Electric signal optimization method of air conditioner electric control system

Abstract

The invention provides an electric signal optimization method for an air conditioner electric control system, and relates to the technical field of air conditioner electric signal optimization, and the method comprises the steps: obtaining an original electric control signal of an air conditioner, and decomposing the original electric control signal into different frequency bands through wavelet packet decomposition to obtain a plurality of sub-band signals; acquiring the sliding energy entropy of each sub-band signal; comparing the sliding energy entropy of each sub-band signal with an entropy threshold value to determine a corresponding denoising algorithm, and outputting the denoised sub-band signal; and carrying out weighted reconstruction on the de-noised sub-band signal, calculating a real-time residual evaluation value of the sub-band signal, comparing the real-time residual evaluation value with a residual threshold, and determining that the reconstructed sub-band signal is output as an output signal or a correction parameter is obtained according to a comparison result for analysis and optimization. According to the method, through dynamic decomposition and feedback closed loop, mapping logic of energy entropy and a multi-modal algorithm and SNR self-adaptive adjustment of weight fusion, the problems of inaccuracy and interference of an existing method on a denoised air conditioner electric signal are solved.

Description

Technical Field

[0001] The present invention belongs to the technical field of air conditioner electrical signal optimization, and particularly relates to an electrical signal optimization method for an air conditioner electronic control system. Background Art

[0002] An air conditioner intelligent electronic control system is an air conditioner control system based on electronic technology and intelligent control technology. By intelligently adjusting parameters such as the temperature, humidity, and wind speed of the air conditioner, comfortable, energy-saving, and environmentally friendly air conditioner control is achieved. Whether it is the normal operation of the air conditioner indoor unit or outdoor unit, and the coordinated control between the electronic control systems, they are all realized by the transmission of electrical signals between each module, such as temperature sensor signals, humidity sensor signals, wind speed sensor signals, power supply signals, control signals, fault signals, remote control signals, etc.

[0003] However, due to electromagnetic interference between various electronic components, sensor pulse fluctuations, voltage instability, etc., electrical signals are extremely vulnerable to noise interference, resulting in errors in the intelligent control and adjustment of the air conditioner. Most of the existing technologies use traditional wavelet decomposition technology to divide the signal into different components, and then only process the part identified as noise. However, the reality is that noise may penetrate into each component. In particular, the high-frequency part may contain useful signals, while the low-frequency part may have noise residues, and the electrical signals cannot be completely and accurately denoised, resulting in inaccurate denoised electrical signals, and further leading to inaccurate regulation of the air conditioner intelligent electronic control system. Summary of the Invention

[0004] In view of the problems raised in the background art, the present invention proposes an electrical signal optimization method for an air conditioner electronic control system.

[0005] To achieve this purpose, the present invention adopts the following technical solutions: An electrical signal optimization method for an air conditioner electronic control system, comprising: Step A: Obtain the original electrical control signal of the air conditioner, and decompose the original electrical control signal into different frequency bands through wavelet packet decomposition to obtain a plurality of sub-band signals; Step B: Obtain the sliding energy entropy of each sub-band signal; Step C: Compare the sliding energy entropy of each sub-band signal with an entropy value threshold to determine the corresponding denoising algorithm, and output each denoised sub-band signal; Step D: Perform weighted reconstruction on the denoised sub-band signals to obtain reconstructed sub-band signals; Step E: Calculate the real-time residual evaluation value for the reconstructed sub-band signal, compare the real-time residual evaluation value with the residual threshold, and determine whether to output the sub-band signal reconstructed in Step D as the final output signal or obtain the correction parameters and optimize Steps A to C based on the correction parameters according to the comparison result. The correction parameters include the new initial decomposition layer number for wavelet packet decomposition after optimization and the new entropy value threshold after optimization.

[0006] Preferably, Step A specifically includes: Determine the initial wavelet packet initial decomposition layer number according to Formula 1: — Formula 1; N represents the initial decomposition layer number; floor() represents the floor function; fs represents the sampling frequency of the original electric control signal; represents the preset frequency resolution, and the minimum frequency band width of the original electric control signal obtained after N-layer decomposition should be greater than the preset frequency resolution.

[0007] Preferably, Step A further includes: Construct a dynamic decomposition tree according to the initial decomposition layer number N, and bisect the frequency band width of the previous layer during each layer of decomposition, so that at the last layer of decomposition, the corresponding number of frequency bands are obtained with the minimum frequency band width, and the original electric control signal is decomposed into sub-band signals of the corresponding number of frequency bands, including: First layer decomposition: Take the sampling frequency of the original electric control signal as the initial frequency band width and perform average bisection to obtain the frequency band widths of the sub-bands in the first layer, and the frequency band width of each sub-band is HZ; Second layer decomposition: Bisect the frequency band widths of the sub-bands in the first layer respectively to obtain the frequency band widths of the sub-bands in the second layer, and the frequency band width of each sub-band is HZ; And so on, decompose successively until the Nth layer of decomposition; Nth layer decomposition: Bisect the frequency band widths of the sub-bands obtained by the (N - 1)th layer of decomposition respectively to obtain the frequency band widths of the sub-bands in the Nth layer, and the frequency band width of each sub-band is HZ, so as to decompose the original electric control signal into HZ with the minimum frequency band width and decompose it into sub-band signals; Obtain the wavelet packet coefficient value of each sub-band signal at time and record it as the sub-band signal amplitude of each sub-band signal 。

[0008] Preferably, step B specifically includes: Calculating the energy integral of each sub-band according to Formula 2: — Formula 2; represents the energy of the K-th sub-band at time t; K represents the sub-band number, used to identify the frequency band position to be analyzed; t represents the current time moment, used to define the end point of the sliding window, t > T; T represents the sliding window width, used to determine the time range of energy statistics, determined according to the air conditioner control period; represents the time point integration variable within the interval [t - T, t], used to traverse all time points within the window; represents the sub-band signal amplitude, obtained by wavelet decomposition of the original electronic control signal in step A.

[0009] Preferably, step B specifically includes: Calculating the energy probability distribution of each sub-band according to Formula 3: — Formula 3; represents the energy probability distribution of the K-th sub-band, used to reflect the weight of this frequency band in the total energy; M represents the total number of sub-bands, determined by the initial decomposition level N in step A, that is, M is the ; j represents the sub-band index, used to traverse all sub-bands; represents the energy of the j-th sub-band at time t; represents the energy of the K-th sub-band at time t.

[0010] Preferably, step B specifically includes: Calculating the sliding energy entropy of each sub-band according to Formula 4: — Formula 4; represents the sliding energy entropy of the K-th sub-band, used to quantify the signal complexity, ; When tends to 1, it indicates that the energy of each sub-band is evenly distributed, indicating that the original electronic control signal belongs to a steady-state signal; When When it tends to 0, it indicates that the energy is concentrated in a few sub-bands, indicating that there are pulse interferences in the original electric control signal.

[0011] Preferably, the step C includes: When , use the morphological filtering algorithm to denoise the current sub-band and output the denoised sub-band signal; When , use the wavelet threshold algorithm to denoise the current sub-band and output the denoised sub-band signal; When , use the LMS adaptive filtering algorithm to denoise the current sub-band and output the denoised sub-band signal; and represents the entropy value threshold, calibrated by the noise samples, , .

[0012] Preferably, the step D specifically includes: It includes using Formula Five and Formula Six to perform weighted reconstruction on each sub-band signal: —Formula Five; —Formula Six; represents the sub-band signal after denoising and reconstruction; represents the Kth sub-band signal after denoising, obtained by denoising the current sub-band signal using the corresponding denoising algorithm in step C; represents the dynamic weight coefficient, used to reflect the credibility of the sub-band signal; exp() represents the natural exponential function; represents the weight adjustment factor, used to reflect the balance between the number of sub-bands and the sensitivity; represents the real-time sub-band signal-to-noise ratio, used to quantify the quality of the sub-band signal; represents the signal-to-noise ratio reference threshold, set according to the noise floor of the air-conditioning system.

[0013] Preferably, the step E specifically includes: Calculate the real-time residual evaluation value according to Formula Seven: —Formula Seven; represents the real-time residual evaluation value; represents the amplitude error weight coefficient, used to control the contribution weight of the static deviation; Denotes the differential error weight coefficient, which is used to control the contribution weight of dynamic changes; Denotes the original electronic control signal; Denotes the target signal after weighted reconstruction in step D; Denotes the first derivative of the original electronic control signal at time t, which is used to represent the instantaneous change rate of the signal.

[0014] Preferably, step D further includes: Comparing the real-time residual evaluation value with the residual threshold: When , the sub-band signal after weighted reconstruction in step D is used as the final output signal for output; When , it includes: Optimizing the initial decomposition layer number N according to formula eight, —Formula eight; Denotes the optimized decomposition layer number; sgn() represents the sign function. When , When , When , ; , denotes the derivative of the real-time residual evaluation value R(t) at time t, which is used to reflect the change trend of the real-time residual evaluation value; Optimizing the entropy threshold according to formula nine, —Formula nine; Denotes the optimized entropy threshold, and the entropy threshold for the first optimization is artificially calibrated based on the entropy threshold calibrated from noise samples at the initial stage. The optimized entropy threshold after the first time is calculated by formula nine; Denotes the learning rate; Denotes the difference between the real-time residual evaluation values before and after optimization; Denotes the difference between the entropy thresholds before and after optimization; Using the optimized decomposition layer number as the new decomposition layer number N and passing it back to step A to construct a dynamic decomposition tree using the new decomposition layer number N; Using the optimized entropy threshold As a new entropy threshold value Feed back to step C and use the new entropy threshold value To expand the filtering range of the morphological filtering algorithm.

[0015] Advantages of the present invention over the prior art: 1. By using the residual analysis in step E to dynamically adjust and optimize the decomposition level in step A, the present invention solves the problem of frequency band aliasing caused by the fixed-level decomposition of the traditional wavelet packet decomposition method, optimizes the entropy threshold value in step C, and calibrates the decision threshold in real time to avoid overfitting or underfitting problems caused by fixed parameters, thereby realizing a closed-loop of dynamic decomposition and feedback; 2. By quantifying the noise characteristics with the entropy value in step B, the present invention realizes the precise matching between the noise characteristics and the denoising algorithm; 3. By using the exponential weight function in step D, the present invention ensures that the subbands with high signal-to-noise ratio dominate the reconstruction result and suppresses the noise influence of the low-quality subbands. Description of the drawings

[0016] Figure 1 is a flowchart of the method for optimizing the electrical signal of the air conditioner electronic control system of the present invention. Specific embodiments

[0017] The technical solution of the present invention will be further described below with reference to the drawings and specific embodiments.

[0018] In order to enable those skilled in the art to better understand the solution of the present invention, the technical solutions in the embodiments of the present invention will be clearly and completely described below with reference to the drawings in the embodiments of the present invention. Obviously, the described embodiments are only a part of the embodiments of the present invention, rather than all the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those of ordinary skill in the art without making creative efforts shall fall within the protection scope of the present invention.

[0019] The terms "first", "second", etc. in the description and claims of the present invention and the above drawings are used to distinguish different objects, rather than to describe a specific order. In addition, the terms "comprising" and "having" and any variations thereof are intended to cover non-exclusive inclusions. For example, a process, method, device, product or terminal that includes a series of steps or units is not limited to the listed steps or units, but optionally further includes steps or units not listed, or optionally further includes other steps or units inherent to these processes, methods, products or terminals.

[0020] References to "embodiments" in this specification mean that specific features, structures, or characteristics described in connection with the embodiments can be included in at least one embodiment of the invention. The phrase appears in various places in the specification and does not necessarily refer to the same embodiment, nor is it an independent or alternative embodiment mutually exclusive with other embodiments. Those skilled in the art will explicitly and implicitly understand that the embodiments described herein can be combined with other embodiments.

[0021] This application proposes an electrical signal optimization method for an air conditioner's electrical control system, including: Step A: Obtain the original electrical control signal of the air conditioner, and decompose the original electrical control signal into different frequency bands through wavelet packet decomposition to obtain multiple sub-band signals; Preferably, step A specifically includes: Determine the initial wavelet packet initial decomposition level according to formula one: —Formula one; N represents the initial decomposition level; floor() represents the floor function; fs represents the sampling frequency of the original electrical control signal; Represents the preset frequency resolution. After N-level decomposition, the minimum frequency band width of the original electrical control signal should be greater than the preset frequency resolution.

[0022] Construct a dynamic decomposition tree according to the initial decomposition level N. At each level of decomposition, divide the frequency band width of the previous level in half, so that at the last level of decomposition, the corresponding number of frequency bands are obtained with the minimum frequency band width, and the original electrical control signal is decomposed into sub-band signals of the corresponding number of frequency bands, including: First-level decomposition: Take the sampling frequency of the original electrical control signal as the initial frequency band width and divide it evenly in half to obtain the frequency band width of the sub-bands in the first layer, and the frequency band width of each sub-band is HZ; Second-level decomposition: Divide the frequency band widths of the sub-bands in the first layer evenly in half respectively to obtain the frequency band widths of the sub-bands in the second layer, and the frequency band width of each sub-band is HZ; And so on, decompose successively until the Nth-level decomposition; Nth-level decomposition: Divide the frequency band widths of the sub-bands obtained from the (N-1)th-level decomposition evenly in half respectively to obtain the frequency band widths of the sub-bands in the Nth layer, and the frequency band width of each sub-band is HZ, so as to decompose the original electrical control signal according to the minimum frequency band width HZ into Sub-band signal Obtain the wavelet packet coefficient value of each sub-band signal at time and record it as the sub-band signal amplitude of each sub-band signal .

[0023] In this embodiment, the purpose of decomposing the original electronic control signal into different frequency bands is to provide an analysis basis for subsequent noise identification. For example, when the air conditioner indoor unit controller detects abnormal fluctuations in the temperature sensor signal, the following interferences may exist: Periodic electromagnetic interference, such as 50HZ power frequency noise; Random pulse interference, such as caused by the start and stop of the compressor; High-frequency noise, such as inverter switching noise.

[0024] Then this method needs to perform wavelet packet decomposition on the original electronic control signal to obtain sub-band signals in different frequency bands, so as to provide a basis for the analysis of each subsequent sub-band signal. Compared with the existing wavelet packet decomposition, the decomposition layer number is a fixed number set artificially. This method calculates the initial decomposition layer number using the sampling frequency and the preset frequency resolution, and at the same time corrects the initial decomposition layer number in the dynamic adjustment of the residual analysis in step E, avoiding frequency band aliasing caused by the fixed decomposition number set artificially. For example, the fixed decomposition layer number may cause the 50HZ power frequency interference to leak into adjacent sub-bands; Specifically, taking the sampling frequency fs of 200HZ as an example, the preset frequency resolution is the accuracy requirement of 2HZ, then the initial decomposition layer number layers, so that at the sampling frequency of 200HZ, after each layer of decomposition bisects the frequency band of the previous layer, the minimum frequency band width obtained after the 6th layer of decomposition is , meeting the accuracy requirement of being greater than the preset frequency resolution .

[0025] Furthermore, select the db4 wavelet basis to construct a dynamic decomposition tree. At each layer of decomposition, bisect the frequency band width of the previous layer, so that at the last layer of decomposition, the corresponding number of frequency bands are obtained with the minimum frequency band width, and the original electronic control signal is decomposed into sub-band signals in the corresponding number of frequency bands. For example: The first layer of decomposition: Take the sampling frequency of 200HZ as the initial frequency band width and bisect it evenly to obtain 2 sub-band frequency band widths. The frequency band width of the first sub-band is 0 - 100HZ, and the frequency band width of the second sub-band is 100 - 200HZ; Second - layer decomposition: The frequency band widths of the first sub - band (0 - 100 HZ) and the second sub - band of the first layer are respectively averaged and divided into two equal parts, obtaining 4 sub - band frequency widths. The frequency band width of the first sub - band is 0 - 50 HZ, the frequency band width of the second sub - band is 50 - 100 HZ, the frequency band width of the third sub - band is 100 - 150 HZ, and the frequency band width of the fourth sub - band is 150 - 200 HZ; And so on, until the 6 - th layer of decomposition, obtaining a total of sub - band signals. The frequency band width of the first sub - band signal is 0 - 3.125 HZ, the frequency band width of the second sub - band is 3.125 - 6.25 HZ, ……, the frequency band width of the 31 - st sub - band is 93.75 - 96.875 HZ …… the frequency band width of the 64 - th sub - band is 196.875 - 200 HZ; Then we decompose the original electrical control signal S(t) into 64 sub - band signals according to the frequency band widths of different sub - bands. In different sub - bands, there may be interference signals and real signals. For example, the first sub - band 0 - 3.125 HZ may contain the real low - frequency signal of temperature change 0.23 HZ, and the 31 - st sub - band 93.75 - 96.875 HZ may contain the compressor interference signal 95.67 HZ.

[0026] Obtain the wavelet packet coefficient value of each sub - band signal at time and record it as the sub - band signal amplitude of each sub - band signal , for example, for a signal with a sampling frequency of 200 HZ, after 6 - layer wavelet packet decomposition, 64 sub - bands are obtained. Each sub - band contains a set of coefficients, and the number of coefficients depends on the length of the original electrical control signal. For example, if the original electrical control signal has 1000 sampling points, then the number of coefficients in each sub - band is approximately 1000. The actual wavelet packet decomposition will reduce the number of points due to down - sampling, but usually, through appropriate processing, the number of points can be kept consistent. For example, through the existing technology of maximum overlap decomposition, then the sub - band signal amplitude is the wavelet packet coefficient value of the i - th sub - band at time .

[0027] Step B: Obtain the sliding energy entropy of each sub - band signal, specifically including: Calculate the energy integral of each sub - band according to Formula Two: -- Formula Two; represents the energy of the K - th sub - band at time t; K represents the sub - band number, used to identify the frequency band position being analyzed; t represents the current time moment, used to define the end point of the sliding window, t > T; T represents the sliding window width, which is used to determine the time range of energy statistics and is determined according to the air conditioner control period; represents the time point integration variable within the interval [t - T, t], which is used to traverse all time points within the window; represents the sub - band signal amplitude, which is obtained by wavelet decomposition of the original electronic control signal in step A.

[0028] In this embodiment, taking the 31st sub - band 93.75 - 96.875HZ as an example, calculate the energy integral of the 31st sub - band. If the signal amplitudes of the 31st sub - band within the window [t - 0.5, t] are: [0.1, - 0.2, - 0.15, 0.3, - 0.25]V, where V represents volts and the sampling interval is 0.1s, then the energy integral ; Here, the selected sliding window width is 0.5s. Because the air conditioner control period is usually in the range of 200ms to 1s, and the period of power frequency noise is 20ms, a window width of 0.5s can cover 25 cycles, meeting the stability of statistics. By calculating the energy integral of the 31st sub - band within the 0.5 - second window, if there is a compressor pulse interference signal, the energy here will suddenly increase.

[0029] Preferably, step B specifically includes: Calculate the energy probability distribution of each sub - band according to formula three: —Formula three; represents the energy probability distribution of the Kth sub - band, which is used to reflect the weight of this frequency band in the total energy; M represents the total number of sub - bands, which is determined by the initial decomposition layer N in step A, that is, M is the ; j represents the sub - band index, which is used to traverse all sub - bands; represents the energy of the jth sub - band at time t; represents the energy of the Kth sub - band at time t.

[0030] In one embodiment, if the total energy of all sub - bands at the current time t , then the energy probability of the 31st sub - band .

[0031] Preferably, step B specifically includes: Calculate the sliding energy entropy of each sub - band according to formula four: —Formula four; Denotes the sliding energy entropy of the K-th subband, which is used to quantify the signal complexity. ; When tends to 1, it indicates that the energy of each subband is evenly distributed, meaning the original electrical control signal belongs to a steady-state signal. When tends to 0, it indicates that the energy is concentrated in a few subbands, meaning there are pulse interferences in the original electrical control signal.

[0032] In one embodiment, if the energy distribution at a certain moment is as follows: Subband 31: ; The energy probability of the other 63 subbands is approximately 0.0154; Then the real-time energy entropy of the 31st subband ; The energy entropy value of 0.72 tends to 1, indicating the existence of moderate noise interference. The entropy value of an ideal noise-free signal should be greater than 0.9.

[0033] From the above example, we can know that when there are continuous pulse interferences in a certain subband signal, its energy ratio will increase significantly, resulting in a decrease in the real-time energy entropy, that is, approaching 0. If it is a steady signal, the energy of each subband is evenly distributed, and the real-time energy entropy value will be close to 1.

[0034] Step C: Compare the sliding energy entropy of each subband signal with the entropy value threshold to determine the corresponding denoising algorithm, and output each denoised subband signal. Preferably, the step C includes: When is true, use the morphological filtering algorithm to denoise the current subband and output the denoised subband signal. When is true, use the wavelet threshold algorithm to denoise the current subband and output the denoised subband signal. When is true, use the LMS adaptive filtering algorithm to denoise the current subband and output the denoised subband signal. And represents the entropy value threshold, which is calibrated by noise samples. , .

[0035] In this embodiment, when is true, it means it is in a high entropy value scenario, and the morphological filtering algorithm is used for denoising. The morphological filtering algorithm is a prior art. This application presents an existing example for explanation. For example, the opening operation formula of the morphological filtering algorithm is , and the closing operation formula of the morphological filtering algorithm is , represents the amplitude of the denoised subband signal obtained after the opening operation, represents the amplitude of the denoised subband signal obtained after the closing operation, represents the dilation operation using the structuring element B, represents the erosion operation using the structuring element B; When , it means it is in the medium entropy value scenario, and the wavelet threshold algorithm is used for denoising. The wavelet threshold algorithm is a prior art. This application presents an existing example for explanation. For example, for the subband signal amplitude , soft threshold processing formula is used for denoising, and the soft threshold processing formula is: ; represents the amplitude of the subband signal after denoising by the wavelet threshold algorithm, sign() represents the sign function. When , , when , , when , , represents the soft threshold, , represents the estimated value of the noise standard deviation, usually obtained by dividing the median of the absolute values of the subband coefficients by the constant 0.6745, that is , N represents the decomposition level; When , it means it is in the low entropy value scenario, and the LMS adaptive filtering algorithm is used for denoising. The LMS adaptive filtering algorithm is a prior art. This application presents an existing example for explanation: The core of the LMS algorithm is the adaptive update formula of the filter coefficients. Let: n is the current time; is the current filter coefficient vector; is the input vector (reference noise signal); is the desired response (i.e., the noisy signal); is the output of the filter, , T is the transpose; is the error signal, ; Then the update formula of the filter coefficients is: ; Among them, is the step size factor, that is, the convergence factor.

[0036] Furthermore, for the entropy value threshold , a value higher than this indicates that the signal structure is stable and suitable for the morphological filtering algorithm. For the entropy value threshold , a value lower than this indicates strong randomness and requires the LMS adaptive filtering algorithm; for the determination of the entropy value threshold , through experimental data collection, we prove that the two values 0.5 and 0.85 are the optimal entropy value thresholds, which can better distinguish slow-varying noise. The experimental data table is as follows:

[0037] As shown in the above table, setting a buffer zone between the distribution intervals, that is, the entropy value threshold (0.5 - 0.85), can avoid critical value oscillation.

[0038] Furthermore, as shown in the complementary relationship table of the three algorithms:

[0039] It can be seen from the complementary relationship table of the algorithms that for different frequency bands and different entropy value scenarios, adopting the corresponding denoising algorithm helps to improve the defect compensation in noise detection, can solve the adaptability defects of traditional methods in complex noise scenarios, and provides a high-robustness signal preprocessing guarantee for air conditioner intelligent control.

[0040] Step D: Perform weighted reconstruction on the denoised subband signals to obtain the reconstructed subband signals; Preferably, the specific steps of step D include: Include using formula five and formula six to perform weighted reconstruction on each subband signal: —Formula five; —Formula six; represents the reconstructed subband signal after denoising and reconstruction; represents the Kth subband signal after denoising, which is obtained by using the corresponding denoising algorithm in step C for denoising; represents the dynamic weight coefficient, which is used to reflect the credibility of the subband signal; exp() represents the natural exponential function; represents the weight adjustment factor, which is used to reflect the balance between the number of subbands and sensitivity; represents the real-time subband signal-to-noise ratio, which is used to quantify the quality of the subband signal; represents the signal-to-noise ratio reference threshold, which is set according to the noise floor of the air conditioner system.

[0041] In this embodiment, Denote the signal-to-noise ratio reference threshold, which is set according to the noise floor of the air-conditioning system. For example, through experiments, when the air conditioner is in the standby state, the average SNR of the system noise floor is 16.2 dB, and the measured standard deviation of the SNR fluctuation is 1.2 dB. Then the signal-to-noise ratio reference threshold , which means that when , the dynamic weight coefficient tends to 1, indicating that the k-th sub-band can be fully trusted. When , the dynamic weight coefficient tends to 0, indicating that the k-th sub-band will be suppressed; Furthermore, for the weight adjustment factor , we usually take a value of 0.5, as shown in the following table:

[0042] Corresponding to the above table, then when , by combining the calculation of the real-time sub-band signal-to-noise ratio and the signal-to-noise ratio reference threshold, the inferior sub-bands can be quickly suppressed during sudden interference. When , a steady-state smooth transition can be achieved.

[0043] Specifically, three sub-band embodiments at the moment of starting air-conditioning refrigeration are provided as follows:

[0044] Reconstructed output: , realizing the dominant output of low-frequency effective signals; Therefore, through real-time signal-to-noise ratio evaluation, different weights are dynamically assigned to sub-bands, breaking through the limitations of traditional fixed weights or average fusion; By setting the signal-to-noise ratio reference threshold based on the system background noise statistics, the dynamic characteristics of the weight adjustment factor are configured; Ensure a balance between noise suppression and signal retention: For the dominant output of high-SNR sub-bands, effective signals can be retained, and for low-SNR sub-bands, strong suppression can be performed to eliminate residual noise, fundamentally solving the problem that effective signals are contaminated by traditional reconstruction methods under strong interference, providing high-fidelity signal recovery ability for the intelligent electronic control system of air conditioners.

[0045] Step E: Calculate the real-time residual error evaluation value for the reconstructed sub-band signals, compare the real-time residual error evaluation value with the residual error threshold, and determine whether to output the sub-band signals reconstructed in step D as the final output signal or obtain correction parameters and optimize steps A to C based on the correction parameters. The correction parameters include the new initial decomposition layer number for wavelet packet decomposition after optimization and the new entropy value threshold after optimization.

[0046] Preferably, step E specifically includes: Calculate the real-time residual error evaluation value according to Equation Seven: —Formula VII; represents the real - time residual bad - evaluation value, which is used to quantify the deviation between the denoised signal and the ideal signal; represents the amplitude error weight coefficient, which is used to control the contribution weight of the static deviation; represents the differential error weight coefficient, which is used to control the contribution weight of the dynamic change; represents the original electronic control signal; represents the sub - band signal after weighted reconstruction in step D; represents the first - order derivative of the original electronic control signal at time t, which is used to represent the instantaneous change rate of the signal.

[0047] In this embodiment, the real - time residual bad - evaluation value can be calculated through Formula VII. The real - time residual bad - evaluation value is used to quantify the deviation between the denoised signal and the ideal signal, that is, to quantify the deviation between the denoised signal and the ideal signal without error as we envisioned through the real - time residual bad - evaluation value, and is used to determine whether it is necessary to optimize the parameters of step A and step C in the follow - up. In an example, if the range of the air - conditioner temperature sensor signal is 0 - 5V, corresponding to 0 - 50°C, the frequency is 200HZ, and the sampling interval is 0.005s; Then, as shown in the calculation of the real - time residual bad - evaluation value in the following table:

[0048] As can be seen from the above table, when the signal changes rapidly ( ), even if the amplitude error is very small, the real - time residual bad - evaluation value is still greater than 1, indicating that the system is extremely sensitive to the dynamic process.

[0049] Preferably, step D further includes: Comparing the real - time residual bad - evaluation value with the residual threshold: Residual threshold In this embodiment, the noise standard deviation is calculated by collecting the air - conditioner standby - state signal. For example, for 1000 sampling - point noise signals noise, the noise standard deviation σ = np.std(noise) is calculated. np.std() represents the function in the NumPy library for calculating the standard deviation of an array, and the experimental data value of the noise standard deviation can be obtained as 0.0167V. Then, through the 3σ criterion covering 99.7% of the noise, the residual threshold .

[0050] When the sub - band signal after weighted reconstruction in step D is used as the final output signal for output; When the real-time residual bad review value is less than or equal to the residual threshold, we believe that under the existing parameter calculations in steps A and C, the denoised subband signal can be used as the final output signal after weighted reconstruction; while when the real-time residual bad review value is greater than the residual threshold, we believe that it is necessary to optimize the decomposition layer number in step A and the entropy value threshold in step C.

[0051] When it includes: Optimize the initial decomposition layer number N according to formula eight, —Formula eight; represents the optimized decomposition layer number; sgn() represents the sign function. When it is ; when it is ; when it is ; represents the derivative of the real-time residual bad review value R(t) at time t, which is used to reflect the change trend of the real-time residual bad review value; For example, in an embodiment, the initial decomposition layer number N = 6, and it is detected that which means the residual continues to increase, then the optimized decomposition layer number increases the initial decomposition layer number to achieve a finer frequency band division.

[0052] Optimize the entropy value threshold according to formula nine, —Formula nine; represents the optimized entropy value threshold. The entropy value threshold for the first optimization is artificially calibrated based on the entropy value threshold calibrated from the noise samples at the beginning. The optimized entropy value threshold after the first time is calculated by formula nine; represents the learning rate; in this embodiment, it is taken as 0.01; represents the difference between the real-time residual bad review values before and after optimization; represents the difference between the entropy value thresholds before and after optimization; The embodiments are shown in the following table:

[0053] As shown in the above table, the entropy value threshold for the first optimization is 0.87, which is the entropy value threshold is manually calibrated on the basis of 0.85. Since it learns by the ratio of the difference between the residuals before and after optimization to the difference between the entropy value thresholds before and after optimization, the entropy value threshold for the first optimization needs to be manually calibrated by experience to form a difference for subsequent automatic optimization. Therefore, the entropy value threshold obtained after the second optimization = 0.885 is calculated through Formula 9 on the basis of the entropy value threshold manually calibrated in the first optimization. That is, starting from the second optimization, the entropy value thresholds for subsequent optimizations can be corrected through Formula 9.

[0054] For example, the entropy value threshold 0.885 obtained in the second optimization is first calculated , that is, the difference between the residual in the first optimization and the residual in the 0th optimization, , that is, the difference between the entropy value threshold in the first optimization and the entropy value threshold in the 0th optimization, . Substituting it into Formula 9, the entropy value threshold obtained in the second optimization is: .

[0055] For example, in an embodiment, when , it means that the entropy value threshold needs to be increased , indicating that the entropy value threshold of the current decision is too low and the entropy value threshold needs to be continuously increased to expand the filtering range of the morphological filter.

[0056] The optimized decomposition level is used as the new decomposition level N and passed back to step A, and a dynamic decomposition tree is constructed using the new decomposition level N; The optimized entropy value threshold is used as the new entropy value threshold and passed back to step C, and the new entropy value threshold is used to expand the filtering range of the morphological filtering algorithm.

[0057] In step E, by constructing a two-dimensional residual, that is, the static amplitude error term to ensure the control accuracy, and by constructing the dynamic differential error term to ensure the response accuracy; through the optimization of the decomposition level N and the entropy value threshold, the defect that the traditional open-loop denoising system cannot adapt to complex working conditions can be fundamentally solved, providing a double guarantee of anti-interference and control accuracy for the air-conditioning electronic control system.

[0058] The technical principle of the present invention has been described above in combination with specific embodiments. These descriptions are only for explaining the principle of the present invention and cannot be construed as limiting the protection scope of the present invention in any way. Based on the explanations herein, those skilled in the art can think of other specific embodiments of the present invention without creative labor, and these embodiments will fall within the protection scope of the present invention.

Claims

1. An electrical signal optimization method for an air conditioner's electrical control system, characterized in that: Step A: Obtain the original electrical control signal of the air conditioner, and decompose the original electrical control signal into different frequency bands through wavelet packet decomposition to obtain multiple sub-band signals; Step B: Obtain the sliding energy entropy of each sub-band signal; Step C: Compare the sliding energy entropy of each sub-band signal with an entropy threshold to determine the corresponding denoising algorithm, and output each denoised sub-band signal; Step D: Perform weighted reconstruction on the denoised sub-band signals to obtain reconstructed sub-band signals; Step E: Calculate the real-time residual evaluation value for the reconstructed sub-band signals, compare the real-time residual evaluation value with a residual threshold, and based on the comparison result, determine whether to output the sub-band signals reconstructed in Step D as the final output signal or obtain correction parameters and optimize Steps A to C based on the correction parameters. The correction parameters include the new initial decomposition layer number for wavelet packet decomposition after optimization and the new entropy threshold after optimization.

2. The electrical signal optimization method for an air conditioner's electrical control system according to claim 1, characterized in that: The specific content of Step A includes: Determine the initial wavelet packet initial decomposition layer number according to Formula 1: — Formula 1; N represents the initial decomposition layer number; floor() represents the floor function; fs represents the sampling frequency of the original electrical control signal; It represents a preset frequency resolution. The minimum frequency band width of the original electrically controlled signal obtained after N-layer decomposition should be greater than the preset frequency resolution.

3. The electrical signal optimization method for an air conditioner's electrical control system according to claim 2, characterized in that: Step A further includes: Construct a dynamic decomposition tree according to the initial decomposition layer number N. When decomposing each layer, divide the bandwidth of the previous layer into two equal parts, so that at the last layer of decomposition, the corresponding number of frequency bands are obtained with the minimum bandwidth, and the original electrical control signal is decomposed into sub-band signals of the corresponding number of frequency bands, including: The first - layer decomposition: The sampling frequency of the original electrical control signal is used as the initial bandwidth for average binary division, obtaining the bandwidths of the sub - bands in the first layer. The bandwidth of each sub - band is HZ; Second - layer decomposition: Average - divide the bandwidths of the sub - bands in the first layer into two equal parts respectively, obtaining the sub - band bandwidths in the second layer. The bandwidth of each sub - band is HZ; And so on, decompose successively until the Nth layer of decomposition; Decomposition of the Nth layer: The bandwidths of the subbands obtained by decomposing the (N - 1)th layer are each evenly divided into two, resulting in the bandwidths of the subbands of the Nth layer. The bandwidth of each subband is HZ, thereby decomposing the original electrically controlled signal into subband signals according to the minimum bandwidth of HZ; The bandwidths of the subbands are each evenly divided into two, resulting in the bandwidths of the subbands of the Nth layer. The bandwidth of each subband is HZ, thereby decomposing the original electrically controlled signal into HZ into subband signals; Obtain the wavelet packet coefficient value of each sub-band signal at time and record it as the sub-band signal amplitude of each sub-band signal .

4. The electrical signal optimization method for an air conditioner's electrical control system according to claim 3, characterized in that: The specific content of Step B includes: Calculate the energy integral of each sub-band according to Formula 2: — Formula II; Indicates the energy of the Kth subband at time t; K represents the sub-band number, which is used to identify the analyzed frequency band position; t represents the current time moment, which is used to define the end point of the sliding window, t > T; T represents the sliding window width, which is used to determine the time range of energy statistics and is determined according to the air conditioner control period; Represents the time-point integration variable within the interval [t - T, t], which is used to traverse all the time points within the window; Indicates the sub-band signal amplitude, which is obtained by wavelet decomposition of the original electronic control signal in step A.

5. The electrical signal optimization method for an air conditioner's electrical control system according to claim 4, characterized in that: The specific content of Step B includes: Calculate the energy probability distribution of each sub-band according to Formula 3: — Formula III; Represents the energy probability distribution of the K-th sub-band, which is used to reflect the weight of this frequency band in the total energy; M represents the total number of subbands, which is determined by the initial decomposition level N in step A, that is, M is the ; j represents the sub-band index, which is used to traverse all sub-bands; represents the energy of the j-th sub-band at time t; Represents the energy of the K-th sub-band at time t.

6. The electrical signal optimization method for an air conditioner's electrical control system according to claim 5, characterized in that: The specific content of Step B includes: Calculate the sliding energy entropy of each sub-band according to Formula 4: — Formula Four; represents the sliding energy entropy of the K-th subband, which is used to quantify the signal complexity, ; When tends to 1, it indicates that the energies of all sub-bands are evenly distributed, indicating that the original electrically controlled signal belongs to a steady-state signal; When tends to 0, it indicates that the energy is concentrated in a few subbands, indicating that there is pulse interference in the original electronic control signal.

7. The electrical signal optimization method for an air conditioner's electrical control system according to claim 6, characterized in that: Step C includes: When occurs, the morphological filtering algorithm is used to denoise the current subband, and the denoised subband signal is output; When the wavelet threshold algorithm is used to denoise the current subband, and the denoised subband signal is output; When the LMS adaptive filtering algorithm is used to denoise the current subband, and the denoised subband signal is output; and represents the entropy value threshold, calibrated by noise samples , .

8. The electrical signal optimization method for an air conditioner's electrical control system according to claim 7, characterized in that: The specific content of Step D includes: Include using Formula 5 and Formula 6 to perform weighted reconstruction on each sub-band signal: — Formula Five; —Formula VI; Indicates the subband signal after denoising and reconstruction; Denotes the K-th subband signal after denoising, obtained by step C using the corresponding denoising algorithm for denoising; Indicates the dynamic weight coefficient, which is used to reflect the credibility of the sub-band signal; exp() represents the natural exponential function; Denotes a weight adjustment factor, which is used to reflect the balance between the number of sub-bands and sensitivity; Indicates the real-time sub-band signal-to-noise ratio, which is used to quantify the quality of the sub-band signal; Denotes the SNR reference threshold, which is set according to the noise floor of the air conditioning system.

9. The method for optimizing electrical signals of an air conditioner's electrical control system according to claim 8, wherein: The specific steps of step E include: Calculating the real-time residual bad evaluation value according to Formula VII: —Formula VII; Indicates the real-time value of residual good and bad reviews; Represents the amplitude error weight coefficient, which is used to control the contribution weight of the static deviation; Denotes the differential error weight coefficient, which is used to control the contribution weight of dynamic changes; Represents the original electronic control signal; Indicates the subband signal after weighted reconstruction in step D; Represents the first derivative of the original electronic control signal at time t, which is used to represent the instantaneous change rate of the signal.

10. The method for optimizing electrical signals of an air conditioner's electrical control system according to claim 9, wherein: Step D further includes: Comparing the real-time residual bad evaluation value with the residual threshold: When the sub-band signal after weighted reconstruction in step D is output as the final output signal; When it includes: Optimize the initial decomposition level N according to Equation VIII, — Equation VIII; Indicates the optimized number of decomposition levels; sgn() represents the sign function. When then When then When then ; , which represents the derivative of the real-time residual good and bad review value R(t) at time t and is used to reflect the changing trend of the real-time residual good and bad review value; Optimize the entropy threshold according to Equation 9 — Equation 9; — Equation 9 represents the optimized entropy value threshold, the entropy value threshold of the first optimization is calibrated manually based on the entropy value threshold calibrated by noise samples initially and the optimized entropy value threshold after the first time is calculated through Formula 9 and is obtained by calculation using Formula 9; denotes the learning rate; Indicates the difference in the real-time residual good and bad review value before and after optimization; Indicates the difference in the entropy value threshold before and after optimization; The optimized number of decomposition levels is fed back to step A as the new number of decomposition levels N, and a dynamic decomposition tree is constructed using the new number of decomposition levels N; Take the optimized entropy threshold value as the new entropy threshold value and backhaul it to step C, and use the new entropy threshold value to expand the filtering range of the morphological filtering algorithm.

Images