Cooling capacity prediction method and system for heating, ventilation and air conditioning and related device

Abstract

A cooling capacity prediction method and system for heating, ventilation and air conditioning and a related device. The method comprises: by means of a denoising algorithm, decomposing an original signal into multiple intrinsic modes and a residual, the multiple intrinsic modes constituting an intrinsic mode function; using the Pearson correlation coefficient to classify the intrinsic modes into a first intrinsic mode dominated by an effective signal and a second intrinsic mode dominated by noise; decomposing the second intrinsic mode by means of wavelet soft thresholding denoising to obtain a third intrinsic mode; reconstructing the first intrinsic mode and the third intrinsic mode to finally obtain denoising data; taking the denoising data as an input of a preset neural network to finally obtain a cooling capacity prediction model, and using the cooling capacity prediction model to predict the cooling capacity of heating, ventilation and air conditioning to obtain a prediction result of the cooling capacity of the heating, ventilation and air conditioning. Compared with the prior art, the present invention solves the problems of nonlinearity, coupling, and long-term dependency in a cooling capacity prediction process, and increases the speed and accuracy of the cooling capacity prediction model.

Description

HVAC cooling capacity prediction method, system and related equipment Technical Field

[0001] The present invention relates to the field of air conditioning, and in particular to a method and system for predicting the cooling capacity of a heating, ventilation and air conditioning system and related equipment. Background Art

[0002] Global carbon emissions primarily come from three different sectors: industry, transportation, and construction. Analysis of United Nations climate change assessment reports indicates that the construction industry accounts for one-third of global energy consumption and contributes approximately 25% of global greenhouse gas emissions. Heating, ventilation, and air conditioning (HVAC) systems are widely used to create a comfortable working and living environment for people. Air conditioning systems contribute the largest portion of a building's annual energy consumption, accounting for approximately 40-50% of the total, and in some regions, even over 70%.

[0003] Existing air-conditioning systems have the problem of high energy consumption. Reducing the energy consumption of air-conditioning systems in public buildings is an important way to reduce the energy consumption of the entire building, and it also has important practical significance.

[0004] Therefore, it is necessary to develop a HVAC cooling capacity prediction method, system and related equipment that can accurately and quickly predict cooling load.

[0005] Summary of the Invention

[0006] The embodiments of the present invention provide a HVAC cooling capacity prediction method, system and related equipment, aiming to provide a HVAC cooling capacity prediction method and system that can accurately and quickly predict cooling load.

[0007] In a first aspect, the present invention provides a method for predicting cooling capacity of a heating, ventilation and air conditioning system, comprising the following steps:

[0008] S1. Decomposing the original signal into a plurality of eigenmodes and residuals by a noise reduction algorithm, wherein the plurality of eigenmodes constitute an eigenmode function;

[0009] S2. using the Pearson correlation coefficient to classify the eigenmode into a first eigenmode dominated by a valid signal and a second eigenmode dominated by noise;

[0010] S3, decomposing the second eigenmode by wavelet soft threshold noise reduction method to obtain a third eigenmode;

[0011] S4. Reconstructing the first eigenmode and the third eigenmode to obtain noise reduction data;

[0012] S5. Using the noise reduction data as input to a preset neural network to ultimately obtain a cooling capacity prediction model, and using the cooling capacity prediction model to predict HVAC cooling capacity to obtain a prediction result of the HVAC cooling capacity.

[0013] Preferably, in step S1, decomposing the original signal comprises the following steps:

[0014] S11, obtaining the minimum point and the maximum point of the original signal, then fitting the minimum point and the maximum point using a cubic spline envelope to obtain the upper envelope and the lower envelope of the original signal, and then calculating the mean envelope according to the upper envelope and the lower envelope;

[0015] S12. Calculate the difference function between the original signal and the mean envelope, define a first condition as the difference between the number of extreme points and the number of zero points of the original signal is 0 or 1, define a second condition as the upper envelope and the lower envelope are symmetrical about the time axis, and the mean envelope is 0 at any point in the signal domain; determine whether the difference function satisfies both the first and second conditions: if so, output a first-order intrinsic mode function; if not, repeat steps S11-S12 until the difference function satisfies both the first and second conditions;

[0016] S13, calculating the difference between the original signal and the first-order intrinsic mode function;

[0017] S14. Repeat the above steps S11-S13 to obtain the corresponding intrinsic mode function and the decomposed original signal.

[0018] Preferably, in step S1, the original signal satisfies the following relationship:

[0019] Among them, the IMF n represents the nth order intrinsic mode signal, residual represents the residual after decomposition, k represents the decomposition order, and X(t) represents the original signal.

[0020] Preferably, in step S2, the Pearson correlation coefficient r satisfies the following relationship:

[0021] in, represents the mean value of the original signal, S X represents the variance of the original signal, IMF i represents the i-th order eigenmode, represents the mean of the eigenmode, S IMF represents the standard deviation of the IMF.

[0022] Preferably, in step S3, the noise reduction algorithm is based on the CEEMADN algorithm.

[0023] Preferably, in step S3, the second eigenmode is decomposed by a wavelet soft threshold denoising method, which specifically includes the following sub-steps:

[0024] Decomposing the original signal by wavelet packet decomposition, using three or four wavelet decomposition layers to decompose the original signal into high-frequency coefficients and low-frequency coefficients;

[0025] Setting a threshold criterion and putting the high frequency coefficient into a soft threshold function for noise reduction processing;

[0026] Perform inverse wavelet transformation, reconstruct the processed high-frequency coefficients, and obtain the third eigenmode.

[0027] Preferably, the preset neural network includes a temporal convolutional network, a bidirectional long short-term memory neural network, and a sparse probabilistic self-attention mechanism network, wherein:

[0028] The temporal convolutional network is used to perform feature extraction on the noise reduction data to obtain a first feature;

[0029] The bidirectional long short-term memory neural network is used to process the time features, periodic patterns and periodic changes in the time series of the first features to obtain a second feature;

[0030] The sparse probabilistic self-attention mechanism network is used to capture and process the long-term dependency in the time series of the second feature to obtain a third feature, and use the third feature as the prediction result.

[0031] In a second aspect, the present invention further provides a HVAC cooling capacity prediction system, comprising the following modules:

[0032] A signal decomposition module decomposes the original signal into a plurality of eigenmodes and residuals through a noise reduction algorithm, wherein the plurality of eigenmodes constitute an eigenmode function;

[0033] a classification module, classifying the eigenmode into a first eigenmode dominated by a valid signal and a second eigenmode dominated by noise using a Pearson correlation coefficient;

[0034] A wavelet soft threshold decomposition module decomposes the second eigenmode by a wavelet soft threshold denoising method to obtain a third eigenmode;

[0035] A reconstruction module reconstructs the first eigenmode and the third eigenmode to obtain noise reduction data;

[0036] A neural network module is preset, and the noise reduction data is used as input of the preset neural network to finally obtain a cooling capacity prediction model, and the cooling capacity prediction model is used to predict the HVAC cooling capacity to obtain a prediction result of the HVAC cooling capacity.

[0037] In a third aspect, the present invention also provides a computer device comprising: a memory, a processor, and a computer program stored on the memory and executable on the processor, wherein when the processor executes the computer program, the steps of the HVAC cooling capacity prediction method as described in any one of the above embodiments are implemented.

[0038] In a fourth aspect, the present invention further provides a computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements the steps in the HVAC cooling capacity prediction method as described in any one of the above embodiments.

[0039] Compared with the existing technology, the HVAC cooling capacity prediction method, system and related equipment provided by the present invention provide a hybrid noise reduction algorithm that can remove noise to the greatest extent while retaining the effective signal of the original signal; at the same time, it provides a cooling capacity prediction model, which solves the problems of nonlinearity, coupling and long-term dependence in the prediction process, and improves the speed and accuracy of the prediction model; and can achieve high-precision prediction with adaptive data noise reduction while utilizing fewer hardware resources, so as to obtain a more accurate predicted cooling capacity at the next moment, and provide an accurate reference for subsequent chiller optimization strategies, fault diagnosis and demand-side management, thereby improving the HVAC operation efficiency, reducing energy consumption and carbon emissions while meeting the comfort of indoor occupants. BRIEF DESCRIPTION OF THE DRAWINGS

[0040] FIG1 is a flowchart of a method for predicting cooling capacity of a heating, ventilation and air conditioning unit according to an embodiment of the present invention;

[0041] FIG2 is a sub-flow diagram of a method for predicting cooling capacity of an HVAC system according to an embodiment of the present invention;

[0042] FIG3 is a schematic diagram of a time convolution structure of a HVAC cooling capacity prediction method provided by an embodiment of the present invention;

[0043] FIG4 is a schematic diagram of the structure of a long short-term memory neural network unit of a method for predicting cooling capacity of an HVAC system provided by an embodiment of the present invention;

[0044] FIG5 is a schematic diagram of a sparse probabilistic self-attention mechanism of a HVAC cooling capacity prediction method provided by an embodiment of the present invention;

[0045] FIG6 is a schematic diagram comparing prediction results of various models of a HVAC cooling capacity prediction method provided by an embodiment of the present invention;

[0046] FIG7 is a schematic diagram of a HVAC cooling capacity prediction system provided by an embodiment of the present invention;

[0047] FIG8 is a schematic diagram of a computer device for predicting HVAC cooling capacity provided by an embodiment of the present invention. DETAILED DESCRIPTION

[0048] The following will clearly and completely describe the technical solutions in the embodiments of the present invention in conjunction with the accompanying drawings. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without making creative efforts are within the scope of protection of the present invention.

[0049] (Example 1)

[0050] Please refer to FIG1 , which is a flowchart of a method for predicting cooling capacity of a heating, ventilation and air conditioning system according to an embodiment of the present invention. The method for predicting cooling capacity of a heating, ventilation and air conditioning system according to an embodiment of the present invention includes the following steps:

[0051] S1. Decomposing an original signal into a plurality of eigenmodes and residuals by a noise reduction algorithm, wherein the plurality of eigenmodes constitute an eigenmode function.

[0052] In this embodiment, the original signal is decomposed using a complete noise-assisted aggregated empirical mode decomposition (CEEMD) AN denoising algorithm. The complete noise-assisted aggregated empirical mode decomposition (CEEMD) denoising algorithm is a further improvement of the empirical mode decomposition (EMD) denoising algorithm and the ensemble empirical mode decomposition (EEMD) denoising algorithm. It can convert positive and negative white noise into adaptive noise, and the eigenmodes obtained at each stage are averaged to obtain the final eigenmode sequences. Compared with the empirical mode decomposition (EMD) denoising algorithm and the ensemble empirical mode decomposition (EEMD) denoising algorithm, the complete noise-assisted aggregated empirical mode decomposition (CEEMD) AN denoising algorithm solves the modal aliasing problem that exists in traditional eigenmode decomposition and greatly reduces the problem of residual white noise in the eigenmodes.

[0053] Please refer to FIG2 , which is a sub-flow chart of a method for predicting cooling capacity of HVAC provided by an embodiment of the present invention. The specific steps are as follows:

[0054] The process of decomposing the original signal by the complete noise-assisted aggregated empirical mode decomposition (CEEMDAN) denoising algorithm is as follows:

[0055] S11. Obtain the minimum point and the maximum point of the original signal, then use the cubic spline envelope to fit the minimum point and the maximum point to obtain the upper envelope and the lower envelope of the original signal, and then calculate the corresponding mean envelope based on the upper envelope and the lower envelope.

[0056] Specifically, the mean envelope A1(t) satisfies the following relationship:

[0057] Among them, E max (t) represents the upper envelope of the original signal, E min (t) represents the lower envelope of the original signal.

[0058] S12. Calculate the difference function between the original signal and the mean envelope, define the first condition as the difference between the number of extreme points and the number of zero points of the original signal is 0 or 1, and define the second condition as the upper envelope and the lower envelope are symmetrical about the time axis and the mean envelope at any point in the signal domain is 0, determine whether the difference function satisfies the first and second conditions at the same time, and if so, output the first-order intrinsic mode function; if not, repeat steps S11-S12 until the difference function satisfies the first and second conditions at the same time.

[0059] S13. Calculate the difference between the original signal and the first-order intrinsic mode function.

[0060] Specifically, the difference between the original signal and the first-order intrinsic mode function satisfies the following relationship: R(t)=X(t)-IMF1;

[0061] Where X(t) represents the original signal, IMF1 represents the first-order intrinsic mode function, and R(t) represents the difference.

[0062] S14. Repeat the above steps to obtain the corresponding intrinsic mode function and the decomposed original signal.

[0063] Specifically, the original signal satisfies the following relationship:

[0064] Among them, the IMF n The nth-order intrinsic mode signal, residual represents the residual after decomposition, and k represents the decomposition order.

[0065] S2. Using the Pearson correlation coefficient, classify the eigenmodes into a first eigenmode dominated by a valid signal and a second eigenmode dominated by noise.

[0066] In this embodiment, in order to divide the eigenmodes decomposed by the complete noise-assisted aggregated empirical mode decomposition (CEEMDAN) denoising algorithm into two categories: signal-dominated and noise-dominated, the Pearson correlation coefficient is used to calculate the correlation between the original signal and the obtained eigenmodes. The Pearson correlation coefficient satisfies the following relationship:

[0067] Wherein, r is the Pearson correlation coefficient, represents the mean value of the original signal, S X represents the variance of the original signal, IMF i The i-th eigenmode, represents the mean of the eigenmode, S IMF represents the standard deviation of the IMF.

[0068] S3. Decompose the second eigenmode using a wavelet soft threshold noise reduction method to obtain a third eigenmode.

[0069] In this embodiment, the specific steps of the process of performing wavelet soft threshold decomposition on the second eigenmode are as follows: performing wavelet packet decomposition on the original signal, using three or four layers of wavelet decomposition to decompose the original signal into high-frequency coefficients and low-frequency coefficients; setting a threshold criterion, placing the high-frequency coefficients into a soft threshold function for noise reduction processing, and keeping the low-frequency coefficients unchanged; performing an inverse wavelet transform, reconstructing the processed high-frequency coefficients, and obtaining the third eigenmode.

[0070] S4. Reconstruct the first eigenmode and the third eigenmode to obtain noise reduction data.

[0071] Specifically, Table [1] shows the performance of each denoising algorithm.

[0072] Table【1】Performance of each denoising algorithm

[0073] The meanings of the symbols in Table [1] are as follows:

[0074] CEEMDAN stands for Completely Noise-Assisted Aggregate Empirical Mode Decomposition Denoising Algorithm;

[0075] EMD stands for empirical mode decomposition denoising algorithm;

[0076] EEMD stands for Ensemble Empirical Mode Decomposition (EEMD) denoising algorithm;

[0077] WaveLet represents the wavelet soft threshold noise reduction algorithm;

[0078] CEEMDAN-WaveLe represents a hybrid algorithm that combines the fully noise-assisted aggregated empirical mode decomposition denoising algorithm with the wavelet soft threshold denoising algorithm.

[0079] CL represents the cooling characteristic data that needs to be processed for noise reduction;

[0080] Tem represents the temperature characteristic data that needs to be processed for noise reduction;

[0081] Rhum represents the humidity characteristic data that needs to be denoised.

[0082] As can be seen from Table 1, the hybrid denoising algorithm proposed in the present invention achieves the best results in terms of indicators such as signal-to-noise ratio, root mean square error, and Pearson correlation coefficient compared with other denoising algorithms. The signal-to-noise ratio is improved by at least 33.88%, the root mean square error is reduced by more than 38%, and the Pearson correlation coefficient is reduced by more than 0.2%.

[0083] This paper combines the advantages of the Complete Noise-Assisted Aggregation Empirical Mode Decomposition (CEEMDAN) denoising algorithm and the wavelet soft-threshold denoising algorithm (WaveLet). Compared with existing technologies, this method more effectively removes noise signals while preserving the original signal, solving the problem of high noise in collected industrial data and reducing the impact of high noise on subsequent input into the preset neural network.

[0084] S5. Using the noise reduction data as input to a preset neural network to ultimately obtain a cooling capacity prediction model, and using the cooling capacity prediction model to predict HVAC cooling capacity to obtain a prediction result of the HVAC cooling capacity.

[0085] In this embodiment, presetting the neural network includes the following steps:

[0086] Please refer to FIG. 3 , which is a schematic diagram of a time convolution structure of a HVAC cooling capacity prediction method provided by an embodiment of the present invention.

[0087] A temporal convolutional network is used to perform feature extraction on the noise reduction data to obtain a first feature.

[0088] Specifically, the temporal convolutional neural network introduces structures such as causal convolution, dilated convolution and residual connection on the basis of one-dimensional convolution, so that it can well capture time-related information when extracting relevant features. It also solves the problem of gradient explosion or disappearance in recursive models to a certain extent, and can also perform parallel calculations.

[0089] The causal convolution structure is a hierarchical structure formed by stacking single one-dimensional convolution layers from bottom to top. In causal convolution, the output of the current layer only operates on the elements of the previous layer at an earlier time. For example, there is a time series data x of length TT , take this sample as input and predict the output y corresponding to each moment based on this input T The modeling process can be expressed as follows: T =f(x0,…,x T );

[0090] To make the output y of the model T In the case of relying only on x0,…,x T Under the condition of the value of , we need to find a network f that minimizes the error between the actual value and the predicted output, but traditional convolutional networks are obviously unable to do this. Therefore, after some restrictions on the convolution operation, a causal convolution that considers the temporal order is formed. The relevant operation is shown in the first and second layer connections in Figure 3.

[0091] The input data enters from the bottom layer, and the time sequence is from left to right. During the convolution process, the output of the second layer at time t is only convolved with the input of the first layer at time t, t-1 and t-2, thus realizing the time sequence convolution.

[0092] Although causal convolution can effectively process time series data and solve the problem of future information "leaking" into the past caused by traditional convolution operations, when the time series data is too long and the network needs to remember more distant information, very deep convolution layers are required, which greatly increases the model complexity and makes training difficult. To solve this problem, the dilated convolution algorithm is used in conjunction with the expanded convolution algorithm.

[0093] Dilated convolution maintains the same kernel size and the number of parameters that need to be updated. Dilated convolution adds a dilation rate parameter to control the dilation size. For example, a 3x3 kernel with a dilation rate of 2 has a receptive field of 7x7. By increasing the dilation rate without changing the kernel size, the receptive field size exponentially increases, allowing the output of each convolution layer to contain a wider range of information.

[0094] As shown in the following formula, expanding the receptive field can be achieved by increasing the convolution kernel size k and increasing the dilation rate parameter d. The receptive field size of the nth layer can be calculated using the following formula.

[0095] Where k n is the size of the convolution kernel of the nth layer, The accumulated offset for each convolution kernel to expand outward.

[0096] The convolution operation on sequence elements is defined as follows:

[0097] At the same time, the expansion rate parameter d increases exponentially with the deepening of the network to ensure that very long-term information can be remembered in the case of deep networks, alleviating the training difficulties caused by the large amount of computation of causal convolution operations.

[0098] Through the above steps, the temporal convolutional neural network extracts relevant features of the denoised data to obtain a first feature.

[0099] Please refer to FIG4 , which is a schematic diagram of the structure of a long short-term memory neural network unit of a HVAC cooling capacity prediction method provided by an embodiment of the present invention;

[0100] Constructing a bidirectional long short-term memory neural network to process the time features, periodic patterns, and periodic changes in the time series of the first feature to obtain a second feature;

[0101] Specifically, the long short-term memory neural network (LSTM) can deeply explore the inherent laws in the time series of the first feature. The unit structure of the LSTM neural network is shown in Figure 4. It includes the forget gate f t , input gate i t and output gate o t . C t-1 to C t This line is called a storage unit and is used for historical data information of the previous moment and the current moment. t-1 is the state information of the previous time step, C t is the state information of the current time step, Indicates the updated value of the unit state; h t-1 is the hidden state of the previous time node, h t is the hidden state of the current node, x t is the input for the current time step.

[0102] The operation process of the LSTM unit of the long short-term memory neural network is divided into three stages. The first stage is the forgetting stage, which determines the C t-1 The degree of being forgotten satisfies the following relationship: t =σ(w f ·[h t-1 ,x t ]+b f );

[0103] f t C t-1 A coefficient that directly determines how much state information of the previous time step is forgotten; the second stage: input gate, determines the information x that needs to be remembered t and h t-1 , Used for storage unit to update the previous node status value C t-1 , while i t Used to control which features are used for update And satisfy the following relationship: i t =σ(w i ·[h t-1 ,x t ]+b i );

[0104] Finally, update the status information of the current time step to satisfy the following relationship:

[0105] The third stage is the output gate, which calculates the predicted value and generates the hidden state of the next time step, satisfying the following relationship: t =σ(w o ·[h t-1 ,x t ]+b o ); h t =o t *tanh(c t );

[0106] Where: σ is the Sigmoid activation function, and its function form satisfies the following relationship:

[0107] Tanh is the hyperbolic tangent function, and the function form satisfies the following relationship:

[0108] In the above relationship, * is the multiplication operation, · is the dot product, and w f 、w i and w o is x t The weight parameter, b f 、b i and b o Indicates h t The bias term.

[0109] The Long Short-Term Memory (LSTM) neural network uses a three-gate structure to control information input and output, allowing for selective information transmission. By connecting LSTM units in series, the LSTM neural network leverages the information filtering capabilities of each unit to learn the inherent time series relationships of data with temporal and periodic characteristics, enabling prediction of related data in the future. This process processes the first feature to generate the second feature.

[0110] A sparse probabilistic self-attention mechanism is used to capture the long-term dependency in the time series of the second feature to obtain a third feature, and the third feature is output as a prediction result.

[0111] Please refer to FIG5 , which is a schematic diagram of a sparse probabilistic self-attention mechanism of a HVAC cooling capacity prediction method provided by an embodiment of the present invention.

[0112] Specifically, the second feature is linearly transformed to obtain the query matrix Q, key-value pair matrix K and value matrix V, which are then split into multiple heads (subspaces) for calculations with different weights. Finally, they are spliced ​​together to obtain the final attention score, which improves computational efficiency while strengthening attention.

[0113] The sparse probabilistic self-attention mechanism evaluates the attention scores based on their sparsity index and sorts them in descending order. The determined number of samples is then used to select the required number of samples from the sparsity-descending sorted results before calculating the attention score. The attention mechanism allocates computing resources to more important tasks when computing power is limited, thereby addressing the problem of information overload.

[0114] Then, dot multiplication, scaling, normalization exponential function Softmax and final dot multiplication are performed to obtain the third feature, which is output as the prediction result.

[0115] The multi-head self-attention mechanism in the existing technology has a long-tail effect, which weakens the effect of the attention mechanism. The sparse probabilistic self-attention mechanism introduces the sampling principle into the attention mechanism, which solves the long-tail effect to a certain extent and reduces its computational complexity from o(L 2 ) is reduced to o(L ln L), shortening the training time.

[0116] Furthermore, a sparse probabilistic self-attention mechanism is used to capture long-term dependencies in time series, allowing the cooling capacity prediction model to focus on more important time steps. This improves the accuracy of the cooling capacity prediction model and provides a reference for energy conservation and carbon reduction efforts, such as optimal control of HVAC systems, fault diagnosis and diagnostic strategies, CO2 emission reduction, and demand-side management.

[0117] Please refer to FIG. 6 , which is a schematic diagram comparing prediction results of various models of a HVAC cooling capacity prediction method provided by an embodiment of the present invention.

[0118] Table 2 shows the performance of each prediction model indicator

[0119]

Table 2

[0120] The meanings of the symbols in Table [2] are as follows:

[0121] ANN (Artificial Neural Network) represents the artificial neural network model;

[0122] BiLSTM (Bi-directional Long Short-Term Memory) represents a bidirectional long short-term memory model;

[0123] BiLSTM-MHSA (Bi-directional Long Short-Term Memory + Multi-headed Self-attention) represents a model that combines a bidirectional long short-term memory model with a multi-headed self-attention mechanism.

[0124] BiLSTM-SPSA (Bi-directional Long Short-Term Memory + Sparse Probabilistic Self-Attention Mechanism) represents a bidirectional long short-term memory model combined with a sparse probabilistic self-attention mechanism;

[0125] CNN-BiLSTM-SPSA (Convolutional Neural Networks + Bi-directional Long Short-Term Memory + Sparse Probabilistic Self-Attention Mechanism) represents a model that combines convolutional neural networks, bidirectional long short-term memory models, and sparse probabilistic self-attention mechanisms.

[0126] TLS: represents the cooling capacity prediction model provided by the present invention.

[0127] RMSE stands for root mean square error, which measures the standard deviation of the difference between the true value and the predicted value;

[0128] MAE: stands for mean absolute error, which is used to evaluate the absolute difference between the true value and the predicted value;

[0129] MAPE: stands for Absolute Percent Error, an indicator used to measure the percentage of deviation between the true value and the predicted value;

[0130] R2: represents the coefficient of determination, which is used to measure the degree of fit between the true value curve and the predicted value curve.

[0131] The prediction results of the prediction model proposed in the embodiment of the present invention are compared with those of the commonly used algorithms. The relevant index values ​​are shown in Table 2. Compared with the commonly used algorithms, the prediction model proposed in the embodiment of the present invention reduces the MAE, MAPE and RMSE by 28.89%-62.48% respectively, while R2 increases by 2%-18.70%. After the introduction of the sparse probabilistic self-attention mechanism, the prediction performance remains basically unchanged and the training time is shortened by 3.5%. The prediction results are shown in Figure 6. It can be seen from the figure that the proposed TLS multivariate prediction model has the best prediction effect, and can achieve effective high-precision prediction at both low and high points. Although the prediction effect of CNN-BiLSTM-SPSA is slightly better than the model proposed in the present invention at some low points, its high-point prediction effect is still far behind that of the proposed TLS prediction model. Other prediction models can predict their basic trends, but the prediction accuracy needs to be improved.

[0132] Compared with the existing technology, the HVAC cooling capacity prediction method, system and related equipment provided by the present invention provide a hybrid noise reduction algorithm that can remove noise to the greatest extent while retaining the effective signal of the original signal; at the same time, it provides a cooling capacity prediction model, which solves the problems of nonlinearity, coupling and long-term dependence in the prediction process, and provides the speed and accuracy of the prediction model; and can achieve high-precision prediction with adaptive data noise reduction while utilizing fewer hardware resources, so as to obtain a more accurate predicted cooling capacity at the next moment, and provide an accurate reference for subsequent chiller optimization strategies, fault diagnosis and demand-side management, thereby improving the HVAC operation efficiency, reducing energy consumption and carbon emissions while meeting the comfort of indoor occupants.

[0133] (Example 2)

[0134] An embodiment of the present invention further provides a HVAC cooling capacity prediction system. Please refer to FIG7 , which is a schematic diagram of the structure of the HVAC cooling capacity prediction system 200 provided by an embodiment of the present invention, which includes:

[0135] 201. A signal decomposition module decomposes an original signal into a plurality of eigenmodes and residuals by using a noise reduction algorithm, wherein the plurality of eigenmodes constitute an eigenmode function.

[0136] 202. A classification module classifies the eigenmode into a first eigenmode dominated by a valid signal and a second eigenmode dominated by noise using a Pearson correlation coefficient.

[0137] 203. A wavelet soft threshold decomposition module performs wavelet soft threshold decomposition on the second eigenmode to obtain a third eigenmode.

[0138] 204. A reconstruction module reconstructs the first eigenmode and the third eigenmode to finally obtain denoised data.

[0139] 205. Preset a neural network module, use the noise reduction data as input of the preset neural network, and finally obtain a cooling capacity prediction model. Use the cooling capacity prediction model to predict the HVAC cooling capacity to obtain a prediction result of the HVAC cooling capacity.

[0140] The system 200 can implement the steps in the HVAC cooling capacity prediction method in the above embodiment and can achieve the same technical effects. Please refer to the description in the above embodiment and will not be repeated here.

[0141] (Example 3)

[0142] An embodiment of the present invention further provides a computer device. Please refer to Figure 8, which is a structural diagram of the computer device provided by an embodiment of the present invention. The computer device 300 includes: a memory 302, a processor 301, and a computer program stored in the memory 302 and executable on the processor 301.

[0143] The processor 301 calls the computer program stored in the memory 302 to execute the steps of the HVAC cooling capacity prediction method provided by the embodiment of the present invention. Referring to FIG8 , the method specifically includes the following steps:

[0144] S1. Decomposing the original signal into a plurality of eigenmodes and residuals by a noise reduction algorithm, wherein the plurality of eigenmodes constitute an eigenmode function;

[0145] S2. using the Pearson correlation coefficient to classify the eigenmode into a first eigenmode dominated by a valid signal and a second eigenmode dominated by noise;

[0146] S3, decomposing the second eigenmode by wavelet soft threshold noise reduction method to obtain a third eigenmode;

[0147] S4. Reconstructing the first eigenmode and the third eigenmode to obtain noise reduction data.

[0148] S5. Using the noise reduction data as input to a preset neural network to ultimately obtain a cooling capacity prediction model, and using the cooling capacity prediction model to predict HVAC cooling capacity to obtain a prediction result of the HVAC cooling capacity.

[0149] The computer device 300 provided in the embodiment of the present invention can implement the steps in the HVAC cooling capacity prediction method in the above embodiment and can achieve the same technical effects. Please refer to the description in the above embodiment and will not be repeated here.

[0150] (Example 4)

[0151] An embodiment of the present invention also provides a computer-readable storage medium, on which a computer program is stored. When the computer program is executed by a processor, the various processes and steps in the HVAC cooling capacity prediction method provided in the embodiment of the present invention are implemented, and the same technical effects can be achieved. To avoid repetition, they will not be described here.

[0152] Those skilled in the art will appreciate that all or part of the processes in the above-described method embodiments can be implemented by instructing related hardware through a computer program. The program can be stored in a computer-readable storage medium, and when executed, the program can include the processes in the above-described method embodiments. The storage medium can be a magnetic disk, an optical disk, a read-only memory (ROM), or a random access memory (RAM).

[0153] It should be noted that, in this document, the terms "comprises," "includes," or any other variations thereof are intended to encompass non-exclusive inclusion, such that a process, method, article, or apparatus comprising a series of elements includes not only those elements but also other elements not explicitly listed, or elements inherent to such process, method, article, or apparatus. In the absence of further limitations, an element defined by the phrase "comprising a ..." does not exclude the presence of other identical elements in the process, method, article, or apparatus comprising the element.

[0154] Through the description of the above embodiments, those skilled in the art can clearly understand that the above-mentioned embodiment methods can be implemented by means of software plus the necessary general hardware platform, and of course can also be implemented by hardware, but in many cases the former is a better embodiment. Based on this understanding, the technical solution of the present invention, or the part that contributes to the prior art, can be embodied in the form of a software product, which is stored in a storage medium (such as ROM / RAM, magnetic disk, optical disk), and includes a number of instructions for enabling a terminal (which can be a mobile phone, computer, server, air conditioner, or network device, etc.) to execute the methods described in each embodiment of the present invention.

[0155] The embodiments of the present invention are described above in conjunction with the accompanying drawings. What is disclosed is only a preferred embodiment of the present invention. However, the present invention is not limited to the above-mentioned specific implementation methods. The above-mentioned specific implementation methods are merely illustrative and not restrictive. Under the guidance of the present invention, ordinary technicians in this field can also make many forms and equivalent changes without departing from the scope of protection of the purpose of the present invention and the claims, which are all within the protection of the present invention.

Claims

1. A method for predicting cooling capacity of HVAC, characterized in that: The following steps are involved: S1. Decomposing the original signal into a plurality of eigenmodes and residuals by a noise reduction algorithm, wherein the plurality of eigenmodes constitute an eigenmode function; S2. using the Pearson correlation coefficient to classify the eigenmode into a first eigenmode dominated by a valid signal and a second eigenmode dominated by noise; S3, decomposing the second eigenmode by wavelet soft threshold noise reduction method to obtain a third eigenmode; S4, reconstructing the first eigenmode and the third eigenmode to obtain noise reduction data; S5. Using the noise reduction data as input of a preset neural network to finally obtain a cooling capacity prediction model, and using the cooling capacity prediction model to predict the cooling capacity of the HVAC to obtain a prediction result of the cooling capacity of the HVAC.

2. The HVAC cooling capacity prediction method according to claim 1, characterized in that: In step S1, decomposing the original signal comprises the following steps: S11, obtaining the minimum point and the maximum point of the original signal, and then fitting the minimum point and the maximum point with a cubic spline envelope to obtain the upper envelope and the lower envelope of the original signal, and then calculating the mean envelope according to the upper envelope and the lower envelope; S12, calculate the difference function between the original signal and the mean envelope, define the first condition as the difference between the number of extreme points and the number of zero points of the original signal is 0 or 1, define the second condition as the upper envelope and the lower envelope are symmetrical about the time axis, and the mean envelope at any point in the signal domain is 0; determine whether the difference function satisfies the first condition and the second condition at the same time: if so, output a first-order intrinsic mode function; if not, repeat steps S11-S12 until the difference function satisfies the first condition and the second condition at the same time; S13, calculating the difference between the original signal and the first-order intrinsic mode function; S14, repeating the above steps S11-S13 to obtain the corresponding intrinsic mode function and the decomposed original signal.

3. The HVAC cooling capacity prediction method according to claim 2, characterized in that: In step S1, the original signal satisfies the following relationship: Among them, IMF n represents the nth order intrinsic mode signal, residual represents the residual after decomposition, k represents the decomposition order, and X(t) represents the original signal.

4. The HVAC cooling capacity prediction method according to claim 3, characterized in that: In step S2, the Pearson correlation coefficient r satisfies the following relationship: in, represents the mean value of the original signal, S X represents the variance of the original signal, IMF i represents the i-th order eigenmode, represents the mean of the eigenmode, S IMF represents the standard deviation of the IMF.

5. The HVAC cooling capacity prediction method according to claim 1, characterized in that: The noise reduction algorithm is based on the CEEMADN algorithm.

6. The HVAC cooling capacity prediction method according to claim 1, characterized in that: In the step S3, the second eigenmode is decomposed by wavelet soft threshold denoising method, which specifically includes the following sub-steps: Decomposing the original signal by wavelet packet, using three or four wavelet decomposition layers to decompose the original signal into high-frequency coefficients and low-frequency coefficients; Setting a threshold criterion, and placing the high frequency coefficients into a soft threshold function for noise reduction processing; Perform inverse wavelet transformation, reconstruct the processed high-frequency coefficients, and obtain the third eigenmode.

7. The HVAC cooling capacity prediction method according to claim 1, characterized in that: The preset neural network includes a temporal convolutional network, a bidirectional long short-term memory neural network, and a sparse probabilistic self-attention mechanism network, wherein: The temporal convolutional network is used to extract features from the noise reduction data to obtain a first feature; The bidirectional long short-term memory neural network is used to process the time features, periodic patterns and periodic changes in the time series of the first feature to obtain a second feature; The sparse probabilistic self-attention mechanism network is used to capture and process the long-term dependency in the time series of the second feature to obtain a third feature, and use the third feature as the prediction result.

8. A HVAC cooling capacity prediction system, characterized in that: Includes the following modules: A signal decomposition module decomposes the original signal into a plurality of eigenmodes and residuals through a noise reduction algorithm, wherein the plurality of eigenmodes constitute an eigenmode function; A classification module, using a Pearson correlation coefficient to classify the eigenmode into a first eigenmode dominated by a valid signal and a second eigenmode dominated by noise; A wavelet soft threshold decomposition module performs wavelet soft threshold decomposition on the second eigenmode to obtain a third eigenmode; A reconstruction module, reconstructing the first eigenmode and the third eigenmode to obtain denoised data; A neural network module is preset, and the noise reduction data is used as an input of the preset neural network to finally obtain a cooling capacity prediction model, and the cooling capacity prediction model is used to predict the HVAC cooling capacity to obtain a prediction result of the HVAC cooling capacity.

9. A computer device, characterized in that: include: A memory, a processor, and a HVAC cooling capacity prediction program stored in the memory and executable on the processor, wherein the processor implements the steps of the HVAC cooling capacity prediction method as described in any one of claims 1 to 7 when executing the HVAC cooling capacity prediction program.

10. A computer-readable storage medium, characterized in that: The computer-readable storage medium stores a HVAC cooling capacity prediction program, and when the HVAC cooling capacity prediction program is executed by the processor, the steps in the HVAC cooling capacity prediction method according to any one of claims 1 to 7 are implemented.

Images