An adaptive order-based air conditioning system active noise reduction method and system

Abstract

The application provides an air conditioning system active noise reduction method and system based on adaptive order, relates to the active noise reduction technical field, and comprises the following steps: collecting environmental noise signals and pre-processing to obtain input signals; the input signals are subjected to adaptive order Volterra nonlinear expansion to obtain expanded signals; the expanded signals are input into a non-normal FIR filter, the filter output signals are obtained through non-normal FIR filter processing, and the filter output signals pass through a secondary path to obtain reverse noise signals; the reverse noise signals are input into a noise environment for noise cancellation, and residual error signals after noise cancellation are collected; filter parameters of the non-normal FIR filter are updated according to the residual error signals. The foregoing steps are repeatedly executed until a preset noise reduction requirement is reached. The application improves the real-time performance, stability and nonlinear noise suppression capacity of the air conditioning system active noise reduction under complex working conditions.

Description

Technical Field

[0001] This invention relates to the field of active noise reduction technology, and in particular to an active noise reduction method and system for air conditioning systems based on adaptive order. Background Technology

[0002] As HVAC systems continue to evolve towards larger scale, greater intelligence, and higher power operation, the aerodynamic noise, mechanical vibration noise, and duct coupling noise generated during operation are becoming increasingly prominent issues. Traditional passive noise reduction methods mainly rely on sound-absorbing materials, sound insulation structures, or damping devices to attenuate noise. However, these methods have limited effectiveness in suppressing low-frequency noise and suffer from problems such as large structural volume, high material costs, and insufficient adaptability. In recent years, active noise cancellation technology, which can cancel low-frequency noise by generating a reverse sound wave with the opposite phase to the original noise, has been gradually applied in fields such as air conditioning ducts, vehicle air conditioning, industrial ventilation, and intelligent buildings. Existing active noise cancellation systems typically use adaptive filtering algorithms combined with finite impulse response (FIR) filters to generate reverse noise and dynamically adjust filter parameters through error feedback mechanisms to improve the system's real-time noise reduction capability. At the same time, Volterra nonlinear modeling, adaptive filtering, and intelligent parameter adjustment in complex noise scenarios are also gradually becoming important research directions in the field of active noise cancellation.

[0003] However, existing active noise cancellation technologies for air conditioning systems still have several shortcomings. First, most traditional active noise cancellation methods employ fixed-order Volterra models or linear FIR filter structures, which cannot dynamically adjust their nonlinear modeling capabilities according to the complexity of the noise environment. When the operating state of the air conditioning system changes, such as changes in fan speed, airflow adjustment, or increased airflow disturbance, the nonlinear coupling components in the environmental noise will increase significantly. Fixed-order models struggle to simultaneously balance the accuracy of complex noise modeling with real-time computational efficiency, easily leading to excessive filter computation or insufficient noise feature representation, thus reducing the noise cancellation effect. Second, traditional standard FIR filters typically employ a serial delay topology, where the critical path length increases with the filter order. In real-time active noise cancellation in large air conditioning systems, this can easily generate significant system processing delays, preventing the generated inverse noise from timely tracking the dynamically changing environmental noise, thereby affecting the noise cancellation accuracy. Furthermore, most existing adaptive filtering algorithms update parameters based on the mean square error criterion, exhibiting poor robustness to non-Gaussian noise, impulse interference, and complex coupled noise environments. Under complex operating conditions in air conditioning systems, they are prone to slow filter convergence, insufficient stability, and even local instability.

[0004] How to solve the above-mentioned technical problems is the challenge facing this invention. Summary of the Invention

[0005] To address the shortcomings of existing technologies, this invention provides an adaptive order-based active noise reduction method and system for air conditioning systems, which improves the real-time performance, stability, and nonlinear noise suppression capabilities of active noise reduction under complex operating conditions.

[0006] The technical solution adopted by this invention to solve its technical problem is as follows: This invention provides an active noise reduction method for air conditioning systems based on adaptive order, comprising the following steps: S1. Collect and preprocess the ambient noise signal to obtain the input signal; S2. Perform an adaptive order Volterra nonlinear expansion on the input signal to obtain the expanded signal; the adaptive order Volterra nonlinear expansion adaptively determines the expansion order based on the signal characteristics. S3. The expanded signal is input into the non-standard FIR filter, and the filter output signal is obtained after processing by the non-standard FIR filter. The filter output signal is then passed through a secondary path to obtain the inverse noise signal. The non-standard FIR filter adopts a non-serial delay topology structure with rearranged delay units and signal paths. S4. Input the reverse noise signal into the noisy environment to cancel the noise, and collect the residual error signal after noise cancellation. S5. Update the filter parameters of the non-standard FIR filter based on the residual error signal; S6. Repeat steps S1 to S5 until the preset noise reduction requirement is met.

[0007] Preferably, the preprocessing includes signal amplification, filtering and noise reduction, normalization, and time-domain smoothing.

[0008] Preferably, step S2, which involves performing an adaptive order Volterra nonlinear expansion of the input signal, specifically includes: The input signal is expanded using Volterra to obtain several signal components of different orders, and the component energy and total energy of each signal component of different orders are calculated respectively. The energy percentage of each signal component is determined based on the ratio of the energy of each signal component to the total energy, and the energy percentage of each signal component is compared with the corresponding preset threshold. When the proportion of component energy is greater than or equal to a preset threshold, the signal component of the corresponding order is retained; otherwise, the signal component of the corresponding order is ignored. The final order of the Volterra nonlinear expansion is determined based on the retained signal components of each order, and the retained signal components of each order are cascaded to obtain the expanded signal.

[0009] Preferably, the expanded signal includes linear terms and several nonlinear terms, as expressed by the following formula:

[0010] In the formula, To control the first The activation switch function for the start and stop states of the subvector. For the first First-order signal components, It is a linear reference component.

[0011] Preferably, the step S3 of obtaining the filtered output signal by processing with a non-normal FIR filter includes: processing the expanded signal using a time offset coefficient and a time offset matrix to obtain the filtered output signal, as expressed by the following formula:

[0012] in, This represents the total dynamic expansion length of the filter. This is the time offset coefficient. ; This is the time offset matrix. , Represents a diagonal matrix; For the current moment The tap weight coefficient vector of the non-canonical FIR filter; express The first time calculated at time The tap weight coefficient of each tap; express The unfolding signal of time.

[0013] Preferably, the specific method for updating the filter parameters of the adaptive filter based on the residual error signal in step S5 is as follows: An adaptive update method based on the maximum correlation entropy criterion is used to iteratively update the tap weight coefficient vector of the non-normalized FIR adaptive filter. The weight coefficient update formula is expressed as:

[0014] in, This is the tap coefficient vector for the next time step. For step size parameters, This is the residual error signal. For kernel width parameter, This is the time offset matrix. This is the extended input vector after secondary path filtering.

[0015] Preferably, in step S6, the specific method for repeating steps S1 to S5 until the preset noise reduction requirement is met is as follows: Within each sampling period, the mean square value of the residual error signal is calculated and compared with a preset error threshold. When the mean square value of the residual error signal is less than the preset error threshold, it is determined that the preset noise reduction requirement has been met, and the iteration stops.

[0016] This invention also provides an active noise reduction system for an air conditioning system based on adaptive order, comprising: The signal acquisition and preprocessing module is used to acquire environmental noise signals and preprocess them to obtain the input signal. The nonlinear expansion module is used to perform adaptive order Volterra nonlinear expansion on the input signal to obtain the expanded signal. The noise reduction signal acquisition module is used to input the expanded signal into a non-standard FIR filter, process it through the non-standard FIR filter to obtain a filtered output signal, and make the filtered output signal obtain an inverse noise signal through a secondary path; The noise reduction execution module is used to input the reverse noise signal into the noisy environment for noise cancellation and to collect the residual error signal after noise cancellation. The parameter update module is used to update the filter parameters of the non-standard FIR filter based on the residual error signal. The iterative execution module is used to repeatedly execute steps S1 to S5 until the preset noise reduction requirements are met.

[0017] The present invention also provides an electronic device, including a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor executes the program to implement the steps of the above-described active noise reduction method for an air conditioning system based on adaptive order.

[0018] The present invention also provides a computer storage medium storing a computer program thereon, which, when executed by a processor, implements the steps of the above-described active noise reduction method for an air conditioning system based on adaptive order.

[0019] The beneficial effects of this invention are: improving the real-time performance, stability, and nonlinear noise suppression capability of active noise reduction in air conditioning systems under complex operating conditions. By performing adaptive-order Volterra nonlinear expansion on the environmental noise signal, and dynamically adjusting the Volterra expansion order according to the energy proportion of each order of nonlinear components in the signal, the system can adaptively select the higher-order nonlinear components to participate in modeling based on the noise complexity under different operating states of the air conditioning system. This improves the characteristic representation capability of complex nonlinear noise while ensuring computational efficiency, solving the technical problem that existing fixed-order Volterra models cannot balance modeling accuracy and real-time performance. A non-standard FIR filter with a non-serial delay topology structure featuring rearranged delay units and signal paths is employed. By introducing time offset coefficients and time offset matrices, the traditional serial delay structure is optimized, shortening the critical computation path of the filter and reducing system processing latency. This allows the generated inverse noise to track dynamically changing environmental noise more promptly, thereby improving the real-time response capability and noise cancellation accuracy in the active noise reduction process of large air conditioning systems. By outputting the inverse noise signal to the noisy environment and acquiring the residual error signal in real time, a closed-loop adaptive feedback mechanism is constructed. This enables the system to continuously adjust the filter parameters based on the current noise cancellation effect, thereby enhancing the active noise reduction system's ability to track dynamic changes in operating conditions such as airflow fluctuations, airflow disturbances, and structural vibrations, and improving system operational stability. An adaptive parameter update method based on the maximum correlation entropy criterion is adopted. By using a Gaussian kernel function to perform nonlinear weighting processing on the residual error signal, the robustness of the adaptive filter to non-Gaussian noise, impulse interference, and complex coupled noise environments is improved, allowing the filter to maintain a fast convergence speed and stable update capability even under complex operating conditions. Attached Figure Description

[0020] Figure 1 This is a diagram illustrating the method steps of the present invention.

[0021] Figure 2 This is a system module diagram of the present invention.

[0022] Figure 3 This is a schematic diagram comparing the noise reduction capabilities of Embodiment 3 of the present invention with those of the fixed-order NCMCC algorithm.

[0023] Figure 4 This is a schematic diagram of the ANR curves of active noise reduction under real transformer noise conditions for Embodiment 3 of the present invention and other algorithms.

[0024] Figure 5 This is a comparison chart of the ANR curves of various algorithms under the impulse noise environment in Embodiment 3 of the present invention.

[0025] Figure 6This is a comparison chart of the active noise reduction capabilities of Embodiment 3 of the present invention and other algorithms under real transformer noise conditions.

[0026] Figure 7 This is a diagram of the internal structure of a computer device according to Embodiment 4 of the present invention. Detailed Implementation

[0027] To clearly illustrate the technical features of this solution, the following detailed implementation method will be used to explain the solution.

[0028] Example 1 This embodiment provides an active noise reduction method for an air conditioning system based on adaptive order, including steps S1 to S6.

[0029] S1. Collect and preprocess the ambient noise signal to obtain the input signal; Specifically, this step is used to acquire environmental noise information generated during the operation of the air conditioning system, which will serve as the raw input data for subsequent active noise reduction processing. It should be noted that during actual operation, the noise generated by the air conditioning system is typically characterized by a high proportion of low frequencies, strong non-stationarity, and significant nonlinearity due to factors such as changes in fan speed, airflow adjustment, airflow disturbance, and changes in duct structure. Therefore, real-time acquisition of environmental noise is necessary to ensure that the subsequent active noise reduction process can track noise changes in a timely manner.

[0030] Furthermore, environmental noise signals can be collected using reference microphones, acoustic sensors, or vibration sensors installed in the air conditioning system. Reference microphones can be positioned inside the air conditioning duct, near the fan, near the air outlet, or along the noise propagation path to obtain reference noise information during noise propagation in advance. Vibration sensors can be installed on the fan casing, duct wall, or supporting structure to collect coupled noise information caused by structural vibration. The coordinated use of different types of sensors can improve the ability to perceive complex noise environments.

[0031] It should also be noted that the environmental noise signal acquired in this step can be either a continuous-time analog signal or a discrete-time digital signal. When the sensor output is an analog signal, the environmental noise signal can be sampled and quantized using an analog-to-digital converter to obtain a discrete input signal. The sampling frequency can be set according to the noise frequency range of the air conditioning system to meet the real-time and frequency domain resolution requirements of the active noise cancellation system. Typically, the sampling frequency should be higher than twice the highest frequency of the environmental noise to ensure complete acquisition of noise information.

[0032] Furthermore, to reduce the impact of environmental interference and random measurement errors on the subsequent active noise reduction process, the input signal can be preprocessed after acquiring the environmental noise signal. Preprocessing includes, but is not limited to, signal amplification, filtering and denoising, normalization, and time-domain smoothing. Filtering and denoising can suppress sensor noise and high-frequency interference signals in the environment; normalization can reduce the impact of input signal amplitude fluctuations under different operating conditions on the stability of subsequent adaptive filtering.

[0033] Furthermore, the environmental noise signal acquired in this step is used as the reference input signal for the active noise reduction system and is input into the subsequent Volterra nonlinear expansion module to construct a nonlinear characteristic representation of the environmental noise. Since the noise of the air conditioning system exhibits significant dynamic changes under different operating conditions, this step continuously collects environmental noise in real time, thereby providing a real-time data foundation for subsequent adaptive order adjustment and inverse noise generation.

[0034] S2. Perform an adaptive order Volterra nonlinear expansion on the input signal to obtain the expanded signal; the adaptive order Volterra nonlinear expansion adaptively determines the expansion order based on the signal characteristics. Specifically, this step is used to process the environmental noise signal collected in step S1. The expanded signal is obtained by performing a nonlinear expansion. , unfold signal It includes linear terms and nonlinear terms.

[0035] It should be noted that in actual system operating environments, noise characteristics are usually time-varying. Most existing active noise control algorithms use fixed-order Volterra filter models. However, fixed-order models face an irreconcilable theoretical contradiction: if the expansion order is set too low, the model fails to adequately characterize nonlinear and harmonic components, severely limiting the noise reduction effect; if the order is set too high, although it can enhance nonlinear modeling capabilities, it inevitably introduces a large number of redundant cross-term calculations, drastically increasing hardware resource consumption and severely reducing the system's real-time data throughput. To overcome the contradiction between computational complexity and modeling accuracy in fixed-order Volterra filters, this embodiment proposes a variable-order Volterra expansion strategy based on signal energy proportion. The core idea of ​​this strategy is to dynamically determine and activate the order that actually contributes to system modeling based on the energy proportion of each order of nonlinear components in the input signal, thereby eliminating redundant high-order terms and achieving adaptive allocation of computational resources.

[0036] Furthermore, in this embodiment, the Volterra filter employs a dual truncation strategy, that is, only retaining the first... The nonlinear term is of order 1, and the memory length of the filter is truncated to... The maximum allowed expansion order of a Volterra filter is defined as follows: At any discrete time To quantify the transient intensity of signals of each order, the first... Volterra nonlinear extended subvectors instantaneous energy The square of the Euclidean norm of the subvector

[0037] In the formula, This represents the standard 2-norm operation for vectors. Based on this, it covers all norms from 1 to... Total energy of the input signal of the first nonlinear component Defined as the linear superposition of the energies of subvectors of each order.

[0038] Furthermore, after clarifying the energy proportions of each order, it is necessary to further construct the unfolding signal. To rigorously eliminate symmetric redundant calculations arising from multiple summations, the first... Volterra nonlinear extended subvectors Only by satisfying the delay constraint Independent product terms constitute

[0039] Based on the energy proportion judgment mechanism, the signal is expanded. The structure is as follows: (The structure is a block-concatenated vector form.)

[0040] In the formula, To control the first The activation switch function for the start / stop state of the subvector of order, whose physical state is determined by the first... Order signal energy ratio With preset decision threshold Jointly decided. Introducing standard indicator functions. Transforming logical decisions into numerical conditions, the activation switch function can be rigorously defined as follows:

[0041] It should be noted that this is achieved by calculating the relative energy percentage. This mechanism can unify noise signals of different absolute intensities to the same evaluation benchmark, achieving an objective measurement of the degree of nonlinear distortion. (In the unfolded signal...) In the cascaded structure, This represents the linear baseline component of the input vector. To ensure the adaptive filter always possesses basic linear noise reduction capability, the switching function corresponding to this component is always in the active state. For higher-order components, when the switching function is activated... At that time, the first Order nonlinear subvector It is fully concatenated into the total extended vector and participates in the weight update; conversely, when At that time, this part of the vector is completely isolated from the computational dimension, precisely blocking all multiplication and addition costs corresponding to this order.

[0042] In summary, based on the dynamically expanded model constructed above, the system can adaptively expand the filtering dimension to capture the distortion feature when it detects an increase in the energy of a specific order of nonlinear characteristic; conversely, it automatically converges the dimension when the nonlinearity tends to weaken. This underlying mechanism not only successfully eliminates the computational waste caused by fixed-order models, but also lays a solid theoretical framework for achieving efficient and intelligent active noise control in complex and variable nonlinear acoustic environments.

[0043] S3. The expanded signal is input into the non-standard FIR filter. After processing by the non-standard FIR filter, the filter output signal is obtained, and the filter output signal is made to obtain the inverse noise signal through the secondary path. The non-standard FIR filter adopts a non-serial delay topology structure of rearranging delay units and signal paths. It should be noted that the core difference between non-canonical FIR filters and traditional canonical FIR filters lies in the fact that traditional canonical FIR filters employ a serial delay structure, where each tap coefficient is delayed at the current time. Multiplying and accumulating the delayed input results in a critical path length that is proportional to the filter order, leading to significant processing delay. The non-standard FIR filter used in this invention optimizes the signal path and delay units by rearranging them. The calculation of the tap weight coefficients is advanced, and a time offset coefficient and a time offset matrix are introduced, which significantly shortens the critical computation path and reduces the overall system processing latency.

[0044] Specifically, the filtered output signal of a non-standard FIR filter. Represented as:

[0045] in, This represents the total dynamic expansion length of the filter. This is the time offset coefficient. ; This is the time offset matrix. , Represents a diagonal matrix; For the current moment The tap weight coefficient vector of the non-canonical FIR filter; express The first time calculated at time The tap weight coefficient of each tap; express The unfolding signal of time; For the current moment The tap weight coefficient vector of the non-canonical FIR filter is constructed by concatenating and rearranging the kernel coefficients of each order in the Volterra series. Specifically, the vector contains the first-order kernel coefficients describing the linearity of the system. Second-order kernel coefficients for capturing quadratic nonlinear distortion up to describing higher-order nonlinear behavior kernel coefficient This vectorized construction corresponds one-to-one with the unfolded signal, allowing complex nonlinear convolution operations to be compactly represented as the inner product of two vectors. This facilitates the direct application of linear adaptive algorithms for unified iterative updates of weights, as shown in the following formula:

[0046] in, The memory length of the filter is truncated.

[0047] It should be noted that introducing time offset coefficients and a time offset matrix into the filtered output signal yields an output form similar to the FIR structure, thus meeting the computational requirements of active noise reduction. Furthermore, after the transformation, the filter error feedback path contains only a finite number of time delays, avoiding unrealizable non-causal time variations. This step ensures the realizability of the filter through the aforementioned causal adaptive framework.

[0048] Furthermore, during the propagation of noise from the loudspeaker to the error microphone in the air conditioning system, it is affected by factors such as the acoustic channel, electroacoustic devices, and the air medium. This effect is usually referred to as the secondary path. If the characteristics of the secondary path are not considered and the loudspeaker is directly driven by the expanded signal, a phase deviation will occur when the generated reverse sound wave and the target noise are spatially superimposed, thereby weakening the noise reduction effect or even amplifying the noise. Therefore, this step further performs secondary path cancellation on the filtered output signal to compensate for the distortion caused by the secondary channel to the sound wave propagation. The formula is as follows:

[0049] in, The impulse response of the secondary path, the filtered signal The anti-noise signal is obtained after secondary path cancellation. , This represents the convolution operation. This filtering process simulates the actual physical path of the inverse sound wave from its generation to its propagation to the error microphone.

[0050] Furthermore, the inverse noise signal obtained in this step is converted from digital to analog and then used to drive the speaker output, where it is superimposed and canceled out with the original noise in the sound field. Simultaneously, this signal is also retained as a reference for coefficient updates in subsequent adaptive algorithms. It should be noted that, due to the use of a non-canonical FIR structure and its time offset correction, this step achieves lower processing latency and higher real-time throughput at the same filter order, enabling the active noise cancellation system to track the dynamic changes in air conditioning system noise more promptly and improve the noise reduction effect.

[0051] S4. Input the reverse noise signal into the noisy environment to cancel the noise, and collect the residual error signal after noise cancellation. Specifically, this step is used to output the reverse noise signal generated in step S3 to the noise propagation area of ​​the air conditioning system, so that the reverse noise signal and the original environmental noise superimpose in space, thereby achieving active noise cancellation. It should be noted that the basic principle of active noise reduction is to use the reverse sound wave with a similar amplitude and opposite phase to the original noise to interfere and superimpose with the original noise, so that the two cancel each other out in the target area, thereby reducing the actual noise energy in the space.

[0052] Furthermore, in this step, the inverse noise signal output by the non-standard FIR filter is converted into an analog audio signal by a digital-to-analog converter module, and then output to the noise propagation environment of the air conditioning system through a loudspeaker, array loudspeaker, or other sound output device. The loudspeaker can be placed near the air conditioning vent, inside the duct, around the equipment housing, or in the noise propagation path area to ensure that the inverse noise can be fully superimposed on the original ambient noise within the target noise reduction area.

[0053] It should be noted that the environmental noise signal forms the main disturbance signal after propagating through the primary acoustic path. The primary acoustic path represents the propagation process of environmental noise from the noise source to the error acquisition location, which is affected by factors such as duct structure, air medium, obstacle reflection, and equipment vibration. The residual error signal at the current moment is obtained by superimposing the reverse noise signal and the main disturbance signal at the error acquisition location, and its expression is:

[0054] in, This represents the residual error signal after noise cancellation; This represents the main disturbance signal after environmental noise propagates along the main path. This indicates a reverse noise signal.

[0055] Furthermore, in this step, the residual error signal can be acquired in real time using an error microphone, an acoustic detection sensor, or an array acoustic sensor. The error microphone can be positioned within the target noise reduction area to detect the actual residual noise level after the ambient noise and the inverse noise are superimposed. By sampling the residual error signal in real time, the noise cancellation effect of the current active noise cancellation system can be reflected in real time.

[0056] It should also be noted that, due to dynamic disturbances such as changes in airflow, airflow fluctuations, and structural vibration coupling during the operation of large air conditioning systems, the propagation characteristics of environmental noise change with the operating state. Therefore, the residual error signal also changes dynamically over time. This embodiment collects the residual error signal in real time and feeds it back to the subsequent non-standard FIR filter parameter update module, enabling the active noise cancellation system to dynamically adjust the filter parameters according to the current noise cancellation effect, thereby improving the system's ability to track changes in complex operating conditions.

[0057] Furthermore, the residual error signal in this step not only reflects the phase matching degree between the current inverse noise and the ambient noise, but also reflects the noise reduction stability of the system under different operating conditions. When the residual error signal continues to decrease, it indicates that the inverse noise output by the current non-standard FIR filter is gradually approaching the target noise characteristics; when the residual error signal fluctuates or increases, it indicates that there is a deviation between the current filter parameters and the ambient noise characteristics, and the filter parameters need to be readjusted through the adaptive update process in subsequent steps.

[0058] It should also be noted that this step outputs the reverse noise signal to the noisy environment in real time and combines it with the residual error signal to construct a closed-loop feedback mechanism, enabling the active noise reduction system to continuously track the noise change process of the air conditioning system, thereby improving the active noise reduction effect and system stability in complex dynamic environments.

[0059] S5. Update the filter parameters of the non-standard FIR filter based on the residual error signal; Specifically, this step is used to update the tap weight coefficient vector of the non-normal FIR (NCFIR) filter in step S3 in real time using the residual error signal collected in step S4, so as to dynamically optimize the matching between the reverse noise signal and the ambient noise, thereby achieving closed-loop active noise reduction.

[0060] It should be noted that the residual error signal can reflect the deviation between the current filter output and the main disturbance signal. The residual error signal is sent back to the non-standard FIR filter and used directly as the feedback basis for the adaptive weight coefficient update.

[0061] Furthermore, to improve the robustness of the algorithm in non-Gaussian impulse noise environments, the maximum correlation entropy criterion is adopted instead of the traditional mean square error criterion.

[0062] Specifically, define the cost function. The expected value of the Gaussian kernel function of the error

[0063] To maximize the cost function and achieve online iterative updates of the weights, an instantaneous cost function is used instead of the statistical expectation, and the stochastic gradient ascent method is applied to derive the weight update formula. This instantaneous cost function relates to the weight vector. The instantaneous gradient can be expanded using the chain rule.

[0064] The partial derivative of the error with respect to the weights is

[0065] in, It is the extended input vector after secondary path filtering, defined as

[0066] The weight update equation can then be obtained:

[0067] It should be noted that this update formula can dynamically adjust the tap weight coefficients according to the magnitude of the current residual error signal and the characteristics of the Volterra expanded signal, enabling the non-standard FIR filter to achieve adaptive optimization in nonlinear complex noise environments, thereby improving the active noise reduction system's ability to track changes in environmental noise and its overall noise reduction effect.

[0068] Furthermore, the adaptive update process in this step is closed-loop. The residual error signal at each moment is used for weight coefficient iteration to ensure continuous optimization of filter parameters. In addition, combined with the adaptive order mechanism in step S2, only the tap weight coefficients corresponding to the currently enabled high-order Volterra components are updated, thereby balancing computational efficiency and noise reduction performance, and realizing the real-time performance and stability of active noise reduction in the air conditioning system.

[0069] S6. Repeat steps S1 to S5 until the preset noise reduction requirement is met.

[0070] Specifically, this step is used to construct a closed-loop iterative mechanism for the active noise reduction system to ensure that the noise of the air conditioning system can be continuously and effectively canceled under different operating conditions. It should be noted that the environmental noise of the air conditioning system has temporal dynamics, including factors such as changes in fan speed, airflow adjustment, airflow disturbance, and structural vibration coupling, which cause the noise characteristics to change continuously over time. Therefore, it is necessary to iteratively execute steps S1 to S5 to collect noise in real time, perform adaptive order Volterra nonlinear expansion, generate inverse noise, measure residual error signals, and adaptively update filter parameters, thereby continuously tracking changes in environmental noise and achieving dynamic active noise reduction.

[0071] Furthermore, in each sampling period, the residual error signal is calculated in this iterative process. mean square value and with the preset error threshold Compare; when If the preset noise reduction requirement is met, it means that the reverse noise output by the system has effectively canceled the environmental noise and met the preset noise reduction requirement, and the iteration stops; otherwise, the system continues to execute steps S1 to S5 for the next round of iteration update.

[0072] It should be noted that the iterative process in this step not only ensures the system's rapid response to transient noise interference, but also enables it to adaptively compensate for changes in noise characteristics during long-term operation, such as fan load fluctuations, airflow disturbances, or changes in pipeline structure, so that the active noise reduction effect remains stable under complex operating conditions.

[0073] Furthermore, the iterative mechanism in this step is closely integrated with the adaptive order Volterra expansion in step S2 and the non-normal FIR filter update in steps S3 and S5, enabling the system to dynamically select high-order nonlinear components to participate in the filtering calculation while ensuring computational efficiency, and to optimize the filter tap weight coefficients in real time, thereby achieving efficient, stable and sustainable noise cancellation in complex environments.

[0074] It should also be noted that by continuously executing steps S1 to S5, this step can form a complete closed-loop adaptive control architecture, enabling the active noise reduction system to monitor and adjust the output reverse noise signal in real time throughout the entire operating cycle, and ultimately achieve noise control under the preset noise reduction index.

[0075] Example 2: See Figure 2 As shown, this embodiment is an active noise reduction system for an air conditioning system based on adaptive order, including... The signal acquisition and preprocessing module is used to acquire environmental noise signals and preprocess them to obtain the input signal. The nonlinear expansion module is used to perform adaptive order Volterra nonlinear expansion on the input signal to obtain the expanded signal. The noise reduction signal acquisition module is used to input the expanded signal into a non-standard FIR filter, process it through the non-standard FIR filter to obtain a filtered output signal, and make the filtered output signal obtain an inverse noise signal through a secondary path; The noise reduction execution module is used to input the reverse noise signal into the noisy environment for noise cancellation and to collect the residual error signal after noise cancellation. The parameter update module is used to update the filter parameters of the non-standard FIR filter based on the residual error signal. The iterative execution module is used to repeatedly execute steps S1 to S5 until the preset noise reduction requirements are met.

[0076] Example 3: To verify the noise reduction performance, convergence speed, and system stability of the proposed adaptive order-based active noise reduction method for air conditioning systems under complex noise environments, comparative experiments were conducted between the proposed method (DVS-VFxNCMCC) and existing active noise reduction algorithms in real transformer noise and impulse noise environments. The compared algorithms included DVS-VFxLMS, ​​DVS-VFxLMP, and DVS-VFxMCC.

[0077] Among them, the DVS-VFxLMS algorithm uses the least mean square (LMS) criterion to update filter parameters; the DVS-VFxLMP algorithm uses the minimum average p-order error criterion to update parameters; the DVS-VFxMCC algorithm uses the maximum correlation entropy (MCC) criterion to update parameters; and the method of this invention (DVS-VFxNCMCC) introduces an adaptive order Volterra nonlinear expansion and a non-normal FIR (NCFIR) filter structure on the basis of the maximum correlation entropy criterion to verify the comprehensive performance improvement effect of this invention in complex nonlinear noise environments.

[0078] During the experiment, the reference noise signal was acquired through a reference microphone placed near the noise source, while the error microphone was placed within the target noise reduction area to acquire the residual error signal in real time. All algorithms used the same sampling frequency, filter order, and initial parameter settings. Active Noise Reduction (ANR) was used as the performance evaluation metric; a lower ANR value indicates stronger active noise reduction capability.

[0079] Figure 3 A comparison of the noise reduction capabilities of DVS-VFxNCMCC algorithms with different thresholds and VFxNCMCC algorithms with a fixed order. Figure 3It can be seen that when the input noise is stable noise of α (α=1.9), the threshold has a certain impact on the variable order expansion algorithm. If the threshold is too large, it will be difficult to enable the order, thus affecting the final noise reduction effect. However, when the threshold is 0.001, the proposed algorithm is still better than the fixed order NCMCC algorithm.

[0080] Figure 4 This is a schematic diagram showing the ANR (Active Noise Reduction) curves of the DVS-VFxLMS algorithm, DVS-VFxLMP algorithm, DVS-VFxMCC algorithm, and the method of this invention (DVS-VFxNCMCC) under real transformer noise conditions. Figure 4 As can be seen, the DVS-VFxLMS algorithm exhibits the slowest convergence rate and the lowest accuracy. In comparison, the DVS-VFxMCC and DVS-VFxLMP algorithms perform moderately well in terms of convergence performance and accuracy, and are relatively better than the DVS-VFxLMS algorithm. Further observation shows that the algorithm combining the adaptive order method and the non-canonical MCC algorithm provided in this embodiment has significantly improved both convergence speed and accuracy, especially in the early stages of algorithm iteration, where its performance is particularly outstanding.

[0081] Figure 5 This is a comparison chart of the ANR curves of various algorithms under impulse noise environment. It can be clearly seen from the chart that the ANR value of the DVS-VFxNCMCC algorithm is lower than that of the other algorithms, which shows that the algorithm has a stronger noise reduction capability than the other algorithms, and verifies the rationality of the improvement.

[0082] Figure 6 This is a comparison chart of the active noise reduction capabilities of the DVS-VFxLMS algorithm, DVS-VFxLMP algorithm, DVS-VFxMCC algorithm, and the method of this invention (DVS-VFxNCMCC) under real transformer noise conditions. It can be seen that the MCC algorithm, combined with a variable-order, non-standard FIR structure, has significantly better noise reduction capabilities than the other algorithms.

[0083] In summary, the active noise reduction method for air conditioning systems based on adaptive order proposed in this invention has a faster convergence speed, lower steady-state residual error, and stronger robustness in complex nonlinear noise environments compared with existing active noise reduction algorithms. It can effectively improve the active noise reduction performance of large air conditioning systems under complex operating conditions.

[0084] Example 4: This embodiment provides a computer device, including a memory and a processor. The memory stores a computer program, and the processor executes the computer program to implement the steps in the above-described method embodiments.

[0085] This computer device can be a server, and its internal structure diagram can be as follows: Figure 7 As shown, the computer device includes a processor, memory, and a network interface connected via a system bus. The processor provides computing and control capabilities. The memory includes non-volatile storage media and internal memory. The non-volatile storage media stores the operating system, computer programs, and a database. The internal memory provides an environment for the operation of the operating system and computer programs in the non-volatile storage media. The database stores server data. The network interface communicates with external terminals via a network connection. When executed by the processor, the computer program implements an active noise reduction method for an air conditioning system based on adaptive order.

[0086] Those skilled in the art will understand that Figure 7 The structure shown is merely a block diagram of a portion of the structure related to the present application and does not constitute a limitation on the computer device to which the present application is applied. Specific computer devices may include more or fewer components than those shown in the figure, or combine certain components, or have different component arrangements.

[0087] Example 5: This embodiment provides a computer-readable storage medium storing a computer program that, when executed by a processor, implements the steps in the above-described method embodiments.

[0088] If the functions implemented by the method are implemented as software functional units and sold or used as independent products, they can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of this invention, or the part that contributes to the prior art or the current technical solution, can be embodied in the form of a software product. This current computer software product is stored in a storage medium and includes several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute all or part of the steps of the methods described in the various embodiments of this invention. The aforementioned storage medium includes various media capable of storing program code, such as USB flash drives, portable hard drives, read-only memory (ROM), random access memory (RAM), magnetic disks, or optical disks.

[0089] The logic and / or steps represented in the flowchart or otherwise described herein, for example, can be considered as a sequenced list of executable instructions for implementing logical functions, and can be embodied in any computer-readable medium for use by, or in conjunction with, an instruction execution system, apparatus, or device (such as a computer-based system, a processor-including system, or other system that can fetch and execute instructions from, an instruction execution system, apparatus, or device). For the purposes of this specification, "computer-readable medium" can be any means that can contain, store, communicate, propagate, or transmit programs for use by, or in conjunction with, an instruction execution system, apparatus, or device.

[0090] More specific examples of computer-readable media (a non-exhaustive list) include: electrical connections (electronic devices) having one or more wires, portable computer disk drives (magnetic devices), random access memory (RAM), read-only memory (ROM), erasable and editable read-only memory (EPROM or flash memory), fiber optic devices, and portable optical disc read-only memory (CDROM). Furthermore, computer-readable media can even be paper or other suitable media on which the program can be printed, because the program can be obtained electronically, for example, by optically scanning the paper or other medium, followed by editing, interpreting, or otherwise processing as necessary, and then stored in computer memory.

[0091] The technical features of this invention not described can be implemented by or using existing technology, and will not be repeated here. Of course, the above description is not a limitation of this invention, and this invention is not limited to the examples above. Any changes, modifications, additions or substitutions made by those skilled in the art within the scope of this invention should also be within the protection scope of this invention.

Claims

1. An active noise reduction method for an air conditioning system based on adaptive order, characterized in that, Includes the following steps: S1. Collect and preprocess the ambient noise signal to obtain the input signal; S2. Perform adaptive order Volterra nonlinear expansion on the input signal to obtain the expanded signal; The adaptive order Volterra nonlinear expansion adaptively determines the expansion order based on the signal characteristics. S3. The expanded signal is input into a non-standard FIR filter, and the filtered output signal is obtained after processing by the non-standard FIR filter. The filtered output signal is then passed through a secondary path to obtain an inverse noise signal. The non-standard FIR filter adopts a non-serial delay topology structure with rearranged delay units and signal paths. S4. Input the reverse noise signal into the noisy environment to cancel the noise, and collect the residual error signal after noise cancellation. S5. Update the filter parameters of the non-standard FIR filter based on the residual error signal; S6. Repeat steps S1 to S5 until the preset noise reduction requirement is met.

2. The active noise reduction method for air conditioning systems based on adaptive order according to claim 1, characterized in that, The preprocessing includes signal amplification, filtering and noise reduction, normalization, and time-domain smoothing.

3. The active noise reduction method for air conditioning systems based on adaptive order according to claim 2, characterized in that, The adaptive order Volterra nonlinear expansion of the input signal in step S2 specifically includes: The input signal is expanded using Volterra to obtain several signal components of different orders, and the component energy and total energy of each signal component of different orders are calculated respectively. The energy percentage of each signal component is determined based on the ratio of the energy of each signal component to the total energy, and the energy percentage of each signal component is compared with the corresponding preset threshold. When the proportion of component energy is greater than or equal to a preset threshold, the signal component of the corresponding order is retained; otherwise, the signal component of the corresponding order is ignored. The final order of the Volterra nonlinear expansion is determined based on the retained signal components of each order, and the retained signal components of each order are cascaded to obtain the expanded signal.

4. The active noise reduction method for air conditioning systems based on adaptive order according to claim 3, characterized in that, The expanded signal includes linear terms and several nonlinear terms, as expressed by the following formula: In the formula, To control the first The activation switch function for the start and stop states of the subvector. For the first First-order signal components, It is a linear reference component.

5. The active noise reduction method for air conditioning systems based on adaptive order according to claim 4, characterized in that, The step S3, which involves processing the unnormalized FIR filter to obtain the filtered output signal, includes: processing the expanded signal using a time offset coefficient and a time offset matrix to obtain the filtered output signal, as expressed by the following formula: in, This represents the total dynamic expansion length of the filter. This is the time offset coefficient. ; This is the time offset matrix. , Represents a diagonal matrix; For the current moment The tap weight coefficient vector of the non-canonical FIR filter; express The first time calculated at time The tap weight coefficient of each tap; express The unfolding signal of time.

6. The active noise reduction method for air conditioning systems based on adaptive order according to claim 5, characterized in that, The specific method for updating the filter parameters of the adaptive filter based on the residual error signal in step S5 is as follows: An adaptive update method based on the maximum correlation entropy criterion is used to iteratively update the tap weight coefficient vector of the non-normalized FIR adaptive filter. The weight coefficient update formula is expressed as: in, This is the tap coefficient vector for the next time step. For step size parameters, This is the residual error signal. For kernel width parameter, This is the time offset matrix. This is the extended input vector after secondary path filtering.

7. The active noise reduction method for air conditioning systems based on adaptive order according to claim 6, characterized in that, In step S6, the specific method for repeating steps S1 to S5 until the preset noise reduction requirement is met is as follows: Within each sampling period, the mean square value of the residual error signal is calculated and compared with a preset error threshold. When the mean square value of the residual error signal is less than the preset error threshold, it is determined that the preset noise reduction requirement has been met, and the iteration stops.

8. An active noise reduction system for an air conditioning system based on adaptive order, characterized in that, The steps for performing the adaptive order-based active noise reduction method for an air conditioning system according to any one of claims 1 to 7 include: The signal acquisition and preprocessing module is used to acquire environmental noise signals and preprocess them to obtain the input signal. The nonlinear expansion module is used to perform adaptive order Volterra nonlinear expansion on the input signal to obtain the expanded signal. The noise reduction signal acquisition module is used to input the expanded signal into a non-standard FIR filter, process it through the non-standard FIR filter to obtain a filtered output signal, and make the filtered output signal obtain an inverse noise signal through a secondary path; The noise reduction execution module is used to input the reverse noise signal into the noisy environment for noise cancellation and to collect the residual error signal after noise cancellation. The parameter update module is used to update the filter parameters of the non-standard FIR filter based on the residual error signal. The iterative execution module is used to repeatedly execute steps S1 to S5 until the preset noise reduction requirements are met.

9. A computer device comprising a memory and a processor, wherein the memory stores a computer program, characterized in that, When the processor executes the computer program, it implements the steps of the active noise reduction method for an air conditioning system based on adaptive order as described in any one of claims 1 to 7.

10. A computer-readable storage medium having a computer program stored thereon, characterized in that, When the computer program is executed by the processor, it implements the steps of the active noise reduction method for an air conditioning system based on adaptive order as described in any one of claims 1 to 7.

Images