CN116187565A - Office building energy consumption prediction method based on improved atomic search optimization BP network - Google Patents

Office building energy consumption prediction method based on improved atomic search optimization BP network Download PDF

Info

Publication number
CN116187565A
CN116187565A CN202310176307.6A CN202310176307A CN116187565A CN 116187565 A CN116187565 A CN 116187565A CN 202310176307 A CN202310176307 A CN 202310176307A CN 116187565 A CN116187565 A CN 116187565A
Authority
CN
China
Prior art keywords
atoms
atom
energy consumption
office building
atomic
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
CN202310176307.6A
Other languages
Chinese (zh)
Inventor
黄云云
张博瑞
肖逢华
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.)
Fuzhou University
Original Assignee
Fuzhou University
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 Fuzhou University filed Critical Fuzhou University
Priority to CN202310176307.6A priority Critical patent/CN116187565A/en
Publication of CN116187565A publication Critical patent/CN116187565A/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
    • G06N3/02—Neural networks
    • G06N3/04—Architecture, e.g. interconnection topology
    • 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
    • G06N3/0418—Architecture, e.g. interconnection topology using chaos or fractal principles
    • 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
    • G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
    • G06N3/00—Computing arrangements based on biological models
    • G06N3/12—Computing arrangements based on biological models using genetic models
    • G06N3/126—Evolutionary algorithms, e.g. genetic algorithms or genetic programming
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
    • G06N7/00—Computing arrangements based on specific mathematical models
    • G06N7/08—Computing arrangements based on specific mathematical models using chaos models or non-linear system models
    • 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—Information and communication technology [ICT] specially adapted for implementation of business processes of specific business sectors, e.g. utilities or tourism
    • G06Q50/08—Construction
    • H—ELECTRICITY
    • H02—GENERATION; CONVERSION OR DISTRIBUTION OF ELECTRIC POWER
    • H02J—CIRCUIT ARRANGEMENTS OR SYSTEMS FOR SUPPLYING OR DISTRIBUTING ELECTRIC POWER; SYSTEMS FOR STORING ELECTRIC ENERGY
    • H02J3/00—Circuit arrangements for ac mains or ac distribution networks
    • H02J3/003—Load forecast, e.g. methods or systems for forecasting future load demand
    • Y—GENERAL TAGGING OF NEW TECHNOLOGICAL DEVELOPMENTS; GENERAL TAGGING OF CROSS-SECTIONAL TECHNOLOGIES SPANNING OVER SEVERAL SECTIONS OF THE IPC; TECHNICAL SUBJECTS COVERED BY FORMER USPC CROSS-REFERENCE ART COLLECTIONS [XRACs] AND DIGESTS
    • Y04—INFORMATION OR COMMUNICATION TECHNOLOGIES HAVING AN IMPACT ON OTHER TECHNOLOGY AREAS
    • Y04S—SYSTEMS INTEGRATING TECHNOLOGIES RELATED TO POWER NETWORK OPERATION, COMMUNICATION OR INFORMATION TECHNOLOGIES FOR IMPROVING THE ELECTRICAL POWER GENERATION, TRANSMISSION, DISTRIBUTION, MANAGEMENT OR USAGE, i.e. SMART GRIDS
    • Y04S10/00—Systems supporting electrical power generation, transmission or distribution
    • Y04S10/50—Systems or methods supporting the power network operation or management, involving a certain degree of interaction with the load-side end user applications

Landscapes

  • Engineering & Computer Science (AREA)
  • Physics & Mathematics (AREA)
  • Theoretical Computer Science (AREA)
  • General Physics & Mathematics (AREA)
  • Health & Medical Sciences (AREA)
  • Life Sciences & Earth Sciences (AREA)
  • Biophysics (AREA)
  • Evolutionary Computation (AREA)
  • General Health & Medical Sciences (AREA)
  • Artificial Intelligence (AREA)
  • Software Systems (AREA)
  • Business, Economics & Management (AREA)
  • General Engineering & Computer Science (AREA)
  • Data Mining & Analysis (AREA)
  • Computing Systems (AREA)
  • Mathematical Physics (AREA)
  • Computational Linguistics (AREA)
  • Biomedical Technology (AREA)
  • Molecular Biology (AREA)
  • Human Resources & Organizations (AREA)
  • Strategic Management (AREA)
  • Economics (AREA)
  • Tourism & Hospitality (AREA)
  • Bioinformatics & Cheminformatics (AREA)
  • Evolutionary Biology (AREA)
  • General Business, Economics & Management (AREA)
  • Bioinformatics & Computational Biology (AREA)
  • Marketing (AREA)
  • Operations Research (AREA)
  • Entrepreneurship & Innovation (AREA)
  • Game Theory and Decision Science (AREA)
  • Development Economics (AREA)
  • Nonlinear Science (AREA)
  • Algebra (AREA)
  • Computational Mathematics (AREA)
  • Mathematical Analysis (AREA)
  • Mathematical Optimization (AREA)
  • Pure & Applied Mathematics (AREA)
  • Power Engineering (AREA)
  • Quality & Reliability (AREA)

Abstract

The invention relates to an office building energy consumption prediction method based on improved atomic search optimization BP network. Firstly, collecting historical energy consumption data of an office building and preprocessing the collected data; optimizing a BP neural network by adopting an improved atomic search algorithm, and predicting the energy consumption of the office building; the improved atomic search algorithm is used for optimizing the weight and the threshold of the BP neural network, constructing an optimal model of the BP neural network, initializing an atomic population by utilizing the Tent chaotic mapping, and enhancing the global search capability of the algorithm. The invention can improve the convergence rate of the BP network and further improve the prediction accuracy of the algorithm.

Description

Office building energy consumption prediction method based on improved atomic search optimization BP network
Technical Field
The invention relates to the technical field of short-term prediction of office building energy consumption, in particular to an office building energy consumption prediction method based on improved atomic search optimization BP network.
Background
Building, industry, traffic are three important components of social energy consumption, called energy consumption "three-megaly". The building energy consumption analysis is the basis of building energy conservation work, and by comprehensively analyzing the building energy consumption, particularly comprehensively analyzing the main building energy consumption systems (heating ventilation and air conditioning systems, heating and lighting systems, other equipment systems and the like) in the operation process, the equipment systems are reasonably optimized and controlled, so that the aim of saving energy is achieved. The energy consumption prediction is carried out on the existing buildings, the existing buildings are expanded or newly built, the prediction analysis on the energy consumption of the buildings is enhanced, whether the buildings meet the energy-saving design specification requirements can be checked, energy-saving potential points are excavated from the buildings, and the method has positive practical significance for determining a feasible energy-saving scheme.
The artificial neural network is a method proposed on the basis of modern neuroscience research results, and has the advantages of strong self-learning capability, capability of fitting any nonlinear function and the like by simulating cerebral nerves to perform information processing. The artificial neural network modeling process can build a building energy consumption prediction model without knowing building geometric parameters or thermal parameters and without making any model assumptions. However, the method has some defects, the problem that the standard BP neural network is easy to be in local optimum in the process of adjusting the weight and the threshold value, and the selection of the weight threshold value of the BP neural network is required to be optimized, so that the accuracy of energy consumption prediction is further improved.
Disclosure of Invention
The invention aims to solve the problems that a standard BP neural network is easy to fall into a local minimum value and the accuracy of an initial weight and a threshold value is reduced due to random setting, so that the office building energy consumption prediction method based on the improved atomic search optimization BP network is provided, the convergence speed of the BP network can be improved, and meanwhile, the prediction accuracy of office building energy consumption can be improved.
In order to achieve the above purpose, the technical scheme of the invention is as follows: an office building energy consumption prediction method based on improved atomic search optimization BP network comprises the following steps:
step 1, acquiring office building energy consumption historical data, and selecting influence factors of office building energy consumption;
step 2, constructing an office building energy consumption short-term prediction initial model of the BP neural network;
step 3, constructing a fitness function;
step 4, optimizing the weight and the threshold of the initial model of the office building energy consumption prediction of the BP neural network by using an improved atomic search algorithm;
and 5, inputting the test sample into an office building energy consumption prediction model of the optimized BP neural network to obtain a power load prediction result.
In an embodiment of the present invention, the specific implementation manner of the step 1 is: taking historical energy consumption data measured every day in a preset time as a sample, and dividing the sample into a training set and a testing set; and 5 factors including the average day temperature, the highest temperature, the lowest temperature, the working day and the humidity are selected as the influencing factors of the office building, and normalization processing is carried out.
In an embodiment of the present invention, the specific implementation manner of the step 2 is: initializing a connection weight between an input layer and an hidden layer and a threshold value of hidden neurons, setting the number of the neurons of the hidden layer, transmitting the input sample data from the input layer to the hidden layer through forward transmission, and selecting an S-shaped function as a transmission function of the hidden layer; and then taking the output of the hidden layer as the input of the output layer, obtaining corresponding output data through a linear transfer function, adjusting the weight of the hidden layer through the back propagation of errors, continuously setting the errors, and finally constructing an office building energy consumption short-term prediction initial model of the BP neural network by the output result.
In an embodiment of the present invention, the specific implementation manner of the step 3 is: the mean value of the mean square error of the training set and the whole test set is selected as a fitness function, and the expression is:
Figure SMS_1
wherein k is the operation times, mse (train) is a training error, mse (test) is a test set error, and an error obtained by summing up ten operations and taking an average value is used as an evaluation index.
In the embodiment of the invention, in the step 4, an improved atomic search algorithm based on Tent chaotic mapping is adopted to optimize an initial model of office building energy consumption prediction of the BP neural network, so as to obtain an optimal prediction model.
In an embodiment of the present invention, the specific implementation manner of the step 4 is:
step 4.1, an atom searching algorithm guides the group to search and optimize through interaction force among atoms in the group according to the motion rule of the atoms; firstly, setting initialization algorithm parameters PS, G, alpha and beta, wherein PS represents an initial atomic scale, G represents a maximum iteration number, alpha represents a depth weight, and beta represents a multiplier weight;
step 4.2, initializing a search space, generating initial speed of atoms, and initializing population positions of the atoms through Tent chaotic mapping;
step 4.3, calculating the initial fitness of newly generated atomic individuals, sequencing the atomic individuals, and selecting the individual with the best fitness as an initial atomic group;
step 4.4, judging whether the updated position exceeds a set range, and adjusting atoms exceeding the set range;
step 4.5, comparing the optimal fitness value after the iteration with the historical optimal fitness value, and updating the historical optimal fitness;
and 4.6, judging whether the maximum searching times are reached, if so, outputting the position of the atom at the moment, taking the position of the atom as the optimal weight and the threshold of the short-term prediction model of the office building energy consumption of the optimized BP neural network, and otherwise, returning to the step 4.3.
In an embodiment of the present invention, the specific implementation manner of step 4.3 is:
the location of each atom in the feasible solution in the population of atoms represents the solution of the problem and is represented in mass; heavier atoms have less acceleration, meaning that they have a better solution; lighter atoms have greater accelerations, indicating that their solutions are poor; at the same time, the interaction forces between atoms are respectively attractive force and repulsive force; in the iterative process of the algorithm, attractive force can guide the atoms to move, and repulsive force can effectively prevent the atoms from converging prematurely; through the interaction between atoms, attraction and repulsion jointly complete global search and local search of an algorithm; in an atomic system, atoms are represented in three main forms: interaction force, geometric constraint and atomic motion, respectively;
the Lennard-Jones potential force is the source of force that produces the interaction force, and the total force on the ith particle is then expressed as:
Figure SMS_2
wherein rand is j ∈[0,1]Kbest is a set of K atoms, expressed as the first K atoms with higher fitness values, and in the initial iteration stage, each atom interacts with other atoms with better fitness as much as possible, i.e. surrounding K neighbor individuals; later in the iteration, each atom should interact with a better-fitting atom to minimize interactions, so K is defined as follows:
Figure SMS_3
where N is the atomic population size, T is the current iteration number, T is the maximum iteration number,
Figure SMS_4
is the interaction force of the ith atom and the jth atom in the ith iteration in the d dimension,/th>
Figure SMS_5
Is defined as shown in the following formula:
Figure SMS_6
wherein the method comprises the steps of
Figure SMS_7
The ratio of the distance between two atoms to the length scalar, i.e., the scalar distance between two atoms; />
Figure SMS_8
Is the position difference vector of the ith atom and the jth atom; r is (r) ij Is the Euclidean distance of the ith and jth atoms; - η (t) represents a depth function for adjusting the attraction and repulsion areas, expressed by the following formula:
Figure SMS_9
where α is the depth weight, function h ij (t) is expressed as:
Figure SMS_10
Figure SMS_11
wherein g 0 Is a lower boundary, u is an upper boundary, g (t) represents a floating factor, and is used for controlling the development process of an algorithm; sigma (t) represents the collision radius, g (t) and sigma (t) are expressed as follows:
Figure SMS_12
Figure SMS_13
x in the formula ij (t) represents the positions of atom i and atom j; geometric constraints in molecular dynamics are linked by covalent bonds between atoms; to solve the optimization problem, assuming that each atom is connected to the atom with the best fitness, the constraint is expressed by the following formula:
Figure SMS_14
wherein the method comprises the steps of
Figure SMS_15
Is the position of the best atom in the t-th iteration; the lagrangian factor λ (t) is defined as follows:
Figure SMS_16
wherein β represents a multiplier weight;
the motion of the atomic system accords with the classical mechanical law, and under the combined action of the interaction force and the geometric constraint force, the acceleration equation of each atom is expressed by the following formula:
Figure SMS_17
wherein a represents atomic acceleration; m represents the mass of an atom;
the mass of an atom in an atom search algorithm is defined as the following formula:
Figure SMS_18
wherein m is i (t) represents the mass of the atoms of the t-th iteration;
Figure SMS_19
representing a quality estimate of the atoms of the t-th iteration; n represents the total number of atoms; fit worst (t) and Fit best (t) the worst fitness value and the best fitness value in the t-th iteration, respectively; fit i (t) represents the objective function value corresponding to the t-th iteration atom;
in the atomic search optimization algorithm, the speed update and the position update of the atoms in the d-dimensional space are respectively shown in the following formulas:
Figure SMS_20
Figure SMS_21
Figure SMS_23
is the speed of atoms; />
Figure SMS_26
Is the position of an atom; />
Figure SMS_28
Is [0,1]Random numbers in between; />
Figure SMS_24
Is the acceleration of the ith atom in the d dimension; d=1, 2, … D, D is the dimension size of the search space; />
Figure SMS_25
Wherein v is max Autonomously setting; />
Figure SMS_27
Wherein->
Figure SMS_29
And->
Figure SMS_22
Representing the lower and upper bounds, respectively, of the search space in the D dimension.
Compared with the prior art, the invention has the following beneficial effects: aiming at the problem that an atomic search algorithm is easy to trap into local optimum, a Tent chaotic mapping method is introduced, so that the diversity of atomic populations is increased, atoms move towards the direction of an optimum solution, the possibility of trapping into the local optimum is effectively reduced, and the optimizing effect of the atomic search algorithm is improved;
according to the invention, the BP neural network is improved by using the optimized atomic search algorithm, the problems of random weight and threshold setting of the BP neural network are solved, the prediction accuracy of the algorithm is improved, and a data support is provided for promoting energy conservation and emission reduction.
Drawings
FIG. 1 is a flow chart of the method of the present invention.
Fig. 2 is a schematic diagram of a flow structure of a BP neural network optimized by an improved atomic search algorithm.
Fig. 3 is a comparison of the results of the Tent-ASO-BP combination method to predict the energy consumption of an office building with the actual values.
FIG. 4 is a graph showing the error between the predicted value and the actual value of the Tent-ASO-BP combination method and the BP neural network.
Detailed Description
The technical scheme of the invention is specifically described below with reference to the accompanying drawings.
The invention provides an office building energy consumption prediction method based on improved atomic search optimization BP network, which comprises the following steps:
step 1, acquiring office building energy consumption historical data, and selecting influence factors of office building energy consumption;
step 2, constructing an office building energy consumption prediction initial model of the BP neural network;
step 3, constructing a fitness function;
step 4, optimizing the weight and the threshold of the initial model of the office building energy consumption prediction of the BP neural network by using an improved atomic search algorithm;
and 5, inputting the test sample into an office building energy consumption prediction model of the optimized BP neural network to obtain a power load prediction result.
As shown in fig. 1-4, the embodiment of the invention provides an office building energy consumption prediction method based on improved atomic search optimization BP network.
Considering that in a BP neural network, the initialization of the parameter weights and thresholds is random and relatively sensitive, unreasonable weights and thresholds can reduce the predictive performance of the neural network. The quality of the initial weight and the threshold value can have influence on forward transmission of signals and back propagation of errors, and a certain deviation exists in the convergence direction.
As shown in FIG. 2, the method optimizes BP by adopting an improved atomic search algorithm to obtain an optimal model aiming at the problem that the BP algorithm is not high in accuracy in actual conditions. Meanwhile, in order to avoid the atomic search algorithm from falling into local optimum in the initial stage, the Tent chaotic mapping is used for improvement in the aspect of initial value selection.
In order to examine the effect of predicting the energy consumption of an office building designed by the invention, the invention is applied to predict the actual energy consumption of a certain office building, as shown in fig. 1, and the method specifically comprises the following steps:
step one: and collecting original data and preprocessing the data. The collected data are 252 days of electricity consumption data of an office building 2021 month 1 to 2021 month 9 in a certain city and daily weather data of the certain city, the energy consumption data of the office building is used for training the energy consumption prediction model designed by the invention, and the prediction result is compared with the sample result and the effect is evaluated. The 5 characteristic variables of the average day temperature, the highest temperature, the lowest temperature, the working day and the humidity are taken as the input variables of the model, and the output characteristic variable is the energy consumption of the office building. Because the data are different in dimension, the data difference is relatively large, and dimensionless processing is needed to be carried out on the data, namely, the normalization of the data is carried out, and the normalization formula can be expressed as follows:
x k ＝(x-x min )/(x max -x min )
where x is the raw dataset before processing, x k For normalized dataset, x max And x min Respectively minimum and maximum values before data processing.
Step two: and establishing an initial BP neural network model. According to sample data, setting the number of neurons of an input layer as 5, setting the number of neurons of an output layer as 1, initially calculating the number of nodes of an hidden layer according to an empirical formula, and then determining the number of nodes of the hidden layer with the best effect by adopting a ten-fold cross validation method, namely training and testing a data set of the whole sample for 10 times, and taking the number of nodes of the hidden layer with the minimum corresponding mean square error. Setting the learning rate alpha=0.01 of the BP neural network, setting the maximum iteration times as 1000, setting the hidden layer activation function as a sigmoid function, setting the output layer activation function as a purelin function, setting the back propagation training function as a tranlm function, then performing preliminary calculation, inputting 5 characteristic variables of the preprocessed daily average temperature, highest temperature, lowest temperature, whether the preprocessed daily average temperature is workday and humidity into the BP neural network, and predicting to obtain the daily electricity consumption value of the office building.
Step three: and constructing an adaptability function. And a ten-fold cross validation method is adopted, a sample data set is divided into ten groups of data, one group of data is used as a test set in each operation, the other nine groups of data are used as training sets, and the average value of errors of the training sets and the test sets is used as the average error of training. The error obtained by adding and averaging ten operations is used as the fitness function of the model.
The expression of the fitness function is as follows:
Figure SMS_30
wherein k is the operation times, mse (train) is a training error, mse (test) is a test set error, and an error obtained by summing up ten operations and taking an average value is used as an evaluation index.
Step four: aiming at the problem that the prediction accuracy of the algorithm is reduced due to the sensitivity of initial value selection in the actual operation of the BP neural network algorithm, an improved atomic search algorithm based on Tent chaotic mapping is adopted to optimize the weight and the threshold of the BP neural network.
Step 4.1, setting the atomic population scale to be 30, the maximum iteration number to be 50, the upper boundary of the initial value to be 3, the lower boundary to be-3, the depth weight to be 50 and the multiplier weight to be 0.2.
Step 4.2 the step of improving the atomic search algorithm is specifically as follows:
the Tent chaotic map is Tent map, and the Tent chaotic map is utilized to replace an initialized atomic population randomly generated in an atomic search algorithm, so that atoms can be more uniformly distributed in a search space, the possibility that the population falls into local optimum can be reduced, and the formula is shown as follows:
Figure SMS_31
wherein beta epsilon (0, 1),
Figure SMS_32
representing the d-dimension position of the ith sparrow under the t-th iteration, wherein the dimension of the chaotic matrix is the same as the dimension of the atomic population matrix, generating the chaotic matrix, and then projecting the chaotic matrix into a weight threshold limiting boundary to obtain the atomic population with chaotic initialization.
Calculating initial atom fitness, and selecting atoms with the best initial fitness and atom positions in the population, and atoms with the worst initial fitness and atom positions in the population;
and 4.3, sorting the fitness of the atomic population, and selecting the individual with the best fitness as the initial population. And updating three forms of interaction force, geometric constraint and atom motion of atoms according to a formula.
(1) The interaction force can be expressed as:
Figure SMS_33
wherein rand is j ∈[0,1]Where Kbest is a set of K atoms, denoted as the first K atoms with higher fitness values, in order to have each atom interact as much as possible with other atoms with better fitness, i.e. surrounding K neighbor individuals, at the beginning of the iteration; later in the iteration, each atom should interact with a better-fitting atom to minimize. According to this design concept, K is defined as follows:
Figure SMS_34
/>
where N is the atomic population size, T is the current iteration number, and T is the maximum iteration number.
Figure SMS_35
Is the interaction force of the ith atom and the jth atom in the d-th dimension in the t-th iteration. />
Figure SMS_36
Is defined as shown in the following formula:
Figure SMS_37
wherein the method comprises the steps of
Figure SMS_38
The ratio of the distance between two atoms to the length scalar, i.e. the scalar distance between two atoms, is indicated. />
Figure SMS_39
Is the position difference vector between the i-th atom and the j-th atom. r is (r) ij Is the euclidean distance of the i-th atom and the j-th atom. - η (t) represents a depth function for adjusting the attraction and repulsion areas, which can be expressed by the following formula:
Figure SMS_40
where α is depthThe weight is set to 50, T is the current iteration number, T is the maximum iteration number, and the function h ij Can be expressed as
Figure SMS_41
Figure SMS_42
Wherein g 0 Is the lower boundary, u is the upper boundary, g (t) represents the floating factor, used to control the development process of the algorithm. Sigma (sigma) (t) Denotes the collision radius, g (t) and sigma (t) The following formula can be used:
Figure SMS_43
Figure SMS_44
(2) The geometric constraint of an atom can be expressed as:
Figure SMS_45
wherein the method comprises the steps of
Figure SMS_46
Is the position of the best atom in the t-th iteration. The lagrangian factor λ (t) may be defined as follows:
Figure SMS_47
wherein beta represents a multiplier weight set to 0.2
(3) In combination with the interaction force between atoms and the geometric constraint force, the acceleration of the ith atom in the d-th dimension at time t can be expressed as:
Figure SMS_48
wherein a represents atomic acceleration; m represents the mass of an atom.
The mass of an atom in an atom search algorithm is defined as the following formula:
Figure SMS_49
wherein m is i (t) represents the mass of the atoms of the t-th iteration;
Figure SMS_50
representing a quality estimate of the atoms of the t-th iteration; n represents the total number of atoms; fit worst (t) and Fit best (t) the worst fitness value and the best fitness value in the t-th iteration, respectively; fit i And (t) represents the objective function value corresponding to the t-th iteration atom.
The velocity update and the position update of atoms in d-dimensional space are respectively shown in the following formulas:
Figure SMS_51
Figure SMS_52
Figure SMS_54
is the speed of atoms; />
Figure SMS_56
Is the position of an atom; />
Figure SMS_58
Is [0,1]Random numbers in between; />
Figure SMS_55
Is the acceleration of the ith atom in the d dimension; d=1, 2, … D, D is searchDimension of space. />
Figure SMS_57
Wherein v is max And (5) autonomous setting. />
Figure SMS_59
Wherein->
Figure SMS_60
And->
Figure SMS_53
Representing the lower and upper bounds, respectively, of the search space in the D dimension.
Step 4.4: judging whether the updated position exceeds a set range or not, and adjusting atoms exceeding the set range;
step 4.5: comparing the optimal fitness value after the iteration with the historical optimal fitness value, and updating the historical optimal fitness value;
step 4.6: judging whether the maximum searching times are reached, if yes, outputting the position of the atom at the moment, taking the position of the atom as the optimal weight and the threshold value of the BP neural network, otherwise, returning to the step 4.3.
Step five: inputting the test sample into an office building energy consumption short-term prediction model of the trained atomic search algorithm optimization BP neural network to obtain a short-term power load prediction result;
in order to illustrate the effect of the office building energy consumption prediction method, the calculation result of the Tent-ASO-BP method is compared with the original BP neural network algorithm, the calculation errors of the test sample building energy consumption true value and the prediction value calculated by the two methods are shown in figure 3, and the calculation errors of the two methods are shown in figure 4. As can be seen from the graph, the prediction result of the BP neural network prediction method based on the improved atomic search algorithm provided by the invention is superior to the prediction result of a standard BP neural network, the MAPE (mean absolute percentage error) of the Tent-ASO-BP method is 3.43%, the MAPE (mean absolute percentage error) of the BP neural network method is 6.59%, the index shows that the prediction error of the improved method is smaller, the fitting degree with a true value is better, and the prediction precision of the office building energy consumption can be effectively improved.
The above is a preferred embodiment of the present invention, and all changes made according to the technical solution of the present invention belong to the protection scope of the present invention when the generated functional effects do not exceed the scope of the technical solution of the present invention.

Claims (7)

1. An office building energy consumption prediction method based on improved atomic search optimization BP network is characterized by comprising the following steps:
step 1, acquiring office building energy consumption historical data, and selecting influence factors of office building energy consumption;
step 2, constructing an office building energy consumption prediction initial model of the BP neural network;
step 3, constructing a fitness function;
step 4, optimizing the weight and the threshold of the initial model of the office building energy consumption prediction of the BP network by using an improved atomic search algorithm;
and 5, inputting the test sample into an office building energy consumption prediction model of the optimized BP network to obtain a power load prediction result.
2. The office building energy consumption prediction method based on the improved atomic search optimization BP network according to claim 1, wherein the specific implementation manner of step 1 is as follows: taking historical energy consumption data measured every day in a preset time as a sample, and dividing the sample into a training set and a testing set; and 5 factors including the average day temperature, the highest temperature, the lowest temperature, the working day and the humidity are selected as the influencing factors of the office building, and normalization processing is carried out.
3. The office building energy consumption prediction method based on the improved atomic search optimization BP network according to claim 1, wherein the specific implementation manner of step 2 is as follows: initializing a connection weight between an input layer and an hidden layer and a threshold value of hidden neurons, setting the number of the neurons of the hidden layer, transmitting the input sample data from the input layer to the hidden layer through forward transmission, and selecting an S-shaped function as a transmission function of the hidden layer; and then taking the output of the hidden layer as the input of the output layer, obtaining corresponding output data through a linear transfer function, adjusting the weight of the hidden layer through the back propagation of errors, continuously setting the errors, and finally constructing an office building energy consumption prediction initial model of the BP network by the output result.
4. The office building energy consumption prediction method based on the improved atomic search optimization BP network according to claim 2, wherein the specific implementation manner of step 3 is as follows: the mean value of the mean square error of the training set and the whole test set is selected as a fitness function, and the expression is:
Figure QLYQS_1
wherein k is the operation times, mse (train) is a training error, mse (test) is a test set error, and an error obtained by summing up ten operations and taking an average value is used as an evaluation index.
5. The method for predicting the office building energy consumption based on the improved atomic search optimization BP network according to claim 1, wherein in the step 4, an improved atomic search algorithm based on Tent chaotic mapping is adopted to optimize an initial office building energy consumption prediction model of the BP network, so as to obtain an optimal prediction model.
6. The office building energy consumption prediction method based on the improved atomic search optimization BP network according to claim 1 or 5, wherein the specific implementation manner of step 4 is as follows:
step 4.1, an atom searching algorithm guides the group to search and optimize through interaction force among atoms in the group according to the motion rule of the atoms; firstly, setting initialization algorithm parameters PS, G, alpha and beta, wherein PS represents an initial atomic scale, G represents a maximum iteration number, alpha represents a depth weight, and beta represents a multiplier weight;
step 4.2, initializing a search space, generating initial speed of atoms, and initializing population positions of the atoms through Tent chaotic mapping;
step 4.3, calculating the initial fitness of newly generated atomic individuals, sequencing the atomic individuals, and selecting the individual with the best fitness as an initial atomic group;
step 4.4, judging whether the updated position exceeds a set range, and adjusting atoms exceeding the set range;
step 4.5, comparing the optimal fitness value after the iteration with the historical optimal fitness value, and updating the historical optimal fitness;
and 4.6, judging whether the maximum searching times are reached, if yes, outputting the position of the atom at the moment, taking the position of the atom as the optimal weight and the threshold of the office building energy consumption prediction model of the optimized BP network, and otherwise, returning to the step 4.3.
7. The method for predicting the office building energy consumption based on the improved atomic search optimization BP network according to claim 6, wherein the specific implementation manner of step 4.3 is as follows:
the location of each atom in the feasible solution in the population of atoms represents the solution of the problem and is represented in mass; heavier atoms have less acceleration, meaning that they have a better solution; lighter atoms have greater accelerations, indicating that their solutions are poor; at the same time, the interaction forces between atoms are respectively attractive force and repulsive force; in the iterative process of the algorithm, attractive force can guide the atoms to move, and repulsive force can effectively prevent the atoms from converging prematurely; through the interaction between atoms, attraction and repulsion jointly complete global search and local search of an algorithm; in an atomic system, atoms are represented in three main forms: interaction force, geometric constraint and atomic motion, respectively;
the Lennard-Jones potential force is the source of force that produces the interaction force, and the total force on the ith particle is then expressed as:
Figure QLYQS_2
wherein rand is j ∈[0,1]Kbest is a set of K atoms, expressed as the first K atoms with higher fitness values, and in the initial iteration stage, each atom interacts with other atoms with better fitness as much as possible, i.e. surrounding K neighbor individuals; later in the iteration, each atom should interact with a better-fitting atom to minimize interactions, so K is defined as follows:
Figure QLYQS_3
where N is the atomic population size, T is the current iteration number, T is the maximum iteration number,
Figure QLYQS_4
is the interaction force of the ith atom and the jth atom in the ith iteration in the d dimension,/th>
Figure QLYQS_5
Is defined as shown in the following formula:
Figure QLYQS_6
wherein the method comprises the steps of
Figure QLYQS_7
The ratio of the distance between two atoms to the length scalar, i.e., the scalar distance between two atoms; />
Figure QLYQS_8
Is the position difference vector of the ith atom and the jth atom; r is (r) ij Is the Euclidean distance of the ith and jth atoms; - η (t) represents a depth function for adjusting the attraction and repulsion areas, expressed by the following formula:
Figure QLYQS_9
where α is the depth weight, function h ij (t) is expressed as:
Figure QLYQS_10
Figure QLYQS_11
wherein g 0 Is a lower boundary, u is an upper boundary, g (t) represents a floating factor, and is used for controlling the development process of an algorithm; sigma (t) represents the collision radius, g (t) and sigma (t) are expressed as follows:
Figure QLYQS_12
Figure QLYQS_13
x in the formula ij (t) represents the positions of atom i and atom j; geometric constraints in molecular dynamics are linked by covalent bonds between atoms; to solve the optimization problem, assuming that each atom is connected to the atom with the best fitness, the constraint is expressed by the following formula:
Figure QLYQS_14
wherein the method comprises the steps of
Figure QLYQS_15
Is the position of the best atom in the t-th iteration; the lagrangian factor λ (t) is defined as follows:
Figure QLYQS_16
wherein β represents a multiplier weight;
the motion of the atomic system accords with the classical mechanical law, and under the combined action of the interaction force and the geometric constraint force, the acceleration equation of each atom is expressed by the following formula:
Figure QLYQS_17
wherein a represents atomic acceleration; m represents the mass of an atom;
the mass of an atom in an atom search algorithm is defined as the following formula:
Figure QLYQS_18
wherein m is i (t) represents the mass of the atoms of the t-th iteration;
Figure QLYQS_19
representing a quality estimate of the atoms of the t-th iteration; n represents the total number of atoms; fit worst (t) and Fit best (t) the worst fitness value and the best fitness value in the t-th iteration, respectively; fit i (t) represents the objective function value corresponding to the t-th iteration atom;
in the atomic search optimization algorithm, the speed update and the position update of the atoms in the d-dimensional space are respectively shown in the following formulas:
Figure QLYQS_20
Figure QLYQS_21
Figure QLYQS_24
is the speed of atoms; />
Figure QLYQS_26
Is the position of an atom; />
Figure QLYQS_27
Is [0,1]Random numbers in between; />
Figure QLYQS_23
Is the acceleration of the ith atom in the d dimension; d=1, 2, … D, D is the dimension size of the search space; />
Figure QLYQS_25
Wherein v is max Autonomously setting;
Figure QLYQS_28
wherein->
Figure QLYQS_29
And->
Figure QLYQS_22
Representing the lower and upper bounds, respectively, of the search space in the D dimension. />
CN202310176307.6A 2023-02-28 2023-02-28 Office building energy consumption prediction method based on improved atomic search optimization BP network Pending CN116187565A (en)

Priority Applications (1)

Application Number Priority Date Filing Date Title
CN202310176307.6A CN116187565A (en) 2023-02-28 2023-02-28 Office building energy consumption prediction method based on improved atomic search optimization BP network

Applications Claiming Priority (1)

Application Number Priority Date Filing Date Title
CN202310176307.6A CN116187565A (en) 2023-02-28 2023-02-28 Office building energy consumption prediction method based on improved atomic search optimization BP network

Publications (1)

Publication Number Publication Date
CN116187565A true CN116187565A (en) 2023-05-30

Family

ID=86434317

Family Applications (1)

Application Number Title Priority Date Filing Date
CN202310176307.6A Pending CN116187565A (en) 2023-02-28 2023-02-28 Office building energy consumption prediction method based on improved atomic search optimization BP network

Country Status (1)

Country Link
CN (1) CN116187565A (en)

Cited By (2)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN116776935A (en) * 2023-06-09 2023-09-19 淮阴工学院 Improved MICN-based photovoltaic power prediction method
CN117436707A (en) * 2023-12-18 2024-01-23 厦门锋联信息技术有限公司 Fire safety management method and system based on artificial intelligence

Cited By (4)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN116776935A (en) * 2023-06-09 2023-09-19 淮阴工学院 Improved MICN-based photovoltaic power prediction method
CN116776935B (en) * 2023-06-09 2024-02-23 淮阴工学院 Improved MICN-based photovoltaic power prediction method
CN117436707A (en) * 2023-12-18 2024-01-23 厦门锋联信息技术有限公司 Fire safety management method and system based on artificial intelligence
CN117436707B (en) * 2023-12-18 2024-03-22 厦门锋联信息技术有限公司 Fire safety management method and system based on artificial intelligence

Similar Documents

Publication Publication Date Title
CN116187565A (en) Office building energy consumption prediction method based on improved atomic search optimization BP network
CN110705743B (en) New energy consumption electric quantity prediction method based on long-term and short-term memory neural network
CN103164742B (en) A kind of server performance Forecasting Methodology based on particle group optimizing neural network
CN110942205B (en) Short-term photovoltaic power generation power prediction method based on HIMVO-SVM
CN114692265B (en) Zero-carbon building optimization design method based on deep reinforcement learning
CN111860989B (en) LSTM neural network short-time traffic flow prediction method based on ant colony optimization
CN106971238A (en) The Short-Term Load Forecasting Method of Elman neutral nets is obscured based on T S
CN112415924A (en) Energy-saving optimization method and system for air conditioning system
CN111079983A (en) Optimization method for vehicle path planning of assembly type construction site
CN115081592B (en) Highway low-visibility prediction method based on genetic algorithm and feedforward neural network
CN114219181A (en) Wind power probability prediction method based on transfer learning
CN113762387B (en) Multi-element load prediction method for data center station based on hybrid model prediction
CN115759359A (en) Elevator fault prediction method based on improved slime algorithm for optimizing Elman neural network
CN109242136A (en) A kind of micro-capacitance sensor wind power Chaos-Genetic-BP neural network prediction technique
CN114970815A (en) Traffic flow prediction method and device based on improved PSO algorithm optimized LSTM
CN114757104A (en) Construction method of series gate group water transfer engineering hydraulic real-time regulation model based on data driving
CN115906399A (en) Improved method for predicting key process quality of product under small sample data
CN108985323A (en) A kind of short term prediction method of photovoltaic power
CN117154690A (en) Photovoltaic power generation power prediction method and system based on neural network
Wang et al. Electric vehicle charging load clustering and load forecasting based on long short term memory neural network
CN109214610A (en) A kind of saturation Methods of electric load forecasting based on shot and long term Memory Neural Networks
CN117313795A (en) Intelligent building energy consumption prediction method based on improved DBO-LSTM
CN111310974A (en) Short-term water demand prediction method based on GA-ELM
CN113762591B (en) Short-term electric quantity prediction method and system based on GRU and multi-core SVM countermeasure learning
CN111401638B (en) Spatial load prediction method based on extreme learning machine and load density index method

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