CN112418495A - Building energy consumption prediction method based on longicorn stigma optimization algorithm and neural network - Google Patents

Building energy consumption prediction method based on longicorn stigma optimization algorithm and neural network Download PDF

Info

Publication number
CN112418495A
CN112418495A CN202011246897.8A CN202011246897A CN112418495A CN 112418495 A CN112418495 A CN 112418495A CN 202011246897 A CN202011246897 A CN 202011246897A CN 112418495 A CN112418495 A CN 112418495A
Authority
CN
China
Prior art keywords
energy consumption
data
longicorn
neural network
building energy
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
Application number
CN202011246897.8A
Other languages
Chinese (zh)
Inventor
胡程磊
仲颖
孙璧文
姬丽雯
王宜雷
董育亮
Current Assignee (The listed assignees may be inaccurate. Google has not performed a legal analysis and makes no representation or warranty as to the accuracy of the list.)
Jiangsu Polytechnic College of Agriculture and Forestry
Original Assignee
Jiangsu Polytechnic College of Agriculture and Forestry
Priority date (The priority date 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 date listed.)
Filing date
Publication date
Application filed by Jiangsu Polytechnic College of Agriculture and Forestry filed Critical Jiangsu Polytechnic College of Agriculture and Forestry
Priority to CN202011246897.8A priority Critical patent/CN112418495A/en
Publication of CN112418495A publication Critical patent/CN112418495A/en
Pending legal-status Critical Current

Links

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
    • G06N3/004—Artificial life, i.e. computing arrangements simulating life
    • 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
    • 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
    • G06Q10/06—Resources, workflows, human or project management; Enterprise or organisation planning; Enterprise or organisation modelling
    • G06Q10/067—Enterprise or organisation modelling
    • 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
    • G06Q50/00—Systems or methods specially adapted for specific business sectors, e.g. utilities or tourism
    • G06Q50/06—Electricity, gas or water supply

Abstract

The invention discloses a building energy consumption prediction method based on a longicorn stigma optimization algorithm and a neural network, which comprises the following four main steps: 1) collecting building energy consumption related data, and performing principal component analysis and normalization pretreatment on the data; 2) determining input and output items and a network structure of a multi-layer feedforward neural network model with an error back propagation learning function; 3) optimizing the connection weight and the threshold of the BP network by using a longicorn algorithm; 4) and carrying out short-term prediction on the building electric power energy consumption by using the BAS-BP prediction model obtained by optimization. The method utilizes the principal component analysis algorithm to carry out principal component analysis on the pre-input variables, and selects the variables meeting the principal component extraction requirement, thereby reducing the input dimension; and optimizing the structure and parameters of the neural network model by using the global optimization capability of the longicorn algorithm. Compared with the existing building energy consumption prediction method, the prediction model provided by the invention has the advantages of simple structure, high prediction precision, short operation time and the like.

Description

Building energy consumption prediction method based on longicorn stigma optimization algorithm and neural network
Technical Field
The invention relates to a building energy consumption prediction method based on a longicorn stigma optimization algorithm and a neural network, and belongs to the field of building energy management.
Background
Building energy systems are a complex class of multivariable, distributed parameter systems. Accurate prediction of building energy consumption level is an important basis and precondition for analyzing the energy-saving potential of buildings and guiding future energy use. Meanwhile, the method has important practical significance for improving the service efficiency of building energy consumption equipment and reducing energy waste. In recent years, with the proposal and wide application of various intelligent optimization technologies, a prediction method of building energy consumption is rapidly developed. In the field of building energy, how to combine an advanced control means to improve and perfect the current building energy consumption prediction method has certain guiding significance for accurate estimation and scientific management of building energy consumption and formulation of relevant policy and regulation.
Over the years, researchers at home and abroad have extensively studied building energy consumption prediction methods, and two major analysis methods, namely a forward method (forward modeling method) and a reverse method (data-driven method), are formed, wherein the forward method usually needs detailed building parameters, energy equipment systems and meteorological data, so that a large amount of expert knowledge is needed, energy consumption calculation is time-consuming, and online operation analysis and control are not facilitated. In the inverse method, if the historical energy consumption data of a building is known, the BP neural network method becomes one of typical prediction methods due to strong self-learning and adaptive capabilities, but as a gradient-based adaptive algorithm, the learning process of the BP neural network has the defects of easy falling into local extremum, slow convergence speed and the like. The intelligent optimization algorithms such as longicorn beard and the like have good global search capability and good convergence, and are easy to combine with other algorithms, so that a chance is brought to improvement of a single neural network structure and optimization of parameters. In addition, because there are many relevant factors affecting building energy consumption, except for the external weather environment, holidays, geographical locations, and the structure of the building itself all affect the building energy consumption, and how to reasonably select the input items of the prediction model and determine the appropriate network structure is also a problem that the neural network prediction method needs further research.
Disclosure of Invention
Aiming at the defects of the existing building energy consumption prediction method, the invention provides the building energy consumption prediction method based on the longicorn stigma optimization algorithm and the neural network, and compared with a single neural network, the building energy consumption prediction method has higher prediction precision. The method is characterized in that the Tianniu whisker algorithm is used for optimizing the connection weight and the threshold of the BP neural network, the advantages of simple realization, high optimization speed, strong global search capability and the like of the Tianniu whisker algorithm are exerted, and the randomness defect existing in the selection problem of the connection weight and the threshold of the BP neural network is overcome, so that the BP neural network has strong convergence, and the learning capability and the generalization capability of the BP neural network are improved.
In the process of building energy consumption prediction model establishment, reasonable input variables are selected as an important link, and the method utilizes the PCA principal component analysis method to screen the pre-input variables, thereby reducing the input dimension, reducing the network scale and having important significance for the analysis and the pretreatment of the building energy consumption related data.
The technical scheme adopted by the invention is as follows:
a building energy consumption prediction method based on a longicorn stigma optimization algorithm and a neural network comprises the following steps:
(1) collecting building energy consumption related data and carrying out data preprocessing;
(2) carrying out variable screening and dimension reduction on the building energy consumption data by adopting a principal component analysis method;
(3) determining the structural parameters of the BP neural network;
(4) optimizing the connection weight and the threshold of each network layer of the BP neural network by using a longicorn algorithm;
(5) and carrying out short-term prediction on the building power energy consumption by using the BAS-BP neural network prediction model obtained by optimization.
In the step (1), the steps of collecting the building energy consumption related data and preprocessing the data are as follows:
(1) acquiring historical building energy consumption data monitored by an energy consumption monitoring system, and acquiring meteorological data information disclosed by a network;
(2) because the building energy consumption data has obvious week periodicity and day periodicity in a short term, namely the energy consumption of a holiday is lower than the energy consumption of a normal working day, the energy consumption of the holiday is less in the morning and at the evening every day, the energy consumption of the noon is less relative to the energy consumption of the morning and the afternoon, noise data are removed according to the periodicity, the removed noise data are filled up through an averaging method, and the noise data are represented by a formula:
Figure BDA0002770327700000031
here, the first and second liquid crystal display panels are,
Figure BDA0002770327700000032
for padded data estimation, yt-1Is the previous cycle data, yt+1Is the data of the next period; according to the local holiday information, a number 1 represents a normal working day, a number 0 represents a rest day, and a holiday mark s is obtained; the sine value sh and cosine value ch of each day time (hour scale) are calculated by the formula:
Figure BDA0002770327700000033
Figure BDA0002770327700000034
here, h (t) is an hour value; and carrying out normalization processing on the energy consumption data set, wherein the formula is as follows:
Figure BDA0002770327700000035
where x is the data to be normalized, xmin、xmaxThe minimum value and the maximum value of the data to be normalized are shown, and y represents a normalized output value.
In the step (2), the steps of screening and reducing the dimension of the building energy consumption data by using a principal component analysis method are as follows:
(1) and (5) standardizing data. Normalizing the input data set with n groups of data and m variables to make all data in the range of [ -1,1 ];
(2) analyzing the principal components, and regarding the data after the standardization processing as a matrix Yn×mAnd performing principal component analysis, wherein the formula is as follows:
Figure BDA0002770327700000041
wherein, c1,c2,...,cmIs Yn×mAnd the principal component vectors of (1) are orthogonal to each other two by two; t is t1,t1,...,tmIs a set of orthonormal vectors, Yn×mIs that it corresponds to the vector tiThe residual data information X is negligible;
C. calculating contribution rate, extracting data principal component, calculating characteristic vector tiCharacteristic value λ ofiAccording to λ1≥λ2≥...≥λmAre sequentially arranged, so that the corresponding characteristic vector t can be obtained1,t1,...,tmAbility to contain initial data information, according to variance contribution rate eta of each principal componentiThe first k principal components can be extracted, and the formula is as follows:
Figure BDA0002770327700000042
in the step (3), the input items of the BP neural network are determined through principal component analysis, namely variables meeting principal component extraction conditions are selected as the input items; the structure of the BP network comprises the number of input neurons, the number of output neurons and the number of hidden layer neurons of the network. The number of the neurons in the hidden layer is reasonably selected according to the existence theorem (Kolmogorov) of the mapping neural network, and the formula is as follows:
Figure BDA0002770327700000043
here, c is the number of hidden layer neurons, a is the number of input layer neurons, b is the number of output layer neurons, and t is an integer constant defined between [1,10 ].
In the step (4), the step of optimizing the connection weight and the threshold of each network layer of the BP neural network by using the longicorn algorithm is as follows:
(1) randomly initializing the orientation of the longicorn stigma and carrying out normalization treatment, wherein the formula is as follows:
Figure BDA0002770327700000051
here, rands () is a random function, k is the spatial dimension,
(2) and creating space coordinates of the left and right whiskers of the longicorn, wherein the formula is as follows:
Figure BDA0002770327700000052
here, xrRepresenting the position coordinates of the longicorn right hair at the t iteration; x is the number oflRepresenting the position coordinates of the longicorn left hair at the t iteration; x is the number oftRepresenting the barycentric coordinates of the longicorn at the t-th iteration; dtThe distance between the left and right whiskers is shown,
(3) determining the advancing direction of the longicorn according to the left and right fitness function values, and updating the centroid coordinate of the longicorn when the longicorn is iterated for t +1 times, wherein the fitness function formula is as follows:
Figure BDA0002770327700000053
here, ypred,iIndicates the predicted value of energy consumption, ydata,iRepresenting the actual energy consumption value, N representing the number of training data sets; centroid coordinate x of longicorn at t +1 iterationst+1The formula is as follows:
Figure BDA0002770327700000054
here, sine () is a sign function, δtRepresenting the step factor at the t-th iteration,
(4) updating the space coordinates of the left and right longicorn whiskers by adopting a linear decreasing weight strategy, wherein the distance formula between the left and right longicorn whiskers is as follows: dt+1＝0.95dtThe step factor formula is: deltat+1＝0.95δt；
(5) Judging whether the termination condition is met: if the current iteration times reach the maximum iteration times or the training error of the network reaches the precision requirement, stopping iteration and outputting an optimization result, otherwise, returning to B to continue iteration optimization;
(6) and outputting an optimization result, namely the optimized connection weight and threshold of the BP network.
In the step (5), the building energy consumption data obtained after principal component analysis is divided into a training data set and a testing data set, and an optimized BAS-BP prediction model is used for building energy consumption prediction, wherein the training data set is used for performing prediction model training, and the testing data set is used for measuring energy consumption prediction errors; in order to evaluate the prediction precision of the building energy consumption, a coefficient of variation CV of the root-mean-square difference and an average absolute percentage error MAPE are used as a measurement index, and the expression is as follows:
Figure BDA0002770327700000061
Figure BDA0002770327700000062
here, ypred,iRepresenting predicted energy consumption value, ydata,iRepresenting the actual energy consumption value in the test data set,
Figure BDA0002770327700000063
represents the average of the actual energy consumption values and N represents the number of test data sets.
The invention relates to a building energy consumption prediction method based on a longicorn whisker optimization algorithm and a neural network, which utilizes principal component analysis to decouple and reduce dimension of a pre-input variable, and combines the longicorn whisker optimization algorithm and the BP neural network to obtain the BAS-BP energy consumption prediction method which has the advantages of simple structure, high prediction precision and the like.
Compared with the prior energy consumption prediction method, the method has the advantages that:
1. the model has simple structure. According to the invention, the input dimensionality is reduced by a PCA principal component analysis method, and the defects of overlong BP neural network training time, overlarge network information redundancy and the like caused by huge training data are well solved;
2. the prediction precision is high. According to the invention, the connection weight and the threshold of the BP neural network are optimized by using a Bethes-taurus algorithm (BAS), so that the learning capability and the generalization capability of the BP neural network are improved, and the prediction precision of the building energy consumption is obviously improved;
3. the running time is short. The longicorn algorithm is used as an efficient intelligent optimization algorithm, only one search individual is needed, and compared with other swarm intelligent algorithms, the longicorn algorithm has the outstanding advantages of simple parameter setting, small operand and the like. The BAS-BP energy consumption prediction method provided by the invention ensures the prediction precision, greatly shortens the operation time and can meet the online energy consumption prediction requirement.
Drawings
FIG. 1 is a flow chart of a building energy consumption prediction method based on a longicorn whisker optimization algorithm and a neural network according to the invention;
FIG. 2 is a simulation diagram of energy consumption prediction for a single BP neural network;
FIG. 3 is a simulation diagram of energy consumption prediction after optimization of BP neural network by using Tianniu whisker algorithm.
Detailed Description
In order to describe the present invention more specifically, the present invention will be described in detail below with reference to the accompanying drawings and specific examples.
Fig. 1 is a flowchart of a building energy consumption prediction method based on a longicorn whisker optimization algorithm and a neural network according to the present invention.
The implementation steps of the method of the invention are described in detail below by using building energy consumption data and corresponding meteorological data provided by the american society of heating, refrigeration and air conditioning engineers (ASHRAE):
and step 0, collecting building energy consumption related data and preprocessing the data.
Step 0.1, obtaining building energy consumption data and corresponding meteorological data provided by a first building energy consumption prediction competition held by American ASHRAE, wherein the data types comprise: the outdoor average dry bulb temperature, the solar radiation value, the relative humidity, the wind speed, the electric power energy consumption, the hot water energy consumption and the cold water energy consumption of the whole building. Relevant meteorological data takes hour intervals, and 4208 groups of data are collected in the data set;
step 0.2, because the building energy consumption data has obvious week periodicity and day periodicity in a short term, namely the energy consumption of a rest day is lower than the energy consumption of a normal working day, the energy consumption in the morning and the evening of each day is less, the energy consumption in the noon is less relative to the energy consumption in the morning and the afternoon, noise data are removed according to the periodicity, the removed data are filled through an averaging method, and the noise data are represented by a formula:
Figure BDA0002770327700000081
here, the first and second liquid crystal display panels are,
Figure BDA0002770327700000082
for padded data estimation, yt-1Is the previous cycle data, yt+1Is the data of the next period; according to the local holiday information, a number 1 represents a normal working day, a number 0 represents a rest day, and a holiday mark s is obtained; calculating the sine of the time of day (in hours)The value sh and the cosine value ch are given by the formula:
Figure BDA0002770327700000083
Figure BDA0002770327700000084
here, h (t) is an hour value; and carrying out normalization processing on the energy consumption data set, wherein the formula is as follows:
Figure BDA0002770327700000091
where x is the data to be normalized, xmin、xmaxThe minimum value and the maximum value of the data to be normalized are shown, and y represents a normalized output value.
Step 1, screening and reducing dimension of building energy consumption data by using a principal component analysis method. The specific method comprises the following steps: performing PCA principal component analysis on the pre-input variables, comprising: after analysis, the cumulative contribution rate of two variables of the outdoor air dry bulb temperature T (t) and the solar radiation value S (t) reaches 97%, and the two variables can be considered to cover most of information in the original data, so that the dimension reduction can be performed on the original data.
And 2, determining the structural parameters of the BP neural network. The specific method comprises the following steps: PCA (principal component analysis) is utilized to select outdoor air dry bulb temperature T (t) and solar radiation value S (t) as two input variables, and then strong periodic regularity of building energy consumption data, local holiday marks and influences of historical energy consumption are combined to finally determine 7 variables such as outdoor air dry bulb temperature T (t), solar radiation value S (t), holiday marks s, sine value sh of each day time (hour level), cosine value ch of each day time (hour level), building power energy consumption value y (t-1) before one hour, building power energy consumption value y (t-2) before two hours and the like as the input variables; output, i.e. hours of the buildingGrade power demand wbe (white building electrical energy); hidden layer neurons of the BP neural network model according to the mapping neural network Presence theorem (Kolmogorov)
Figure BDA0002770327700000092
t is defined as [1,10]]An integer constant between 7 and 13, and determining the number of hidden layer neurons to be 9 in combination with practical application.
And 3, optimizing the connection weight and the threshold of the BP neural network by using a longicorn algorithm. The method comprises the following specific steps:
step 3.1, randomly initializing the orientation of the longicorn stigma and carrying out normalization treatment: in this example, the normalization processing formula is:
Figure BDA0002770327700000101
here, rands () is a random function, k is the spatial dimension; initializing parameters: in this example, the initial step length is set to 30, the initial distance between the left and right whiskers is 6, and the maximum iteration number is set to 100;
step 3.2, creating space coordinates of the longicorn left and right whiskers, wherein the formula is as follows:
Figure BDA0002770327700000102
here, xrRepresenting the position coordinates of the longicorn right hair at the t iteration; x is the number oflRepresenting the position coordinates of the longicorn left hair at the t iteration; x is the number oftRepresenting the barycentric coordinates of the longicorn at the t-th iteration; dtIndicating the distance between the two left and right whiskers.
Step 3.3, determining the advancing direction of the longhorn beetle according to the left and right fitness function values, and updating the barycentric coordinates of the longhorn beetle during t +1 iterations, wherein the fitness function formula is as follows:
Figure BDA0002770327700000103
here, ypred,iIndicates the predicted value of energy consumption, ydata,iRepresenting the actual energy consumption value, N represents the number of training data sets (3208 in this example); centroid coordinate x of longicorn at t +1 iterationst+1The formula is as follows:
Figure BDA0002770327700000104
here, sine () is a sign function, δtRepresenting the step factor at the t-th iteration.
And 3.4, updating the space coordinates of the left and right longicorn whiskers by adopting a linear decreasing weight strategy, wherein the distance formula between the left and right longicorn whiskers is as follows: dt+1＝0.95dtThe step factor formula is: deltat+1＝0.95δt。
Step 3.5, judging whether the termination condition is met: if the current iteration times reach the maximum iteration times or the training error of the network reaches the precision requirement, stopping iteration and outputting an optimization result, otherwise, returning to the step 3.2 to continue iteration optimization.
And 3.6, outputting an optimization result, and assigning a connection weight and a threshold value to the BP network.
And 4, carrying out short-term prediction on the building electric power energy consumption by using the BAS-BP neural network prediction model obtained through optimization. The building energy consumption data is divided into a training data set and a testing data set. The test data set is used for representing energy consumption prediction errors (in the example, the energy consumption data comprises 4208 groups of data, the front 3208 group of data is selected for network training, and the rear 1000 groups of data are used for network testing; the maximum iteration number of the BP neural network is set to be 100, and the learning rate is set to be 0.001); training and testing are respectively carried out on the BP network before and after optimization by using training and testing data, the prediction accuracy of the building energy consumption is evaluated by using a variation Coefficient (CV) of a root-mean-square difference and an average absolute percentage error (MAPE) as measurement indexes, and the expression is as follows:
Figure BDA0002770327700000111
Figure BDA0002770327700000112
here, ypred,iRepresenting predicted energy consumption value, ydata,iRepresenting the actual energy consumption value in the test data set,
Figure BDA0002770327700000113
representing the average of the actual energy consumption values and N representing the number of test data sets (in this case 1000).
In order to compare the building energy consumption prediction effect, a single BP neural network prediction model is selected as a reference. By applying the energy consumption prediction model provided by the invention, the variation coefficient of the root mean square difference of the short-term prediction of the building energy consumption can be improved by 25.7% at most and can be improved by 16.7% on average.
Fig. 2 and fig. 3 are comparison diagrams of the predicted building energy consumption value and the actual energy consumption value obtained by simulating the BP network before and after optimization. Fig. 2 depicts a simulation diagram of energy consumption prediction of a single BP neural network, and fig. 3 depicts a simulation diagram of energy consumption prediction after optimization of the BP neural network using the longicorn whisker algorithm. As can be seen from the figure, the invention optimizes the connection weight and the threshold of the BP neural network by using a Tianniu whiskers algorithm (BAS), can obviously improve the structure of the BP network and improve the prediction precision of the building energy consumption.
While the invention has been described with reference to specific embodiments, it will be apparent to those skilled in the art that various changes and modifications can be made therein without departing from the spirit and scope thereof. Therefore, various modifications and changes that fall within the scope of the claims of the present invention should fall within the scope of the present invention.

Claims (6)

1. A building energy consumption prediction method based on a longicorn stigma optimization algorithm and a neural network is characterized by comprising the following steps: the method comprises the following steps:
(1) collecting building energy consumption related data and carrying out data preprocessing;
(2) carrying out variable screening and dimension reduction on the building energy consumption data by adopting a principal component analysis method;
(3) determining the structural parameters of the BP neural network;
(4) optimizing the connection weight and the threshold of each network layer of the BP neural network by using a longicorn algorithm;
(5) and carrying out short-term prediction on the building power energy consumption by using the BAS-BP neural network prediction model obtained by optimization.
2. The building energy consumption prediction method based on the longicorn whisker optimization algorithm and the neural network as claimed in claim 1, wherein: in the step (1), the building energy consumption related data comprises historical building power energy consumption data and local building meteorological data, and the preprocessing of the building energy consumption data comprises the following steps: noisy data culling and data normalization, wherein:
the eliminated noise data is filled by an averaging method, and is expressed by a formula as follows:
Figure FDA0002770327690000011
here, the first and second liquid crystal display panels are,
Figure FDA0002770327690000013
for padded data estimation, yt-1Is the previous cycle data, yt+1Is the data of the next period;
the normalization of the energy consumption data is formulated as:
Figure FDA0002770327690000012
where x is the data to be normalized, xmin、xmaxThe minimum value and the maximum value of the data to be normalized are shown, and y represents a normalized output value.
3. The building energy consumption prediction method based on the longicorn whisker optimization algorithm and the neural network as claimed in claim 1, wherein: in the step (2), the steps of screening and reducing the dimension of the building energy consumption data by using a principal component analysis method are as follows:
(1) normalizing data, namely normalizing the input data set with n groups of data and m variables to enable all data to be in a range of [ -1,1 ];
(2) analyzing principal components, and regarding the normalized data as a matrix Yn×mAnd performing principal component analysis, wherein the formula is as follows:
Figure FDA0002770327690000021
wherein, c1,c2,...,cmIs Yn×mAnd the principal component vectors of (1) are orthogonal to each other two by two; t is t1,t1,...,tmIs a set of orthonormal vectors, Yn×mIs that it corresponds to the vector tiThe residual data information X is negligible;
(3) calculating contribution rate, extracting data principal component, calculating characteristic vector tiCharacteristic value λ ofiAccording to λ1≥λ2≥...≥λmAre sequentially arranged, so that the corresponding characteristic vector t can be obtained1,t1,...,tmAbility to contain initial data information, according to variance contribution rate eta of each principal componentiThe first k principal components can be extracted, and the formula is as follows:
Figure FDA0002770327690000022
4. the building energy consumption prediction method based on the longicorn whisker optimization algorithm and the neural network as claimed in claim 1, wherein: in the step (3), the structure of the BP network includes the number of input neurons, the number of output neurons, and the number of neurons in the hidden layer of the network, wherein the number of input neurons is determined by data principal component information extracted after principal component analysis; the number of the neurons of the hidden layer is reasonably selected according to the existence theorem of the mapping neural network, and the formula is as follows:
Figure FDA0002770327690000023
here, c is the number of hidden layer neurons, a is the number of input layer neurons, b is the number of output layer neurons, and t is an integer constant defined between [1,10 ].
5. The building energy consumption prediction method based on the longicorn whisker optimization algorithm and the neural network as claimed in claim 1, wherein: in the step (4), the step of optimizing the connection weight and the threshold of the BP neural network by using the longicorn whisker algorithm is as follows:
(1) randomly initializing the orientation of the longicorn stigma and carrying out normalization treatment, wherein the formula is as follows:
Figure FDA0002770327690000031
here, rands () is a random function, k is the spatial dimension;
(2) creating space coordinates of the longicorn left and right whiskers, wherein the formula is as follows:
Figure FDA0002770327690000032
here, xrRepresenting the position coordinates of the longicorn right hair at the t iteration; x is the number oflRepresenting the position coordinates of the longicorn left hair at the t iteration; x is the number oftRepresenting the barycentric coordinates of the longicorn at the t-th iteration; dtRepresenting the distance between the left and right whiskers;
(3) determining the advancing direction of the longicorn according to the left and right fitness function values, and updating the centroid coordinate of the longicorn when the longicorn is iterated for t +1 times, wherein the fitness function formula is as follows:
Figure FDA0002770327690000033
here, ypred,iIndicates the predicted value of energy consumption, ydata,iRepresenting the actual energy consumption value, N representing the number of training data sets;
centroid coordinate x of longicorn at t +1 iterationst+1The formula is as follows:
Figure FDA0002770327690000034
here, sine () is a sign function, δtRepresents the step factor at the t-th iteration;
(4) updating the space coordinates of the left and right whiskers of the longicorn by adopting a linear decreasing weight strategy, wherein the distance formula between the left and right whiskers is as follows: dt+1＝0.95dtThe step factor formula is: deltat+1＝0.95δt。
(5) Judging whether a termination condition is met: if the current iteration times reach the maximum iteration times or the training error of the network reaches the precision requirement, stopping iteration and outputting an optimization result, otherwise, returning to the step (2) to continue iteration optimization;
(6) and outputting an optimization result, namely the optimized connection weight and threshold of the BP network.
6. The building energy consumption prediction method based on the longicorn whisker optimization algorithm and the neural network as claimed in claim 1, wherein: in the step (5), the building energy consumption data obtained through principal component analysis is divided into a training data set and a testing data set, and an optimized BAS-BP prediction model is used for building energy consumption prediction, wherein the training data set is used for performing prediction model training, and the testing data set is used for measuring energy consumption prediction errors; in order to evaluate the prediction precision of the building energy consumption, a variation coefficient CV of a root-mean-square difference and an average absolute percentage error MAPE are used as measurement indexes, and the expression is as follows:
Figure FDA0002770327690000041
Figure FDA0002770327690000042
here, ypred,iRepresenting predicted energy consumption value, ydata,iRepresenting the actual energy consumption value in the test data set,
Figure FDA0002770327690000043
represents the average of the actual energy consumption values and N represents the number of test data sets.
CN202011246897.8A 2020-11-10 2020-11-10 Building energy consumption prediction method based on longicorn stigma optimization algorithm and neural network Pending CN112418495A (en)

Priority Applications (1)

Application Number Priority Date Filing Date Title
CN202011246897.8A CN112418495A (en) 2020-11-10 2020-11-10 Building energy consumption prediction method based on longicorn stigma optimization algorithm and neural network

Applications Claiming Priority (1)

Application Number Priority Date Filing Date Title
CN202011246897.8A CN112418495A (en) 2020-11-10 2020-11-10 Building energy consumption prediction method based on longicorn stigma optimization algorithm and neural network

Publications (1)

Publication Number Publication Date
CN112418495A true CN112418495A (en) 2021-02-26

Family

ID=74782325

Family Applications (1)

Application Number Title Priority Date Filing Date
CN202011246897.8A Pending CN112418495A (en) 2020-11-10 2020-11-10 Building energy consumption prediction method based on longicorn stigma optimization algorithm and neural network

Country Status (1)

Country Link
CN (1) CN112418495A (en)

Cited By (3)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN113052379A (en) * 2021-03-17 2021-06-29 河北工程大学 Tomato water demand prediction method based on BAS-RBF
CN113177675A (en) * 2021-05-28 2021-07-27 西安建筑科技大学 Air conditioner cold load prediction method based on optimization neural network of longicorn group algorithm
CN113705891A (en) * 2021-08-27 2021-11-26 南京林业大学 Urban business office complex building parking demand prediction method based on MRA-BAS-BP algorithm

Citations (4)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN104484715A (en) * 2014-11-28 2015-04-01 江苏大学 Neural network and particle swarm optimization algorithm-based building energy consumption predicting method
CN109492748A (en) * 2018-09-26 2019-03-19 广东工业大学 A kind of Mid-long term load forecasting method for establishing model of the electric system based on convolutional neural networks
CN110728401A (en) * 2019-10-10 2020-01-24 郑州轻工业学院 Short-term power load prediction method of neural network based on squirrel and weed hybrid algorithm
CN110852902A (en) * 2019-11-14 2020-02-28 合肥工业大学 Photovoltaic power generation power prediction method based on BAS-BP

Patent Citations (4)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN104484715A (en) * 2014-11-28 2015-04-01 江苏大学 Neural network and particle swarm optimization algorithm-based building energy consumption predicting method
CN109492748A (en) * 2018-09-26 2019-03-19 广东工业大学 A kind of Mid-long term load forecasting method for establishing model of the electric system based on convolutional neural networks
CN110728401A (en) * 2019-10-10 2020-01-24 郑州轻工业学院 Short-term power load prediction method of neural network based on squirrel and weed hybrid algorithm
CN110852902A (en) * 2019-11-14 2020-02-28 合肥工业大学 Photovoltaic power generation power prediction method based on BAS-BP

Non-Patent Citations (1)

* Cited by examiner, † Cited by third party
Title
胡程磊: "数据驱动的建筑电能耗预测方法研究", 《中国优秀硕士学位论文全文数据库工程科技Ⅱ辑》 *

Cited By (5)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN113052379A (en) * 2021-03-17 2021-06-29 河北工程大学 Tomato water demand prediction method based on BAS-RBF
CN113177675A (en) * 2021-05-28 2021-07-27 西安建筑科技大学 Air conditioner cold load prediction method based on optimization neural network of longicorn group algorithm
CN113177675B (en) * 2021-05-28 2023-05-23 西安建筑科技大学 Air conditioner cooling load prediction method based on longicorn group algorithm optimization neural network
CN113705891A (en) * 2021-08-27 2021-11-26 南京林业大学 Urban business office complex building parking demand prediction method based on MRA-BAS-BP algorithm
CN113705891B (en) * 2021-08-27 2024-02-13 南京林业大学 Urban commercial complex building parking demand prediction method based on MRA-BAS-BP algorithm

Similar Documents

Publication Publication Date Title
Zhang et al. A systematic feature selection procedure for short-term data-driven building energy forecasting model development
Khafaga et al. Forecasting Energy Consumption Using a Novel Hybrid Dipper Throated Optimization and Stochastic Fractal Search Algorithm
CN112418495A (en) Building energy consumption prediction method based on longicorn stigma optimization algorithm and neural network
CN110705743B (en) New energy consumption electric quantity prediction method based on long-term and short-term memory neural network
CN104484715A (en) Neural network and particle swarm optimization algorithm-based building energy consumption predicting method
CN107590562A (en) A kind of Short-Term Load Forecasting of Electric Power System based on changeable weight combination predicted method
Delcroix et al. Autoregressive neural networks with exogenous variables for indoor temperature prediction in buildings
CN111260126B (en) Short-term photovoltaic power generation prediction method considering correlation degree of weather and meteorological factors
CN111598224B (en) Air conditioner cold load prediction method for optimizing BP neural network based on simulated annealing algorithm
CN105719028A (en) Method for dynamic prediction of air-conditioning loads based on multi-factor chaos support vector machine
CN113657660A (en) Heat source load prediction method based on substation load and heat supply network hysteresis model
CN104331737A (en) Office building load prediction method based on particle swarm neural network
CN113837488B (en) Method, system and equipment for predicting energy consumption data
CN112149890A (en) Comprehensive energy load prediction method and system based on user energy label
CN112365056A (en) Electrical load joint prediction method and device, terminal and storage medium
CN116796141A (en) GBDT regression model-based office building energy consumption prediction method
CN116307076A (en) Industrial park energy efficiency management and control method based on Internet of things
CN111815039A (en) Weekly scale wind power probability prediction method and system based on weather classification
CN116227689A (en) Photovoltaic panel pollution degree state evaluation and cleaning decision system
Girimurugan et al. Application of Deep Learning to the Prediction of Solar Irradiance through Missing Data
CN114119273A (en) Park comprehensive energy system non-invasive load decomposition method and system
CN114240687A (en) Energy hosting efficiency analysis method suitable for comprehensive energy system
CN114091776A (en) K-means-based multi-branch AGCNN short-term power load prediction method
Kumar et al. Forecasting indoor temperature for smart buildings with ARIMA, SARIMAX, and LSTM: A fusion approach
CN113570414A (en) Electricity price prediction method for optimizing deep neural network based on improved Adam algorithm

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