CN113537638A - Short-term wind pressure prediction method and abnormal data completion method and device for high-rise building - Google Patents
Short-term wind pressure prediction method and abnormal data completion method and device for high-rise building Download PDFInfo
- Publication number
- CN113537638A CN113537638A CN202110927781.9A CN202110927781A CN113537638A CN 113537638 A CN113537638 A CN 113537638A CN 202110927781 A CN202110927781 A CN 202110927781A CN 113537638 A CN113537638 A CN 113537638A
- Authority
- CN
- China
- Prior art keywords
- wind pressure
- pressure data
- data
- neural network
- network model
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Pending
Links
- 230000002159 abnormal effect Effects 0.000 title claims abstract description 55
- 238000000034 method Methods 0.000 title claims abstract description 51
- 238000003062 neural network model Methods 0.000 claims abstract description 69
- 238000000354 decomposition reaction Methods 0.000 claims abstract description 37
- 238000012549 training Methods 0.000 claims abstract description 30
- 230000008569 process Effects 0.000 claims description 11
- 230000004913 activation Effects 0.000 claims description 7
- 238000004364 calculation method Methods 0.000 claims description 5
- 238000012216 screening Methods 0.000 claims description 3
- 230000000644 propagated effect Effects 0.000 claims description 2
- 238000013528 artificial neural network Methods 0.000 abstract 1
- 230000006870 function Effects 0.000 description 18
- 238000010586 diagram Methods 0.000 description 6
- 238000012545 processing Methods 0.000 description 5
- 238000013473 artificial intelligence Methods 0.000 description 1
- 230000001174 ascending effect Effects 0.000 description 1
- 230000005540 biological transmission Effects 0.000 description 1
- 230000000295 complement effect Effects 0.000 description 1
- 230000000694 effects Effects 0.000 description 1
- 230000007774 longterm Effects 0.000 description 1
- 238000010801 machine learning Methods 0.000 description 1
- 239000011159 matrix material Substances 0.000 description 1
- 238000012986 modification Methods 0.000 description 1
- 230000004048 modification Effects 0.000 description 1
- 238000012544 monitoring process Methods 0.000 description 1
- 238000012546 transfer Methods 0.000 description 1
- 238000012800 visualization Methods 0.000 description 1
Images
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06Q—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES; SYSTEMS OR METHODS SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES, NOT OTHERWISE PROVIDED FOR
- G06Q10/00—Administration; Management
- G06Q10/04—Forecasting or optimisation specially adapted for administrative or management purposes, e.g. linear programming or "cutting stock problem"
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/04—Architecture, e.g. interconnection topology
- G06N3/045—Combinations of networks
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/08—Learning methods
- G06N3/084—Backpropagation, e.g. using gradient descent
Landscapes
- Engineering & Computer Science (AREA)
- Theoretical Computer Science (AREA)
- Physics & Mathematics (AREA)
- General Physics & Mathematics (AREA)
- Business, Economics & Management (AREA)
- Software Systems (AREA)
- Health & Medical Sciences (AREA)
- Computational Linguistics (AREA)
- Data Mining & Analysis (AREA)
- Evolutionary Computation (AREA)
- General Health & Medical Sciences (AREA)
- Molecular Biology (AREA)
- Computing Systems (AREA)
- General Engineering & Computer Science (AREA)
- Biomedical Technology (AREA)
- Mathematical Physics (AREA)
- Artificial Intelligence (AREA)
- Life Sciences & Earth Sciences (AREA)
- Biophysics (AREA)
- Strategic Management (AREA)
- Economics (AREA)
- Human Resources & Organizations (AREA)
- Game Theory and Decision Science (AREA)
- Development Economics (AREA)
- Entrepreneurship & Innovation (AREA)
- Marketing (AREA)
- Operations Research (AREA)
- Quality & Reliability (AREA)
- Tourism & Hospitality (AREA)
- General Business, Economics & Management (AREA)
- Testing Of Devices, Machine Parts, Or Other Structures Thereof (AREA)
- Wind Motors (AREA)
Abstract
本发明公开了一种高层建筑风压短期预测方法、异常数据补全方法及装置，包括采集高层建筑的历史风压数据，并采用经验模态分解算法将所述原始风压数据序列分解成多个IMF分量和一个RES分量；将分解得到的IMF分量输入BiGRU神经网络模型中进行迭代训练，优化模型参数，得到训练好的BiGRU神经网络模型；采集高层建筑的实时风压数据，将实时风压数据输入训练好的BiGRU神经网络模型中，通过BiGRU神经网络模型得到高层建筑的风压预测数据；本发明基于经验模态分解算法与BiGRU模型实现，将分解后的IMF分量输入到BiGRU深度神经网络模型中，风压预测数据的预测精度高，能够准确评估高层建筑的表面风压。
The invention discloses a method for short-term prediction of wind pressure of high-rise buildings, a method and device for completing abnormal data, including collecting historical wind pressure data of high-rise buildings, and using an empirical mode decomposition algorithm to decompose the original wind pressure data sequence into multiple IMF components and a RES component; input the decomposed IMF components into the BiGRU neural network model for iterative training, optimize the model parameters, and obtain a trained BiGRU neural network model; collect real-time wind pressure data of high-rise buildings, The data is input into the trained BiGRU neural network model, and the wind pressure prediction data of the high-rise building is obtained through the BiGRU neural network model; the present invention is realized based on the empirical mode decomposition algorithm and the BiGRU model, and the decomposed IMF components are input into the BiGRU deep neural network In the model, the wind pressure prediction data has high prediction accuracy and can accurately evaluate the surface wind pressure of high-rise buildings.
Description
Technical Field
The invention relates to the technical field of building structure wind pressure prediction, in particular to a high-rise building wind pressure short-term wind pressure prediction method, an abnormal data completion method and an abnormal data completion device.
Background
The wind pressure sensors have important significance for evaluating the surface pressure of the building, but some wind pressure sensors fail or are abnormal in the long-term operation process, so that data are lost, and the data are difficult to recover.
At present, the method for predicting the missing wind pressure data of the high-rise building based on an artificial intelligence method is mainly divided into two methods: one method is a 'shallow layer' machine learning method, the wind pressure data has high nonlinearity and non-stationarity, and a 'shallow layer' model has certain limitation on short-term prediction of wind load, cannot process massive monitoring data and has low accuracy; the other method is a traditional deep neural network model and has the characteristics of universality, high efficiency and the like, but the accuracy needs to be further improved. Therefore, it is necessary to further improve the precision of the traditional deep neural network model and develop the accurate and real-time high-rise building wind pressure prediction work.
Disclosure of Invention
In view of the above, the present invention provides a short-term prediction method based on high-rise building wind pressure, an abnormal data completion method and an abnormal data completion device, so as to solve the problem of low accuracy in prediction and completion of high-rise building wind pressure data in the prior art.
In order to achieve the above object, an aspect of the present invention provides a short-term wind pressure prediction method for a high-rise building, which specifically includes the following steps:
s1: collecting historical wind pressure data of a high-rise building, and integrating the historical wind pressure data into an original wind pressure data sequence according to a time sequence;
s2: decomposing the original wind pressure data sequence into a plurality of IMF components and an RES component by adopting an empirical mode decomposition algorithm;
s3: inputting the IMF component obtained by decomposition into a BiGRU neural network model for iterative training, and optimizing model parameters to obtain a trained BiGRU neural network model;
s4: and acquiring real-time wind pressure data of the high-rise building, inputting the real-time wind pressure data into the trained BiGRU neural network model, and obtaining wind pressure prediction data of the high-rise building through the BiGRU neural network model.
Further, in step S2, the historical wind pressure data is integrated into an original wind pressure data sequence according to a time sequence, and the original wind pressure data sequence is subjected to empirical mode decomposition, which includes the following specific steps:
s201: screening out all maximum value points and minimum value points in the original wind pressure data sequence, and respectively fitting the maximum value points and the minimum value points to obtain an upper envelope line and a lower envelope line of the original wind pressure data sequence;
s202: calculating the mean envelope of the original wind pressure data sequence according to the upper envelope line and the lower envelope line, and calculating the difference value between the original wind pressure data sequence and the mean envelope to obtain a first wind pressure data sequence;
s203: judging whether the first wind pressure data sequence meets the condition that the IMF component is established, if so, executing the step S204, and if not, repeatedly executing the steps S201-S202 on the first wind pressure data sequence until the ith wind pressure data sequence obtained after repeating the m times meets the condition that the IMF component is established;
s204: taking the ith wind pressure data sequence as a first IMF component of the original wind pressure data sequence, and separating the first IMF component from the original wind pressure data sequence to obtain a first residual wind pressure data sequence;
s205: and judging whether the first residual wind pressure data sequence is a monotonic function, if so, completing the decomposition, otherwise, repeating the steps S201-S204 on the first residual wind pressure data sequence until the jth residual wind pressure data sequence obtained by repeating the n times is the monotonic function, so as to obtain n IMF components and one RES component by decomposition.
Further, in step S203, the condition that the IMF component is satisfied is:
in the whole time interval, the number of the maximum value points and the minimum value points on the first wind pressure data sequence is equal to or different from the number of zero crossing points by one at most;
and in the whole time course, the mean value of the upper envelope line and the lower envelope line at any point on the first wind pressure data sequence is zero.
Further, in step S3, the BiGRU neural network model includes two GRU submodels with the same structure, the two GRU submodels process the input IMF component along the time positive sequence and the time negative sequence, and the training process includes:
s301: inputting all IMF components obtained by decomposition in the step S2 into two GRU submodels respectively, and calculating reset gates and update gates of the two GRU submodels respectively;
s302: respectively calculating candidate activation states of the two GRU submodels, and calculating according to the candidate activation states, the reset gate and the update gate to obtain hidden layer output of the corresponding GRU submodels;
s303: fusing the hidden layer outputs of the two GRU submodels obtained by calculation to obtain a predicted value of the BiGRU neural network model;
s304: and reversely transmitting the predicted value of the BiGRU neural network model obtained in the step S303 to the BiGRU neural network model for iterative training, and finishing training until the loss function of the BiGRU neural network model tends to be stable to obtain the trained BiGRU neural network model.
Further, in step S303, the predicted value H of the BiGRU neural network modeltThe calculation formula of (2) is as follows:
wherein:
hidden layer transmission for forward propagation of GRU submodel at time tDischarging;
outputting a hidden layer which is propagated backwards for another GRU sub-model at the time t; alpha is alphat,βtWeights output by the hidden layers of forward propagation and backward propagation of the two GRU submodels at the time t are respectively; btAnd the offset corresponding to the hidden layer state of the BiGRU neural network model at the time t.
Further, in step S304, the loss function is a mean square error function, and its expression is:
wherein: x is the number oftIMF component, y, of the BiGRU neural network model is input for time ttAnd N is the iteration times during model training, and is wind pressure prediction data of the high-rise building to be predicted at the moment t.
The invention also provides a high-rise building wind pressure short-term abnormal data completion method, which comprises the steps of predicting real-time wind pressure data by adopting the high-rise building wind pressure short-term prediction method to obtain wind pressure prediction data;
and judging whether the real-time wind pressure data is abnormal or missing, and when the real-time wind pressure data is abnormal or missing, adopting the wind pressure prediction data to complete.
The third aspect of the present invention further provides a high-rise building wind pressure short-term abnormal data completion device, including:
the data acquisition module is used for acquiring the wind pressure data of the high-rise building;
the data decomposition module is used for carrying out empirical mode decomposition on the wind pressure data acquired by the data acquisition module to obtain a plurality of IMF components and an RES component;
the model training module is used for inputting all IMF components obtained by decomposition of the data decomposition module into the BiGRU neural network model for iterative training to obtain a trained BiGRU neural network model;
the prediction data output module is used for outputting wind pressure prediction data; and
and the abnormal data completion module is used for judging whether the real-time wind pressure data is abnormal or missing and completing the real-time wind pressure data by using the wind pressure prediction data when the real-time wind pressure data is abnormal or missing.
And the display module is used for displaying the wind pressure prediction data output by the prediction data output module and the abnormal data judged by the abnormal data completion module.
The method is realized based on an empirical mode decomposition algorithm and a BiGRU model, and because the empirical mode decomposition algorithm can decompose complex nonlinear wind pressure data into linear combinations of IMF components with limited frequencies from high to low, and each decomposed IMF component comprises local characteristic signals of different time scales of original wind pressure data, the decomposed IMF components are input into the BiGRU deep neural network model, so that time dimension characteristics can be extracted, and the aim of improving prediction accuracy is fulfilled. In addition, abnormal data or missing data collected by the wind pressure sensor can be supplemented through predicted wind pressure prediction data obtained through prediction, and the surface wind pressure of the high-rise building can be accurately evaluated.
Additional advantages, objects, and features of the invention will be set forth in part in the description which follows and in part will become apparent to those having ordinary skill in the art upon examination of the following or may be learned from practice of the invention. The objectives and other advantages of the invention may be realized and attained by the means of the instrumentalities and combinations particularly pointed out hereinafter.
Drawings
Fig. 1 is a flowchart of a short-term wind pressure prediction method for a high-rise building according to an embodiment of the present invention.
Fig. 2 is a flowchart of step S2.
Fig. 3 is a schematic structural diagram of the BiGRU neural network model in step S3.
Fig. 4 is a flowchart of step S3.
Fig. 5 is a schematic structural diagram of the GRU submodel in step S3.
Fig. 6 is a flowchart of a short-term abnormal wind pressure data completion method for a high-rise building according to a second embodiment of the present invention.
Fig. 7 is a schematic diagram illustrating an embodiment of the abnormal data completion process in fig. 6.
Fig. 8 is a system block diagram of a high-rise building wind pressure short-term abnormal data completion apparatus according to a third embodiment of the present invention.
Detailed Description
The following is further detailed by way of specific embodiments:
example 1
Fig. 1 is a flowchart of a short-term wind pressure prediction method for a high-rise building according to this embodiment. The short-term wind pressure prediction method for the high-rise building specifically comprises the following steps:
s1: and (6) data acquisition.
And acquiring historical wind pressure data of a high-rise building by using a wind pressure sensor, and integrating the historical wind pressure data into an original wind pressure data sequence according to a time sequence. The original wind pressure data sequence corresponding to the wind pressure sensor can be expressed as:
X(t)＝{Xt-k,Xt-k+1,…,Xt-1} (1)
wherein: xt-1And k is the length of the original wind pressure data sequence.
S2: and (5) decomposing the data.
Specifically, an Empirical Mode Decomposition (EMD) algorithm is used to decompose the original wind pressure data sequence into a plurality of IMF components and an RES component. The EMD algorithm substantially smoothes the non-stationary signal, and in this embodiment, because the wind pressure data of the high-rise building is random, the obtained original wind pressure data sequence shows a non-stationary ascending or descending trend, so that the EMD algorithm can be used to gradually decompose the fluctuations and trends of different scales in the original wind pressure data sequence to generate a series of data sequences with different feature scales, each data sequence being an IMF component.
As shown in fig. 2, the specific step of performing empirical mode decomposition on the original wind pressure data sequence in step S2 is as follows:
s201: and calculating an upper envelope line and a lower envelope line of the original wind pressure data sequence.
Firstly, all maximum points max [ X (t) ] in the original wind pressure data sequence X (t) are screened out]All maximum points max [ X (t) ] are scaled by cubic spline function]Fitting an upper envelope m of the original wind pressure data sequence X (t)1(t)max(ii) a Then, screening all minimum value points min [ X (t) ] in the original wind pressure data sequence X (t)]All minimum values min [ X (t) ] are calculated using cubic spline function]Fitting a lower envelope m of the original wind pressure data sequence X (t)1(t)min。
S202: and calculating the mean envelope to obtain a first wind pressure data sequence.
According to the upper envelope m1(t)maxAnd a lower envelope m1(t)minCalculating the mean value envelope m of the original wind pressure data sequence X (t)1(t)：
Mean envelope m obtained based on the above equation (2)1(t) calculating the original wind pressure data sequence X (t) and the mean value envelope m1(t) obtaining a first wind pressure data sequence d by the difference value between the first wind pressure data sequence d and the second wind pressure data sequence d1(t)：
d1(t)＝X(t)-m1(t) (3)
S203: and judging the condition of the first wind pressure data sequence.
Determining the first wind pressure data sequence d obtained in step S2021(t) whether two conditions are satisfied for the IMF component to hold:
in the whole time course, the first wind pressure data sequence d1(t) upper maximum value point max [ d ]1(t)]And minimum value min [ d ]1(t)]The number of the zero-crossing points is equal to or different from the number of the zero-crossing points by one at most;
in the whole time course, the first wind pressure numberAccording to sequence d1(t) an upper envelope m at any point on the line2(t)maxAnd a lower envelope m2(t)minMean envelope m of2(t)＝0。
If the first wind pressure data sequence d1(t) if the condition that the IMF component is satisfied, the process proceeds to step S204.
If the first wind pressure data sequence d1(t) if the IMF component is not satisfied, the first wind pressure data sequence d is processed1(t) repeating the steps S201-S202 as another original wind pressure data sequence until the ith wind pressure data sequence d obtained after repeating the m timesi(t) (where i is 1,2, …, m, m is the number of times steps S201-S202 are performed, i.e. d1(t) is a wind pressure data sequence obtained after the steps S201-S202 are executed once, dmAnd (t) is a wind pressure data sequence obtained after executing the steps S201-S202 m times) until the condition that the IMF component is satisfied.
S204: and separating IMF components, and calculating a residual wind pressure data sequence of the original wind pressure data sequence.
The wind pressure data sequence d meeting the IMF component establishment condition obtained in the step S203 is processedi(t) as the first IMF component, denoted IMF1(t) and IMF the first IMF component1(t) separating the original wind pressure data sequence X (t) to obtain a first residual wind pressure data sequence r1(t)：
r1(t)＝X(t)-IMF 1(t) (5)
S205: and judging the condition of the residual wind pressure data sequence.
According to the convergence condition of the EMD algorithm, judging the first residual wind pressure data sequence r obtained by decomposition in the step S2041(t) is a monotonic function.
If yes, the decomposition of the original wind pressure data sequence X (t) is completed, otherwise the residual wind pressure data sequence r is used1(t) repeating the steps S2031-S2034 as a new original wind pressure data sequence until repeating the jth remaining wind pressure data sequence r obtained n timesj(t) (where j is 1,2, …, n, n is the number of times steps S201-S204 are performed, i.e., n isr1(t) is a residual wind pressure data sequence obtained after the steps S201-S204 are executed once, rn(t) is a residual wind pressure data sequence obtained after the steps S201-S204 are executed for n times) is a monotone function. According to the convergence condition of EMD algorithm, when the decomposed residual wind pressure data sequence is a monotonic function, the time period is longer than the recording length of the original wind pressure data sequence X (t), so that the last residual wind pressure data sequence r obtained by decomposition can be usedn(t) is used as a trend term of the original wind pressure data sequence x (t), i.e. RES component, and further the original wind pressure data sequence x (t) can be expressed as:
n IMF components IMF obtained by decompositioni(t)＝{IMF1(t),IMF2(t),…,IMFn(t) } is to predict in the BiGRU neural network model of step S3 described below.
S3: and (5) training a model.
In the prediction of the wind pressure data, the wind pressure data at a certain moment is related to the previous moment and a certain future moment at the same time, the influence factors of the past wind pressure data and the influence factors of the future wind pressure data are related to the prediction of the wind pressure data at the current moment, the wind pressure prediction data can be more accurate, and the BiGRU neural network model can adaptively sense the characteristic information of the upper time sequence and the lower time sequenceiAnd (t) inputting the BiGRU neural network model to extract the dependency characteristics of the wind pressure data before and after time, predicting to obtain the wind pressure prediction data at the current moment, reversely inputting the wind pressure prediction data into the BiGRU neural network model to perform iterative training, and optimizing model parameters to obtain the trained BiGRU neural network model.
Specifically, as shown in fig. 3, the BiGRU neural network model is a schematic structural diagram of the BiGRU neural network model, the BiGRU neural network model is formed by combining two GRU submodels with the same structure, the two GRU submodels are respectively marked as a forward GRU submodel and a backward GRU submodel, the forward GRU submodel processes the IMF component input to the BiGRU neural network model in a time positive sequence, and the backward GRU submodel is used for processing the IMF component input to the BiGRU neural network model in a time negative sequence. The BiGRU neural network model comprises an input layer, a hidden layer and an output layer, wherein the input layer is used for inputting IMF components, the output layer is used for outputting wind pressure prediction data, and the output of the hidden layer can be obtained by weighted summation of the forward hidden layer output of a forward GRU sub-model and the backward hidden layer output of a backward GRU sub-model. Therefore, the processing of the input IMF component by the BiGRU neural network model can also be regarded as the processing of the IMF component by the two GRU submodels respectively.
As shown in fig. 4, the specific process of training the BiGRU neural network model in step S3 is as follows:
s301: a reset gate and an update gate of the GRU submodel are computed.
Specifically, as shown in fig. 5, all the IMF components IMF decomposed in step S2 are IMFi(t) inputting the forward GRU submodel and the backward GRU submodel, and respectively calculating to obtain a reset gate r corresponding to the forward GRU submodel and the backward GRU submodel at the time t according to the hidden layer output of the forward GRU submodel and the backward GRU submodel at the time t-1t：
And respectively calculating to obtain the updating gate z corresponding to the forward GRU submodel and the backward GRU submodel at the time tt：
Wherein: wr,Wz,Ur,UzA model parameter matrix to be trained; sigma is a Sigmoid activation function; x is the number oftInputting IMF components of the GRU submodels for the time t;
hidden layer outputs of the forward GRU submodel and the backward GRU submodel at the time t-1 are respectively output; → denotes processing in time order, and ← denotes processing in time order.
S302: and calculating hidden layer output of the GRU submodel.
Inputting IMF components of the forward GRU submodel and the backward GRU submodel according to the time t, and calculating in the step S301 to obtain a reset gate r at the time ttAnd hidden layer output of the forward GRU submodel and the backward GRU submodel at the time t-1, and candidate activation states corresponding to the forward GRU submodel and the backward GRU submodel at the time t are respectively obtained through calculation
Then according to the candidate activation state at the time t
Reset gate rtAnd an update gate ztCalculating to obtain hidden layer output h of the corresponding forward GRU submodel and the backward GRU submodel at the time tt：
Wherein: w and U are model parameter matrixes to be trained; tan h is a hyperbolic tangent function;
representing the product of the matrices.
S303: and calculating wind pressure prediction data.
The forward hidden layer output of the forward GRU sub-model calculated in the step S302 at the time t
And backward hidden layer output of backward GRU submodel at t moment
Carrying out weighted summation, and calculating to obtain a predicted value H of the BiGRU neural network modelt：
Wherein: alpha is alphatWeight, β, of the forward hidden layer output for the forward GRU submodel at time ttThe weight is output to a backward hidden layer of the GRU sub-model at the time t; btAnd the offset corresponding to the hidden layer state of the BiGRU neural network model at the time t.
The predicted value H of the BiGRU neural network modeltNamely the wind pressure prediction data yt。
S304: and (5) performing iterative training.
Wind pressure prediction data y of the BiGRU neural network model obtained in the step S303tReversely transmitting the signal to a BiGRU neural network model for iterative training until the loss function of the BiGRU neural network model tends to be stable, and finishing the training to obtain the signalAnd obtaining the trained BiGRU neural network model. In this embodiment, the loss function is a mean square error function, and its expression is:
wherein: x is the number oftIMF component, y, of the BiGRU neural network model is input for time ttAnd N is the iteration times during model training, and is wind pressure prediction data of the high-rise building to be predicted at the moment t.
S4: and predicting wind pressure data.
And acquiring real-time wind pressure data of the high-rise building, inputting the real-time wind pressure data into the trained BiGRU neural network model, and obtaining wind pressure prediction data of the high-rise building through the BiGRU neural network model.
In the embodiment, the EMD algorithm is adopted to gradually decompose the fluctuation and the trend of different scales in the original wind pressure data sequence to generate a series of data sequences with different characteristic scales, the data sequences are IMF components obtained through decomposition, then the IMF components are used as the input of a BiGRU neural network model to carry out model training, finally, the wind pressure data is predicted through the BiGRU neural network model, and the BiGRU neural network model can link the past wind pressure data influence factors and the future wind pressure data influence factors with the prediction of the wind pressure data at the current moment in the prediction of the wind pressure data, so that the wind pressure prediction data is more accurate.
Example two
As shown in fig. 6, a flowchart of a high-rise building wind pressure short-term abnormal data completion method according to this embodiment is shown, and the high-rise building wind pressure short-term abnormal data completion method according to this embodiment is implemented based on the high-rise building wind pressure short-term prediction method according to the first embodiment. The high-rise building wind pressure short-term abnormal data completion method of the embodiment includes steps S1 'to S4' corresponding to the method of steps S1 to S4 in the embodiment, in addition, the high-rise building wind pressure short-term abnormal data completion method of the embodiment further includes step S5 'to complete abnormal data after step S4', and the specific process of the embodiment is as follows:
as shown in fig. 7, the method of steps S1 '-S4' is adopted to predict the real-time wind pressure data, so as to obtain wind pressure prediction data.
S5': and (5) completing abnormal data.
Specifically, whether the wind pressure data acquired by the wind pressure sensor in real time is abnormal or missing is judged, and when the real-time wind pressure data is abnormal or missing, the abnormal data or the missing data is completed by adopting the wind pressure prediction data at the corresponding moment.
In the embodiment, the wind pressure prediction data obtained by prediction is adopted to complement the abnormal data and the missing data of the surface of the high-rise building to be detected within a certain period of time, so that the high precision is achieved, and the surface wind pressure of the high-rise building can be accurately evaluated.
EXAMPLE III
Fig. 8 is a system block diagram of the high-rise building wind pressure short-term abnormal data completion apparatus according to the present embodiment, and is used to implement the prediction method according to the first embodiment and the abnormal data completion method according to the second embodiment. The high-rise building wind pressure short-term abnormal data completion device of the embodiment comprises a data acquisition module 100, a data decomposition module 200, a model training module 300, a prediction data output module 400, an abnormal data completion module 500 and a display model 600, so that the short-term prediction of the wind pressure of the high-rise building and the completion of the wind pressure abnormal data of the surface of the high-rise building within a certain period of time are realized.
The data acquisition module 100 is used for acquiring wind pressure data of a high-rise building to be tested, the wind pressure data comprises historical wind pressure data and real-time wind pressure data, and the data acquisition module 100 is also used for transmitting the historical wind pressure data to the data decomposition module and transmitting the real-time wind pressure data to the prediction data output module 400. In this embodiment, the data acquisition module 100 is preferably a wind pressure sensor disposed on an outer facade of a high-rise building.
The data decomposition module 200 is configured to perform empirical mode decomposition on the wind pressure data acquired by the data acquisition module 100 to obtain a plurality of IMF components and an RES component, and input all the IMF components into the model training module 300 for model training.
The model training module 300 is configured to predict the wind pressure data by inputting all IMF components into the data decomposition module 200 as inputs of the BiGRU neural network model, and reversely transfer the predicted wind pressure predicted data to the BiGRU neural network model for iterative training until a loss function of the BiGRU neural network model becomes stable, that is, stop training to obtain the trained BiGRU neural network model. In this embodiment, the loss function is a mean square error function.
The prediction data output module 400 is configured to input the real-time wind pressure data acquired by the data acquisition module 100 into the trained BiGRU neural network model for prediction, and output the wind pressure prediction data to the abnormal data completion module 500 and the display module.
The abnormal data completion module 500 is configured to determine whether the real-time wind pressure data is abnormal or missing, and when the real-time wind pressure data is abnormal or missing, perform completion on the abnormal data or the missing data by using the wind pressure prediction data at the corresponding moment.
The display module 600 is configured to display the wind pressure prediction data output by the prediction data output module 400 and the abnormal data determined by the abnormal data completion module 500, so as to realize visualization of the wind pressure prediction data and the abnormal data.
In this embodiment, the data decomposition module 200, the model training module 300, the predicted data output module 400, the abnormal data completion module 500, and the display module 600 may all be integrated on a computer, so that the apparatus has a simple structure and a low cost.
The foregoing is merely an example of the present invention and common general knowledge of known specific structures and features of the embodiments is not described herein in any greater detail. It should be noted that, for those skilled in the art, without departing from the structure of the present invention, several changes and modifications can be made, which should also be regarded as the protection scope of the present invention, and these will not affect the effect of the implementation of the present invention and the practicability of the present invention.
Claims (9)
1. A short-term prediction method for wind pressure of a high-rise building is characterized by comprising the following steps:
s1: collecting historical wind pressure data of a high-rise building, and integrating the historical wind pressure data into an original wind pressure data sequence according to a time sequence;
s2: decomposing the original wind pressure data sequence into a plurality of IMF components and an RES component by adopting an empirical mode decomposition algorithm;
s3: inputting the IMF component obtained by decomposition into a BiGRU neural network model for iterative training, and optimizing model parameters to obtain a trained BiGRU neural network model;
s4: and acquiring real-time wind pressure data of the high-rise building, inputting the real-time wind pressure data into the trained BiGRU neural network model, and obtaining wind pressure prediction data of the high-rise building through the BiGRU neural network model.
2. The short-term prediction method for wind pressure of high-rise building according to claim 1, wherein in step S2, the empirical mode decomposition of the original wind pressure data sequence comprises:
s201: screening out all maximum value points and minimum value points in the original wind pressure data sequence, and respectively fitting the maximum value points and the minimum value points to obtain an upper envelope line and a lower envelope line of the original wind pressure data sequence;
s202: calculating the mean envelope of the original wind pressure data sequence according to the upper envelope line and the lower envelope line, and calculating the difference value between the original wind pressure data sequence and the mean envelope to obtain a first wind pressure data sequence;
s203: judging whether the first wind pressure data sequence meets the condition that the IMF component is established, if so, executing the step S204, and if not, repeatedly executing the steps S201-S202 on the first wind pressure data sequence until the ith wind pressure data sequence obtained after repeating the m times meets the condition that the IMF component is established;
s204: taking the ith wind pressure data sequence as a first IMF component of the original wind pressure data sequence, and separating the first IMF component from the original wind pressure data sequence to obtain a first residual wind pressure data sequence;
s205: and judging whether the first residual wind pressure data sequence is a monotonic function, if so, completing the decomposition, otherwise, repeating the steps S201-S204 on the first residual wind pressure data sequence until the jth residual wind pressure data sequence obtained by repeating the n times is the monotonic function, so as to obtain n IMF components and one RES component by decomposition.
3. The short-term prediction method for wind pressure of high-rise building according to claim 2, wherein in step S203, the condition that the IMF component is established is:
in the whole time interval, the number of the maximum value points and the minimum value points on the first wind pressure data sequence is equal to or different from the number of zero crossing points by one at most;
and in the whole time course, the mean value of the upper envelope line and the lower envelope line at any point on the first wind pressure data sequence is zero.
4. The short-term prediction method for wind pressure in high-rise buildings according to claim 1, wherein in step S3, the BiGRU neural network model includes two GRU submodels with the same structure, and the two GRU submodels process the input IMF components in the time forward order and the time reverse order respectively, and the training process is as follows:
s301: inputting all IMF components obtained by decomposition in the step S2 into two GRU submodels respectively, and calculating reset gates and update gates of the two GRU submodels respectively;
s302: respectively calculating candidate activation states of the two GRU submodels, and calculating according to the candidate activation states, the reset gate and the update gate to obtain hidden layer output of the corresponding GRU submodels;
s303: fusing the hidden layer outputs of the two GRU submodels obtained by calculation to obtain a predicted value of the BiGRU neural network model;
s304: and reversely transmitting the predicted value of the BiGRU neural network model obtained in the step S303 to the BiGRU neural network model for iterative training, and finishing training until the loss function of the BiGRU neural network model tends to be stable to obtain the trained BiGRU neural network model.
5. The short-term prediction method for wind pressure of high-rise building according to claim 4, wherein in step S303, the predicted value H of the BiGRU neural network modeltThe calculation formula of (2) is as follows:
wherein:
outputting a hidden layer for forward propagation of the GRU sub-model at the time t;
outputting a hidden layer which is propagated backwards for another GRU sub-model at the time t; alpha is alphat,βtWeights output by the hidden layers of forward propagation and backward propagation of the two GRU submodels at the time t are respectively; btAnd the offset corresponding to the hidden layer state of the BiGRU neural network model at the time t.
6. The short-term prediction method for wind pressure of high-rise building according to claim 4, wherein in step S304, the loss function is a mean square error function, and the expression is:
wherein: x is the number oftIMF component, y, of the BiGRU neural network model is input for time ttAnd N is the iteration times during model training, and is wind pressure prediction data of the high-rise building to be predicted at the moment t.
7. A method for complementing short-term abnormal wind pressure data of high-rise buildings is characterized by comprising
Predicting real-time wind pressure data by adopting a high-rise building wind pressure short-term prediction method according to any one of claims 1 to 6 to obtain wind pressure prediction data;
and judging whether the real-time wind pressure data is abnormal or missing, and when the real-time wind pressure data is abnormal or missing, adopting the wind pressure prediction data to complete.
8. The utility model provides a short-term abnormal data completion device of high-rise building wind pressure which characterized in that includes:
the data acquisition module is used for acquiring the wind pressure data of the high-rise building;
the data decomposition module is used for carrying out empirical mode decomposition on the wind pressure data acquired by the data acquisition module to obtain a plurality of IMF components and an RES component;
the model training module is used for inputting all IMF components obtained by decomposition of the data decomposition module into the BiGRU neural network model for iterative training to obtain a trained BiGRU neural network model;
the prediction data output module is used for outputting wind pressure prediction data; and
and the abnormal data completion module is used for judging whether the real-time wind pressure data is abnormal or missing and completing the real-time wind pressure data by using the wind pressure prediction data when the real-time wind pressure data is abnormal or missing.
9. The high-rise building wind pressure short-term abnormal data completion device according to claim 8, further comprising a display module for displaying the wind pressure prediction data output by the prediction data output module and the abnormal data judged by the abnormal data completion module.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202110927781.9A CN113537638A (en) | 2021-08-13 | 2021-08-13 | Short-term wind pressure prediction method and abnormal data completion method and device for high-rise building |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202110927781.9A CN113537638A (en) | 2021-08-13 | 2021-08-13 | Short-term wind pressure prediction method and abnormal data completion method and device for high-rise building |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN113537638A true CN113537638A (en) | 2021-10-22 |
Family
ID=78091676
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202110927781.9A Pending CN113537638A (en) | 2021-08-13 | 2021-08-13 | Short-term wind pressure prediction method and abnormal data completion method and device for high-rise building |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN113537638A (en) |
Cited By (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN114548161A (en) * | 2022-02-15 | 2022-05-27 | 山西理工智联科技有限公司 | Dense medium separation clean coal ash content prediction method and device, electronic equipment and medium |
| CN114912365A (en) * | 2022-06-02 | 2022-08-16 | 重庆大学 | Wind pressure short-term prediction method and device based on parallel CNN-GRU and computer equipment |
Citations (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN102368230A (en) * | 2011-10-31 | 2012-03-07 | 北京天地融科技有限公司 | Mobile memory and access control method thereof as well as system |
| CN106126896A (en) * | 2016-06-20 | 2016-11-16 | 中国地质大学（武汉） | The mixed model wind speed forecasting method learnt based on empirical mode decomposition and the degree of depth and system |
| CN111860982A (en) * | 2020-07-06 | 2020-10-30 | 东北大学 | A short-term wind power prediction method for wind farms based on VMD-FCM-GRU |
| CN112884237A (en) * | 2021-03-11 | 2021-06-01 | 山东科技大学 | Power distribution network prediction auxiliary state estimation method and system |
-
2021
- 2021-08-13 CN CN202110927781.9A patent/CN113537638A/en active Pending
Patent Citations (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN102368230A (en) * | 2011-10-31 | 2012-03-07 | 北京天地融科技有限公司 | Mobile memory and access control method thereof as well as system |
| CN106126896A (en) * | 2016-06-20 | 2016-11-16 | 中国地质大学（武汉） | The mixed model wind speed forecasting method learnt based on empirical mode decomposition and the degree of depth and system |
| CN111860982A (en) * | 2020-07-06 | 2020-10-30 | 东北大学 | A short-term wind power prediction method for wind farms based on VMD-FCM-GRU |
| CN112884237A (en) * | 2021-03-11 | 2021-06-01 | 山东科技大学 | Power distribution network prediction auxiliary state estimation method and system |
Non-Patent Citations (2)
| Title |
|---|
| 李映雪: ""基于递归神经网络的空气质量指数预测"", 《万方学位论文》 * |
| 黄佳星: ""基于机器学习方法的风荷载预测研究"", 《中国优秀硕士学位论文全文数据库 工程科技II辑》 * |
Cited By (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN114548161A (en) * | 2022-02-15 | 2022-05-27 | 山西理工智联科技有限公司 | Dense medium separation clean coal ash content prediction method and device, electronic equipment and medium |
| CN114912365A (en) * | 2022-06-02 | 2022-08-16 | 重庆大学 | Wind pressure short-term prediction method and device based on parallel CNN-GRU and computer equipment |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN107590565A (en) | A kind of method and device for building building energy consumption forecast model | |
| CN114358192B (en) | Multi-source heterogeneous landslide data monitoring and fusing method | |
| CN111985719B (en) | Power load prediction method based on improved long-term and short-term memory network | |
| CN112381279B (en) | Wind power prediction method based on VMD and BLS combined model | |
| CN112001110A (en) | Structural damage identification monitoring method based on vibration signal space real-time recursive graph convolutional neural network | |
| CN106529185B (en) | A combined prediction method and system for displacement of ancient buildings | |
| CN113537638A (en) | Short-term wind pressure prediction method and abnormal data completion method and device for high-rise building | |
| CN113159088B (en) | Fault monitoring and diagnosis method based on multi-feature fusion and width learning | |
| CN113688770B (en) | Method and device for supplementing long-term wind pressure missing data of high-rise building | |
| RU2600099C1 (en) | Method of neural network forecasting of change of values of function with its complementary wavelet processing and device for its implementation | |
| CN111784061A (en) | Training method, device and equipment for power grid engineering cost prediction model | |
| CN111914470A (en) | Multi-monitoring time series regression prediction method for energy chemical production system | |
| CN113988210A (en) | Distorted data restoration method, device and storage medium for structural monitoring sensor network | |
| CN116303786B (en) | Block chain financial big data management system based on multidimensional data fusion algorithm | |
| CN114897277B (en) | Burst-type landslide displacement prediction method based on LSTM | |
| CN105447596A (en) | Optimized power data short-term prediction method | |
| Lola et al. | Improving the performance of ann-arima models for predicting water quality in the offshore area of kuala terengganu, terengganu, malaysia | |
| CN111104298A (en) | LSTM-based power grid server running state prediction device | |
| CN117217548A (en) | Water quality prediction method based on CEEMDAN-LSTM | |
| Ballı et al. | An application of artificial neural networks for prediction and comparison with statistical methods | |
| CN111695260B (en) | Material performance prediction method and system | |
| CN117493980B (en) | Bearing fault diagnosis method integrating feature extraction and sequencing causal discovery | |
| CN118153433A (en) | Event trigger model learning method based on physical information neural network | |
| CN117371321A (en) | Internal plasticity depth echo state network soft measurement modeling method based on Bayesian optimization | |
| CN113779724B (en) | A method and system for intelligent prediction of faults of filling and packaging machine |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| RJ01 | Rejection of invention patent application after publication | ||
| RJ01 | Rejection of invention patent application after publication |
Application publication date: 20211022 |