is ​​then iteratively updated using the OIALM algorithm to obtain a low-rank matrix.

7. The BIM-based energy consumption prediction and monitoring system as described in claim 6, characterized in that: The step of outputting the initial predicted energy consumption through the prediction model includes: A mode folding operation is performed on the low-rank matrix to obtain a stable time series. Row right padding is then performed on the series to obtain the TCN input sequence. This sequence is input into the TCN network to obtain a deep feature vector. The deep feature vector is then input into a bidirectional GRU to obtain the hidden state. An importance score is obtained at each time point through an attention mechanism. The attention weights of the importance scores are calculated using the Softmax function. The importance scores and attention weights are then weighted and summed to obtain a context vector. This context vector is then input into the fully connected output layer to obtain the initial prediction energy consumption.

8. The BIM-based energy consumption prediction and monitoring system as described in claim 7, characterized in that: The process of optimizing model parameters using the Golden Jackal optimization algorithm to generate the final predicted energy consumption includes: Initialize the hyperparameter space to be optimized, including the kernel size, dilation rate and regularization coefficient of TCN, the number of hidden units and layers of bidirectional GRU, and the key vector dimension of the attention mechanism. Use the Golden Jackal optimization algorithm to iteratively update the hyperparameter space to be optimized to obtain the optimal hyperparameter space and generate the final predicted energy consumption.

9. The BIM-based energy consumption prediction and monitoring system as described in claim 8, characterized in that: The process involves obtaining a BIM model via an API interface and acquiring real-time meteorological data through an IoT platform, followed by preprocessing of both the BIM model and the IoT platform, including: By acquiring BIM models through API interfaces and real-time meteorological data through IoT platforms, the BIM models and IoT platforms are preprocessed to obtain structured energy consumption parameter matrices and extended environmental sequences.

10. The BIM-based energy consumption prediction and monitoring system as described in claim 9, characterized in that: The calculation of transient heat flux energy of the component, transient room load, and HVAC equivalent load yields a dynamic energy consumption drive sequence, including: The structured energy consumption parameter matrix and the extended environment sequence are concatenated to obtain an alignment matrix. All alignment matrices within time t are then sorted according to time. The dimensions are stacked to obtain a unified spatiotemporal matrix, and a dynamic energy consumption driving sequence is constructed. This includes calculating the transient heat flux energy of components, the transient load of the room, and the equivalent load of HVAC. Combined with the solar heat gain, the normalization process is performed and a vector horizontal splicing operation is carried out to obtain the dynamic energy consumption driving sequence.

Images