CN-121998322-A - Building energy consumption prediction method, system, equipment and medium

CN121998322ACN 121998322 ACN121998322 ACN 121998322ACN-121998322-A

Abstract

The application relates to the technical field of building energy consumption prediction, in particular to a method, a system, equipment and a medium for predicting building energy consumption, which comprise the steps of calculating a pearson correlation coefficient according to sensor information in an initial layer; the method comprises the steps of carrying out feature aggregation through a tensor graph rolling network, taking an average value under the condition of carrying out global average pooling treatment on a plurality of fusion feature tensor information, splicing intra-layer local coding feature vector information and initial global attribute information, carrying out global weight treatment according to floor space information, carrying out attention score calculation on global node feature information through a graph attention network, adjusting first global side weight information according to attention parameters, carrying out aggregation on global node feature information and second global side weight information, carrying out coding operation on space fusion feature sequence information through a time sequence model, carrying out building energy consumption prediction on high-dimensional features of the space-time fusion feature vector information through a fully connected network, and carrying out building energy consumption prediction accuracy of a single building.

Inventors

BIAN JI
WANG DEPEI
DONG CHUNXIAO
LIU HUI

Assignees

深圳市中宏低碳建筑科技有限公司

Dates

Publication Date: 20260508
Application Date: 20260114

Claims (10)

1. A building energy consumption prediction method, comprising: acquiring sensor information in an initial layer, and calculating the Pearson correlation coefficient of every two sensors in the layer according to the sensor information in the initial layer to obtain inner layer edge weight information; acquiring initial layer attribute information, and performing feature aggregation on initial layer sensor information, initial layer attribute information and layer inner side weight information through a tensor graph convolution network to obtain fusion feature tensor information; Taking an average value under the condition of carrying out global average pooling treatment on a plurality of fusion feature tensor information to obtain intra-layer local coding feature vector information; Acquiring initial global attribute information, and splicing intra-layer local coding feature vector information and the initial global attribute information to obtain global node feature information; The method comprises the steps of acquiring floor spacing information, carrying out global weight processing according to the floor spacing information to obtain first global side weight information, carrying out attention score calculation on global node characteristic information through a graph attention network to obtain attention parameters, and adjusting the first global side weight information according to the attention parameters to obtain second global side weight information; The method comprises the steps of acquiring global node characteristic information and second global side weight information, and acquiring space fusion characteristic sequence information; and reducing the dimension of the high-dimension features of the space-time fusion feature vector information through the fully connected network to obtain the predicted value of the building energy consumption at the target moment.
2. The building energy consumption prediction method according to claim 1, further comprising: And obtaining a true energy consumption value, and calculating the loss between the predicted value of the building energy consumption at the target moment and the true energy consumption value according to the average absolute error to obtain an energy consumption loss value for parameter updating.
3. The building energy consumption prediction method according to claim 1, further comprising: and acquiring historical data information, and preprocessing the historical data information to obtain initial intra-layer sensor information, initial intra-layer attribute information and initial global attribute information.
4. The building energy consumption prediction method according to claim 1, wherein the specific step of aggregating the global node feature information and the second global side weight information to obtain the spatial fusion feature sequence information is: acquiring time step information, and sequencing global node characteristic information according to the time step information to obtain global fusion characteristic sequence information; And splicing the global fusion characteristic sequence information with the corresponding time step information to obtain the spatial fusion characteristic sequence information.
5. The building energy consumption prediction method according to claim 1, wherein the specific steps of obtaining the initial intra-layer attribute information, and performing feature aggregation on the initial intra-layer sensor information, the initial intra-layer attribute information and the intra-layer weight information through a tensor convolution network to obtain the fused feature tensor information are as follows: acquiring initial layer attribute information, and splicing initial layer sensor information and initial layer attribute information to obtain initial node characteristic information; combining the initial node characteristic information and the intra-layer edge weight information into intra-layer edge weight data association information; And carrying out space association feature aggregation on the intra-layer weighted data association information through a tensor convolution network to obtain fusion feature tensor information.
6. The building energy consumption prediction method according to claim 1, wherein the specific steps of obtaining inter-floor distance information, performing global weight processing according to the inter-floor distance information, and obtaining first global side weight information are as follows: Acquiring floor space information, and performing reciprocal calculation on the floor space information to obtain the reciprocal information of the floor space; and acquiring a weight fixed parameter, and performing global weight processing according to the weight fixed parameter and the floor space reciprocal information to obtain first global side weight information.
7. The building energy consumption prediction method according to claim 1, wherein the step of reducing the dimension of the high-dimension feature of the space-time fusion feature vector information to obtain the predicted value of the building energy consumption at the target moment through the fully connected network comprises the following specific steps: linearly mapping the space-time fusion feature vector information through a first layer of fully-connected network to obtain a first dimension feature; Performing linear transformation and compression on the first dimension characteristic through a second layer of full-connection network to obtain a second dimension characteristic; And linearly regressing the second dimension characteristic through a third-layer full-connection network to obtain the predicted value of the building energy consumption at the target moment.
8. A building energy consumption prediction system, comprising: The intra-layer weight analysis module is used for acquiring initial intra-layer sensor information, calculating the pearson correlation coefficient of every two sensors in the layer according to the initial intra-layer sensor information, and acquiring intra-layer weight information; The intra-layer feature fusion module is used for acquiring initial intra-layer attribute information, and carrying out feature aggregation on the initial intra-layer sensor information, the initial intra-layer attribute information and the intra-layer weight information through a tensor convolution network to obtain fusion feature tensor information; the global average pooling processing module is used for taking the average value under the condition of carrying out global average pooling processing on the plurality of fusion characteristic tensor information to obtain intra-layer local coding characteristic vector information; The node characteristic splicing module is used for acquiring initial global attribute information, splicing intra-layer local coding characteristic vector information and the initial global attribute information to obtain global node characteristic information; The global side weight analysis module is used for acquiring floor space information, performing reciprocal calculation on the floor space information to obtain first global side weight information, performing attention score calculation on global node characteristic information through a graph attention network to obtain attention parameters, and adjusting the first global side weight information according to the attention parameters to obtain second global side weight information; The space-time fusion feature processing module is used for aggregating the global node feature information and the second global side weight information to obtain space fusion feature sequence information; And the energy consumption prediction module is used for reducing the dimension of the high-dimension feature of the space-time fusion feature vector information through the fully connected network to obtain the building energy consumption predicted value at the target moment.
9. An electronic device comprising a memory storing a computer program and a processor arranged to run the computer program to perform the building energy consumption prediction method of any of claims 1-7.
10. A computer readable storage medium, characterized in that a computer program is stored in the computer readable storage medium, wherein the computer program is arranged to perform the building energy consumption prediction method of any of claims 1-7 at run-time.

Description

Building energy consumption prediction method, system, equipment and medium Technical Field The application relates to the technical field of building energy consumption prediction, in particular to a building energy consumption prediction method, a system, equipment and a medium. Background With the continuous advancement of urban process, buildings are used as main carriers of energy consumption, and energy consumption efficiency management is increasingly paid attention to. The building inner space is three-dimensional, the function subregion is various, and each floor equipment operation and personnel's activity mutually independent influence each other, leads to its whole energy consumption to demonstrate dynamic, inhomogeneous complex characteristic. The current building relies on sensor networks deployed in various subareas to monitor energy consumption, and the core is to collect and collect meter data of different subareas such as water, electricity, gas and the like in real time, essentially complete the recording and statistics of historical and current energy consumption states, generate various data reports and provide basis for energy consumption audit and post analysis. However, this statistical-based monitoring mode is limited in that it is mainly oriented towards the past and the present, and it is difficult to form an effective prognosis for the future. Due to the lack of mining and modeling data, the system cannot only rely on current and historical statistics, so that the system is passive in operation management, the equipment operation strategy or starting demand response cannot be adjusted in advance before the electricity consumption peak comes, the energy purchasing plan or the charging and discharging time sequence of the energy storage system cannot be optimized according to the predicted load curve, the precursor of the potential energy efficiency bottleneck is difficult to accurately locate, a manager can usually take countermeasures only after abnormality occurs or peak occurs, the energy saving scheduling is delayed, the optimization efficiency is low, the operation cost and the risk are correspondingly increased, and the problems are to be solved. Disclosure of Invention In order to overcome the statistical limitation of the existing monitoring technology, realize the effective prediction of the energy consumption of the single multi-storey building and improve the accuracy of the energy consumption prediction of the single building, the application provides a building energy consumption prediction method, system, equipment and medium, which adopts the following technical scheme: in a first aspect, the present application provides a building energy consumption prediction method, including: acquiring sensor information in an initial layer, and calculating the Pearson correlation coefficient of every two sensors in the layer according to the sensor information in the initial layer to obtain inner layer edge weight information; acquiring initial layer attribute information, and performing feature aggregation on initial layer sensor information, initial layer attribute information and layer inner side weight information through a tensor graph convolution network to obtain fusion feature tensor information; Taking an average value under the condition of carrying out global average pooling treatment on a plurality of fusion feature tensor information to obtain intra-layer local coding feature vector information; Acquiring initial global attribute information, and splicing intra-layer local coding feature vector information and the initial global attribute information to obtain global node feature information; The method comprises the steps of acquiring floor spacing information, carrying out global weight processing according to the floor spacing information to obtain first global side weight information, carrying out attention score calculation on global node characteristic information through a graph attention network to obtain attention parameters, and adjusting the first global side weight information according to the attention parameters to obtain second global side weight information; The method comprises the steps of acquiring global node characteristic information and second global side weight information, and acquiring space fusion characteristic sequence information; and reducing the dimension of the high-dimension features of the space-time fusion feature vector information through the fully connected network to obtain the predicted value of the building energy consumption at the target moment. Preferably, the method further comprises: And obtaining a true energy consumption value, and calculating the loss between the predicted value of the building energy consumption at the target moment and the true energy consumption value according to the average absolute error to obtain an energy consumption loss value for parameter updating. Preferably, the method further comprises