Disclosure of Invention
Aiming at the technical problems in the prior art, the invention provides an air conditioner cold load prediction optimization method, an air conditioner cold load prediction optimization system and air conditioner cold load prediction equipment, and aims to solve the technical problems that in the existing air conditioner load prediction, as the input weights of a network input layer and an hidden layer and the threshold value of the hidden layer are randomly generated, the generalization capability of a prediction model is poor and the prediction result is unstable.
In order to achieve the above purpose, the invention adopts the following technical scheme:
the invention provides an air conditioner cold load prediction optimization method, which comprises the following steps:
acquiring influence component data of the air conditioner cold load, and determining main influence index data of the air conditioner cold load;
the method comprises the steps of training and optimizing a limit learning machine network by adopting main influence index data of air conditioner cold load; the training optimization process adopts a whale optimization algorithm to optimize weight parameters and threshold parameters of the extreme learning machine, so that the optimized extreme learning machine is obtained;
and acquiring main influence index data of the air conditioner cold load, inputting the main influence index data into the optimized extreme learning machine, and outputting to obtain an air conditioner cold load prediction optimization result.
Further, the data of the influence components of the air conditioner cooling load comprise the indoor temperature of the building and CO 2 Concentration, total level radiation, outdoor air temperature, relative humidity, wet bulb temperature, and wind speed.
Further, the process of determining the main impact index data of the air conditioner cooling load specifically includes:
and preprocessing the influence component data of the air conditioner cold load by adopting a random forest algorithm to obtain main influence index data of the air conditioner cold load.
Furthermore, the influence component data of the air conditioner cold load is preprocessed by adopting a random forest algorithm, and the method comprises the following steps of:
selecting the data of the influencing constituent elements of the air conditioner cooling load by adopting a single variable selection mode; establishing a random forest model by using the influence constituent element data of each air conditioner cold load and the air conditioner cold load value respectively; taking the determined coefficient as an evaluation index of the random forest model;
the determining coefficient is the influence degree of the air conditioner cold load influence factor on the air conditioner cold load value; training a random forest model by adopting a model training mode of 5-fold cross validation, and screening to obtain main influence index data of the air conditioner cold load.
Furthermore, the weight parameters and the threshold parameters of the limit learning machine are optimized by adopting a whale algorithm, and the method specifically comprises the following steps:
setting prediction precision parameters of the extreme learning machine, and determining a prediction precision parameter range of the extreme learning machine;
setting optimization ranges of influence parameters of a whale optimization algorithm according to the prediction precision parameter ranges of the extreme learning machine;
obtaining a predicted value of the extreme learning machine by using main influence index data of the air conditioner cold load and whale population information; calculating fitness values of all individuals according to the predicted value of the extreme learning machine, selecting the current optimal fitness individual, and setting the position of the individual as the current optimal;
iteratively updating the individual position by utilizing a contraction surrounding mechanism, a spiral position updating mechanism and an exploration mechanism of a whale optimization algorithm;
after each iteration update, the optimized prediction precision parameters are transmitted to a limit learning machine;
judging whether the iteration cycle times reach a preset value, if so, stopping optimizing the prediction precision parameter to obtain an optimal weight parameter and a threshold value parameter of the extreme learning machine; if not, the iterative update continues.
Further, the prediction precision parameters of the extreme learning machine comprise input weight parameters and hidden layer threshold parameters; wherein, the input weight parameter is set as a random number in the range of [ -1,1 ]; the hidden layer threshold parameter is set to a random number in the range of 0, 1.
Further, the whale optimization algorithm influence parameters comprise population size, maximum iteration times and upper and lower limits of whale population positions; the fitness function of the whale optimization algorithm is the mean square error.
Further, the individual position is subjected to an iterative updating process by utilizing a contraction surrounding mechanism, a spiral updating position mechanism and an exploration mechanism of a whale optimizing algorithm;
if the current iteration times t<Maximum number of iterations T max Updating the input weight and the hidden layer threshold;
when the random variable p is less than 0.5, if the coefficient vector |A| is more than or equal to 1, the whale will give up the hunting and search again; if the coefficient vector |A| <1, whale will attack the prey;
when the random variable p is more than or equal to 0.5, the individual position is updated in a spiral way.
The invention also provides an air conditioner cold load prediction optimization system, which comprises an influence index data module, a model optimization module and an output module;
the influence index data module is used for acquiring influence component data of the air conditioner cold load and determining main influence index data of the air conditioner cold load;
the model optimization module is used for training and optimizing the limit learning machine network by adopting main influence index data of the air conditioner cold load; the training optimization process adopts a whale optimization algorithm to optimize weight parameters and threshold parameters of the extreme learning machine, so that the optimized extreme learning machine is obtained;
and the output module is used for acquiring the main impact index data of the air conditioner cooling load, inputting the main impact index data into the optimized extreme learning machine, and outputting and obtaining the air conditioner cooling load prediction optimization result.
The invention also provides air conditioner cold load prediction optimizing equipment, which comprises a memory, a processor and executable instructions which are stored in the memory and can run in the processor; and the air conditioner cold load prediction optimization method is realized when the processor executes the executable instruction.
Compared with the prior art, the invention has the beneficial effects that:
the invention provides an air conditioner cold load prediction optimization method and system, which are characterized in that main influence index data of an air conditioner cold load is determined, and a limit learning machine is trained and optimized by adopting the main influence index data, so that the dimension of an input variable of the improved limit learning machine is reduced, the convergence speed is increased, and the operation cost is saved; meanwhile, the whale optimization algorithm is combined with the extreme learning machine, so that the mean square error of the air conditioner cold load prediction of the extreme learning machine is reduced, and the prediction performance is effectively improved.
Furthermore, the importance degree of the air conditioner cold load influencing constituent elements is evaluated by adopting a random forest algorithm, and the main influence index influencing the air conditioner cold load is determined, so that the pre-screening of characteristic data is better realized, the prediction precision of a model is improved, the dimensionality of an input variable of an improved extreme learning machine is reduced, the convergence rate is improved, and the operation cost is saved.
Furthermore, a whale algorithm is introduced to optimize the weight and the threshold of the extreme learning machine, so that the problem of optimal selection of the extreme learning machine in the air conditioner cold load prediction parameters is solved; the whale optimization algorithm is organically combined with the extreme learning machine algorithm, a prediction model based on random forests and improved extreme learning machines is constructed, the mean square error of the air conditioner cold load prediction of the extreme learning machines is reduced, and the prediction performance is effectively improved.
The invention provides an air conditioner cold load prediction optimization method and system, which utilize the organic combination of an extreme learning machine and a whale optimization algorithm to improve the prediction effect of the extreme learning machine in a mode of optimizing relevant parameters of the extreme learning machine; considering the complexity of the cold load data set and the influence degree of the prediction index on the prediction effect, the characteristic index of the data set is evaluated and screened by adopting a random forest algorithm, so that the influence of excessive useless characteristic data on the prediction effect is effectively avoided; the invention is adopted to predict the air conditioner cold load, and is respectively compared with three prediction models of a support vector machine, a convolutional neural network and a limit learning machine before improvement.
Detailed Description
In order to make the technical problems, technical schemes and beneficial effects solved by the invention more clear, the following specific embodiments are used for further describing the invention in detail. It should be understood that the specific embodiments described herein are for purposes of illustration only and are not intended to limit the scope of the invention.
The invention provides an air conditioner cold load prediction optimization method, which comprises the following steps:
step 1, acquiring data of influencing components of the cooling load of an air conditioner; wherein the data of the influencing components of the air conditioner cooling load comprise the indoor temperature of the building and CO 2 Concentration, total horizontal radiation, outdoor air temperature, relative humidity, wet bulb temperature, and wind speed; and preprocessing the influence component data of the air conditioner cold load by adopting a random forest algorithm to obtain main influence index data of the air conditioner cold load.
The preprocessing process is carried out on the influence constituent element data of the air conditioner cold load by adopting a random forest algorithm, and the preprocessing process is specifically as follows:
selecting the data of the influencing constituent elements of the air conditioner cooling load by adopting a single variable selection mode; establishing a random forest model by using the influence constituent element data of each air conditioner cold load and the air conditioner cold load value respectively; taking the determined coefficient as an evaluation index of the random forest model; the determining coefficient is the influence degree of the air conditioner cold load influence factor on the air conditioner cold load value; training a random forest model by adopting a model training mode of 5-fold cross validation, and screening to obtain main influence index data of the air conditioner cold load.
Step 2, training and optimizing a limit learning machine network by adopting main influence index data of the air conditioner cold load; the training optimization process adopts a whale optimization algorithm to optimize weight parameters and threshold parameters of the extreme learning machine, so that the optimized extreme learning machine is obtained; the specific process is as follows:
setting prediction precision parameters of the extreme learning machine, and determining a prediction precision parameter range of the extreme learning machine; the prediction precision parameters of the extreme learning machine comprise input weight parameters and hidden layer threshold parameters; wherein, the input weight parameter is set as a random number in the range of-1 to 1; the hidden layer threshold parameter is set to a random number in the range of 0 to 1.
Setting optimization ranges of influence parameters of a whale optimization algorithm according to the prediction precision parameter ranges of the extreme learning machine; the whale optimization algorithm influence parameters comprise population size, maximum iteration times and upper and lower limits of whale population positions; the fitness function of the whale optimization algorithm is the mean square error.
Obtaining a predicted value of the extreme learning machine by using main influence index data of the air conditioner cold load and whale population information; calculating fitness values of all individuals according to the predicted value of the extreme learning machine, selecting the current optimal fitness individual, and setting the position of the individual as the current optimal;
iteratively updating the individual position by utilizing a contraction surrounding mechanism, a spiral position updating mechanism and an exploration mechanism of a whale optimization algorithm;
after each iteration update, the optimized prediction precision parameters are transmitted to a limit learning machine;
judging whether the iteration cycle times reach a preset value, if so, stopping optimizing the prediction precision parameter to obtain an optimal weight parameter and a threshold value parameter of the extreme learning machine; if not, continuing to update iteratively; in the invention, an iterative updating process is carried out on the individual position by utilizing a contraction surrounding mechanism, a spiral updating position mechanism and an exploration mechanism of a whale optimizing algorithm, and if t is less than Tmax, the input weight and the hidden layer threshold are updated; when the random variable p is less than 0.5, if the coefficient vector |A| is more than or equal to 1, the whale will give up the hunting and search again; if the coefficient vector |A| <1, whale will attack the prey; when the random variable p is more than or equal to 0.5, the individual position is updated in a spiral way.
And step 3, acquiring main influence index data of the air conditioner cooling load, inputting the main influence index data into the optimized extreme learning machine, and outputting and obtaining an air conditioner cooling load prediction optimization result.
The invention also provides an air conditioner cold load prediction optimization system, which comprises an influence index data module, a model optimization module and an output module; the influence index data module is used for acquiring influence component data of the air conditioner cold load and determining main influence index data of the air conditioner cold load; the model optimization module is used for training and optimizing the limit learning machine network by adopting main influence index data of the air conditioner cold load; the training optimization process adopts a whale optimization algorithm to optimize weight parameters and threshold parameters of the extreme learning machine, so that the optimized extreme learning machine is obtained; and the output module is used for acquiring the main impact index data of the air conditioner cooling load, inputting the main impact index data into the optimized extreme learning machine, and outputting and obtaining the air conditioner cooling load prediction optimization result.
The invention also provides air conditioner cold load prediction optimizing equipment which comprises a processor, a memory and a computer program which is stored in the memory and can run on the processor. The steps in the air conditioner cold load prediction optimization method are realized when the processor executes the computer program. Or the processor realizes the functions of each module in the air conditioner cold load prediction optimizing equipment when executing the computer program.
The computer program may be divided into one or more modules, which are stored in the memory and executed by the processor to accomplish the present invention, for example. The one or more modules may be a series of computer program instruction segments capable of performing specific functions for describing the execution of the computer program in the air conditioner cold load prediction optimization device. For example, the computer program may be divided into an impact index data module, a model optimization module and an output module, where each module specifically functions as follows: the influence index data module is used for acquiring influence component data of the air conditioner cold load and determining main influence index data of the air conditioner cold load;
the model optimization module is used for training and optimizing the limit learning machine network by adopting main influence index data of the air conditioner cold load; the training optimization process adopts a whale optimization algorithm to optimize weight parameters and threshold parameters of the extreme learning machine, so that the optimized extreme learning machine is obtained;
and the output module is used for acquiring the main impact index data of the air conditioner cooling load, inputting the main impact index data into the optimized extreme learning machine, and outputting and obtaining the air conditioner cooling load prediction optimization result.
The air conditioner cold load prediction optimizing device can be computing devices such as a desktop computer, a notebook computer, a palm computer and a cloud server. The air conditioner cold load prediction optimization device may include, but is not limited to, a processor, a memory.
It will be appreciated by those skilled in the art that the foregoing is an example of an air conditioner cooling load prediction optimizing device, and does not constitute a limitation of the air conditioner cooling load prediction optimizing device, and may include more or fewer components, or combine some components, or different components, for example, the air conditioner cooling load prediction optimizing device may further include an input/output device, a network access device, a bus, and so on.
The processor may be a central processing unit (CentralProcessingUnit, CPU), other general purpose processors, digital signal processors (DigitalSignalProcessor, DSP), application specific integrated circuits (ApplicationSpecificIntegratedCircuit, ASIC), off-the-shelf programmable gate arrays (Field-ProgrammableGateArray, FPGA) or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components, or the like. The general-purpose processor may be a microprocessor or the processor may be any conventional processor or the like, which is a control center of the air conditioner cold load prediction optimizing apparatus, and various interfaces and lines are used to connect various parts of the entire air conditioner cold load prediction optimizing apparatus.
The memory may be used to store the computer program and/or module, and the processor may implement various functions of the air conditioner cooling load prediction optimizing device by running or executing the computer program and/or module stored in the memory and invoking data stored in the memory.
The memory may mainly include a storage program area and a storage data area, wherein the storage program area may store an operating system, an application program (such as a sound playing function, an image playing function, etc.) required for at least one function, and the like; the storage data area may store data (such as audio data, phonebook, etc.) created according to the use of the handset, etc. In addition, the memory may include high-speed random access memory, and may also include non-volatile memory, such as a hard disk, memory, plug-in hard disk, smart memory card (SmartMediaCard, SMC), secure digital (SecureDigital, SD) card, flash card (FlashCard), at least one magnetic disk storage device, flash memory device, or other volatile solid state storage device.
The module integrated with the air conditioner cooling load prediction optimizing device can be stored in a computer readable storage medium if implemented in the form of a software functional unit and sold or used as a separate product.
Based on such understanding, the present invention may implement all or part of the above-mentioned flow of the air conditioner cold load prediction optimization method, or may be implemented by instructing related hardware through a computer program, where the computer program may be stored in a computer readable storage medium, and the computer program may implement the steps of the air conditioner cold load prediction optimization method when executed by a processor. Wherein the computer program comprises computer program code which may be in source code form, object code form, executable file or some intermediate form etc.
The computer readable medium may include: any entity or device capable of carrying the computer program code, a recording medium, a U disk, a removable hard disk, a magnetic disk, an optical disk, a computer memory, a Read-only memory (ROM), a random access memory (RAM, randomAccessMemory), an electrical carrier signal, a telecommunication signal, a software distribution medium, and so forth.
It should be noted that the computer readable medium contains content that can be appropriately scaled according to the requirements of jurisdictions in which such content is subject to legislation and patent practice, such as in certain jurisdictions in which such content is subject to legislation and patent practice, the computer readable medium does not include electrical carrier signals and telecommunication signals.
Examples
As shown in fig. 1, two large commercial buildings in a city are taken as research objects, and the embodiment provides an air conditioner cooling load prediction optimization method, which comprises the following steps:
step 1, acquiring component data of an air conditioner cooling load; wherein, the components of the air conditioner cooling load comprise the indoor temperature of the building and CO 2 Concentration, total level radiation, outdoor air temperature, relative humidity, wet bulb temperature, and wind speed.
Step 2, preprocessing the influence component data of the air conditioner cold load by adopting a random forest algorithm to obtain main influence index data of the air conditioner cold load; the method comprises the steps of preprocessing the influence composition requirements of air conditioner cold load by adopting a random forest algorithm, selecting air conditioner cold load influence factor data by adopting a univariate selection mode, and establishing a random forest model by each air conditioner cold load influence factor data and an air conditioner cold load value; taking the determined coefficient as an evaluation index of the random forest model; the determining coefficient is the influence degree of the air conditioner cold load influence factor on the air conditioner cold load value; and screening the predicted characteristic data by adopting a model training mode of 5-fold cross validation to obtain main influence index data of the air conditioner cooling load.
As shown in fig. 2, the preprocessing process for the components of the air conditioner cooling load by adopting a random forest algorithm specifically comprises the following steps:
step 21, acquiring data of influence constituent elements of the air conditioner cold load, and constructing an initial training set; extracting N new sub-sample sets randomly and in a replaced mode by applying a boost method from an initial training set; and establishing N classification regression trees according to the N sub-sample sets.
Step 22, setting the feature dimension of each sub-sample set as M, and setting a constant M which is less than or equal to M; selecting m features at each node of each classification regression tree; and calculating the information content contained in each feature, and selecting one feature with the highest classification capability from m features to perform node splitting.
And step 23, making each classification regression tree perform sound field to the maximum extent, and having no pruning process.
Step 24, integrating all the classification regression trees into a random forest; when the classification problem is processed, the final output is determined at most according to a certain classification result number; when a regression prediction is established using a random forest, the final result is determined by the average of all tree outputs.
In this embodiment, the calculation formula of the determination coefficient is as follows:
wherein R is a determining coefficient, y i Is the true value of the response variable in the original data, y * Is the average value of the true values, f i N is the number of response variables, i is the ith response variable, which is the predicted value of the response variable.
In the embodiment, in the process of selecting the features, a model training mode of 5-fold cross validation is applied; decomposing the whole initial training set into 5 parts of sub-sample sets, alternately taking 4 parts of the sub-sample sets as training sets and 1 part of the sub-sample sets as test sets to train a model, finally determining coefficients for the output of 5 times of training, and taking an average value; the stability of the model can be improved, and the credibility of the feature selection result is increased; and sequencing the influence component data of the air conditioner cold load corresponding to the decision coefficient according to the size, and acquiring main influence index data of the air conditioner cold load.
Step 3, training and optimizing a limit learning machine network by adopting main influence index data of the air conditioner cold load; the training optimization process adopts a whale optimization algorithm to optimize weight parameters and threshold parameters of the extreme learning machine, so that the optimized extreme learning machine is obtained; adopting whale optimization algorithm to perform optimization process, specifically comprising the following steps:
step 31, setting prediction precision parameters of the extreme learning machine, and determining a prediction precision parameter range of the extreme learning machine; the prediction precision parameters of the extreme learning machine comprise input weight parameters and hidden layer threshold parameters; wherein, the input weight parameter is set as a random number in the range of-1 to 1; the hidden layer threshold parameter is set to a random number in the range of 0 to 1.
Step 32, setting optimization ranges of influence parameters of a whale optimization algorithm according to the prediction precision parameter ranges of the extreme learning machine; the whale optimization algorithm influence parameters comprise population size, maximum iteration times and upper and lower limits of whale population positions; the fitness function of the whale optimization algorithm is the mean square error.
Obtaining a predicted value of the extreme learning machine by using main influence index data of the air conditioner cold load and whale population information; and calculating the fitness value of all individuals according to the predicted value of the extreme learning machine, selecting the current optimal fitness individual, and setting the position of the individual as the current optimal.
Because the input weight of the extreme learning machine and the hidden layer threshold have randomly selected characteristics, in order to find the optimal parameters in the process of model establishment; in this embodiment, the whale optimization algorithm is selected to iteratively optimize the above two parameters, and a mean-square error (mean-square error) is set as a fitness function of the optimization algorithm; the smaller the fitness function value is, the more accurate the predicted value is, and the more optimal the obtained input weight and hidden layer threshold value of the extreme learning machine are; the mean square error calculation formula is as follows:
wherein y is ij Is true value, Y j J is the test sample number, which is the predicted value of the test sample.
Step 34, iteratively updating the individual position by utilizing a contraction surrounding mechanism, a spiral updating position mechanism and an exploration mechanism of a whale optimization algorithm; wherein, if the current iteration times t<Maximum number of iterations T max Then the input weight and the hidden layer threshold are further processedNew; when the random variable p<When the coefficient vector |A| is more than or equal to 1, the whale will give up the hunting and search again; if coefficient vector |A|<1, whale will attack prey; when the random variable p is more than or equal to 0.5, the individual position is updated in a spiral way.
And 35, after each iteration update, transmitting the optimized prediction precision parameters to a limit learning machine.
Step 36, judging whether the iteration cycle number reaches a preset value, if so, stopping optimizing the prediction precision parameter to obtain the optimal weight parameter and the prefabrication parameter of the extreme learning machine; if not, the iterative update continues.
The extreme learning machine is widely applied in the field of artificial intelligence, the traditional feedforward neural network is improved, and the output weight can be obtained through one-step calculation; meanwhile, the connection weight from the input layer to the hidden layer and the threshold value of the hidden layer can be randomly generated, and the adjustment is not needed after the setting is finished; meanwhile, the connection weight between the hidden layer and the output layer is not required to be adjusted iteratively, but is determined once by solving an equation set, so that the operation speed is faster than that of the traditional machine learning algorithm.
As shown in fig. 3, fig. 3 shows a network structure block diagram of the extreme learning machine in the embodiment, including an input layer, a single hidden layer and an output layer; assume that an input layer, a hidden layer, and an output layer are used for n, l, and m nodes, respectively; in N sample sets (x i ,y i ) Wherein x is i =[x 1i ,x 2i ,…,x ni ],y i =[y 1i ,y 2i ,…,y mi ]Where i=1, 2, …, N, the input matrix of the training samples is x= [ X 1 ,x 2 ,…,x N ]The output matrix is y= [ Y ] 1 ,y 2 ,…,y N ]The activation function of the hidden layer is f (x).
The mathematical model of the extreme learning machine is shown below,
the above can be simplified as:
Hβ=Y
wherein beta is i Output weights from the ith hidden layer node to the output neurons, w i B for inputting the input weights of the neurons to the ith hidden layer node i A threshold value of the i-th hidden layer node, y j The output value of the j training sample is H, which is the hidden layer output matrix of the extreme learning machine; the expression of the hidden layer output matrix of the extreme learning machine is as follows:
in the extreme learning machine, due to the input weight w i And threshold b of hidden layer i Is given randomly and is fixed, when the activation function f (x) and the neural network structure are determined, the output matrix H of the hidden layer is uniquely determined, and the output weight can be obtained by solving the least square solution of the linear equation set:
wherein H is + Moore-Penrose generalized inverse of the output matrix H of the hidden layer.
The whale optimization algorithm is a new meta-heuristic algorithm proposed by the Australian scholars Seyedali Mirjalili in 2016; after the prey is found, the whale first submerges into the bottom of the prey and then forms unique bubbles along a circular path; meanwhile, the whale is positioned from the upstream to the sea surface, and the prey is swallowed in a smaller range through bubbles; the whale optimization algorithm works in three parts: shrink wrap, bubble mesh hunting, and search for hunting.
In the shrink wrapping phase, the whale will first wrap around the prey, which can be described by the following equation:
D=|CX*(t)-X(t)|
X(t+1)=X*(t)-AD
wherein t is the number of current iterations, X (t) is the optimal position vector of the current whale, X (t) is the position vector of the current whale, and X (t+1) is the target position vector of the next iteration; A. c is a coefficient vector, and is defined as follows:
A=2ar-a;C=2r
where r is a random vector within the interval [0,1], and the value of a decreases linearly from 2 to 0.
In the development stage, two processes of a shrink wrapping mechanism and a spiral position updating mechanism are realized, and the mathematical model is as follows:
shrink wrapping mechanism: in this process, A is some random number between [ a, -a ], which decreases in value from 2 to 0 with the iterative process.
Spiral update location mechanism: in this mechanism, the whale approaches the prey in a spiral motion, and the simulation equation of this process is shown as follows:
X(t+1)=De bl cos(2πl)+X*(t)
where d= |cx (t) -X (t) | is the distance between whale and the current optimal position, and the constant b is used to characterize the shape of the spiral, and l is a random number in [ -1,1 ].
When hunting is performed, the whale performs the two predation strategies according to 50% probability:
where p represents a random variable between [0,1], the value of a is set to decrease as whale approaches the prey.
If |A| <1, whale will attack the prey; if |A| >1, whale will discard the prey and search again.
In the exploration stage, a random value |A| is set to be more than or equal to 1, and the mathematical model of the stage is as follows:
D=|CX rand -X(t)|
X(t+1)=X rand -AD
wherein X is rand Is a random agent location vector in the population.
The training and optimizing process of the extreme learning machine is as follows:
collecting the data of the cold load influence constituent elements of the large commercial building air conditioner, and selecting the parameters with high cold load correlation of the large commercial building air conditioner as the input variables of the optimized extreme learning machine by utilizing a random forest algorithm; meanwhile, in order to avoid error influence of difference of dimensions among data on an experimental model, normalization processing is carried out on the data samples.
Setting independent operation times u, randomly initializing a whale population 1 within the range of input weight and hidden layer threshold values of the extreme learning machine, and initializing a whale population 2 in a chaotic sequence mode.
And respectively calculating output layer weight matrixes of the two whale populations.
And calculating individual fitness values of the two whale populations, and recording optimal individuals X of the two whale populations respectively. For two whale populations, respectively calculating convergence factors and updating vector coefficients A; a whale individual performs a food search within the enclosure; and respectively updating the optimal individuals X of the two populations.
After the two whale populations independently run for u times, new immigration operators are adopted to exchange individuals of the two whale populations. And setting independent evolution algebra m, and continuously executing evolution operation according to an independent operation mode. Judging whether a termination condition is met, if so, calculating an optimized output layer weight, and evaluating a prediction effect by using test set sample data; otherwise, returning to the individual exchange step, and continuing to execute.
And 4, acquiring main influence index data of the air conditioner cooling load, inputting the main influence index data into the optimized extreme learning machine, and outputting and obtaining an air conditioner cooling load prediction optimization result.
Test results:
according to the embodiment, prediction model learning and testing are performed by using building cold load related data acquired by a public building energy consumption monitoring platform in a certain city.
As shown in fig. 4-5, the parameter influence coefficient diagrams of the large commercial buildings 1 and 2 are shown in fig. 4-5, and it can be seen from fig. 4-5 that 13 large commercial building cold load related parameters have different influence coefficients on cold load values, the eastern window wall ratio and the western window wall ratio have small difference, but the eastern window wall ratio influence coefficient is obviously higher than the western window wall ratio; while the eastern window wall ratio and the northbound window wall ratio of the large commercial building 2 are completely consistent, the eastern window wall ratio influence coefficient is still higher than the northbound. In addition, as can be seen from fig. 4-5, the related parameters such as load at the previous moment, outdoor dry bulb temperature, room personnel flow condition, window wall ratio, outdoor relative humidity, solar irradiance at the previous moment and the like have higher influence proportion on the cold load, and the influence of parameters such as customs, lighting use condition, solar irradiance and the like is smaller.
As shown in the accompanying drawings 6-7, the comparison situation of the large commercial building prediction result diagrams is shown in the accompanying drawings 6-7, and compared with the three prediction models of the support vector machine GRNN, the convolutional neural network RF-GRNN and the extreme learning machine PWOA-ELM before improvement and the difference between the two commercial building cold load prediction values obtained by the RF-PWOA-ELM in the embodiment and the real value is minimum, partial values almost completely coincide, and the fitting effect is better.
Meanwhile, in order to verify the model prediction effect, a root mean square error (Root Mean Square Error, RMSE) and an average absolute percentage error (Mean Absolute Percentage Error, MAPE) are selected as main evaluation indexes of the model prediction accuracy, and the formula is as follows:
as shown in fig. 8, a comparison chart of prediction accuracy in the present embodiment is shown in fig. 8, and it can be seen from fig. 8 that RMSE and MAPE of the rf-PWOA-ELM prediction model are 2.8735, 0.2% and 4.7721, 0.45% respectively, for the large commercial building 1 and the large commercial building 2, which are the smallest compared with other methods; in addition, as can be seen from fig. 8, compared with PWOA-ELM, two evaluation indexes of RMSE and MAPE of the RF-PWOA-ELM prediction model are respectively reduced by 5.621, 0.88% and 12.267 and 1.29%, and compared with GRNN, two evaluation indexes of the RF-GRNN prediction model are respectively reduced by 19.7248, 3.05% and 12.508 and 1.04%, so that it can be obtained that the random forest can well perform the dimension reduction processing on the cold load related parameters, and the selected input parameters can better establish the prediction model, thereby effectively increasing the model prediction precision.
The above embodiment is only one of the implementation manners capable of implementing the technical solution of the present invention, and the scope of the claimed invention is not limited to the embodiment, but also includes any changes, substitutions and other implementation manners easily recognized by those skilled in the art within the technical scope of the present invention.