Transformer substation ventilation and noise reduction method and device based on adaptive filtering algorithm

Abstract

The invention relates to a transformer substation ventilation and noise reduction method and device based on a self-adaptive filtering algorithm. The method comprises the following steps: carrying out data acquisition on noise of a transformer substation ventilation system; establishing a filtering-XLMS algorithm model; constructing a model based on an improved variable step size LMS algorithm; and analyzing and processing the noise data of the ventilation system of the transformer substation, and outputting a reverse counteracting signal through the controller to counteract the noise of the ventilation system of the transformer substation. According to the principle of sound wave interference offset, corresponding waveforms can be generated in a self-adaptive mode according to fan noise changes to drive the loudspeaker to form negative signals, fan noise signals are offset, and the purpose of active noise elimination and reduction of the transformer substation ventilation system is achieved. The transformer substation noise pollution abatement problem is solved, the problem that transformer substation noise disturbs residents in an urban area is solved, and the high-quality service level of a power grid is improved.

Description

Technical Field

[0001] The present invention belongs to the technical field of noise reduction of transformer substation ventilation systems, and in particular relates to a method and device for reducing noise of transformer substation ventilation based on an adaptive filtering algorithm. Background Art

[0002] With the improvement of urbanization level, the demand for electricity has increased rapidly, and the number of newly built and expanded substations has increased. The substations originally deployed in the suburbs have gradually been surrounded by cities, and even penetrated into the city center. Due to the shortage of urban land resources, the site selection of new substations has increased. As a result, some new substations are close to urban residential areas, resulting in the noise generated by the operation of substations seriously affecting the living standards and physical health of nearby residents.

[0003] A significant portion of substation noise comes from the exhaust fan noise of the ventilation system. Exhaust fan noise is aerodynamic noise caused by the relative motion between the fan's impeller and the surrounding air during operation. Exhaust fans are primarily high-power axial flow fans, which are large and numerous, and operate 24 / 7, generating power-frequency noise, making them the primary source of substation noise. Furthermore, with increasingly stringent environmental regulations and growing environmental awareness, substation noise has become a major source of complaints. Summary of the Invention

[0004] The purpose of the present invention is to overcome the deficiencies of the prior art and to provide a method and device for reducing noise in substation ventilation based on an adaptive filtering algorithm.

[0005] The object of the present invention is achieved by: a method for reducing noise in substation ventilation based on an adaptive filtering algorithm, the method comprising the following steps:

[0006] Step 1: Collect data on the noise of the substation ventilation system;

[0007] Step 2: Establish the filtering-XLMS algorithm model;

[0008] Step 3: Construct a model based on the improved variable step-size LMS algorithm;

[0009] Step 4: Analyze and process the noise data of the substation ventilation system, and output a reverse cancellation signal to cancel the noise of the substation ventilation system.

[0010] In step 2, let the reference signal be x(n), the primary signal be p(n), the output signal of the secondary sound source be y(n), and the error signal after the secondary sound source and the primary noise source are superimposed be e(n);

[0011] The FXLMS algorithm is used, which makes the reference signal pass through a secondary path H S(z) The same additional path is used to update the weight coefficients in the LMS algorithm, solving the problem that the error signal and the reference signal x(n) cannot be correctly "aligned" in time due to the delay of the error signal e(n); (The system control diagram is shown in Figure 2). Figure 1 As shown, H s (z), W(z) is an adaptive filter that uses a finite impulse response structure.)

[0012] The relationship between the reference signal and the primary signal is:

[0013] x(n)=p(n)*h1(n)

[0014] At the error sensor, the received primary sound field and secondary sound field signals are:

[0015] d(n)=p(n)*h3(n)

[0016] s(n)=y(n)*h2(n)

[0017] If the length of the transversal filter is L, the secondary signal is the output of the filter:

[0018]

[0019] Assume that the primary noise has a local stationary characteristic, so that the adaptive filter weight coefficient can be considered to be basically unchanged within the L period, then

[0020]

[0021] Among them, r(n) is called the filter-X signal, and the column vector composed of it is called the filter-X signal vector.

[0022] r(n)=[r(n),r(n-1),…,r(n-l+1)] T

[0023] r(n)=X(n)*h2(n)

[0024] Therefore, the signal received by the error sensor can be expressed as:

[0025] e(n)=d(n)+s(n)=d(n)+r τ (n)W(n)

[0026] Since the error signal is a random signal, in order to find the optimal weight coefficient of the filter, we should first set a criterion (objective function) to be achieved, and then derive the optimal filter transfer function under this criterion. The most commonly used criterion is the minimum mean square error criterion, which is to set the objective function of the control system to

[0027] J(n)=E[e 2 (n)]=E[d 2 (n)]+2P T W+W T RW

[0028] According to the principle of steepest descent method, the recursive filter weight coefficient is:

[0029]

[0030] Where μ is a parameter that controls the stability and convergence speed of the adaptive process, that is, the step size factor. In order to facilitate the real-time implementation of the system, the square of a single error sample e is taken. 2 The gradient of (n) is used as the gradient estimate of the mean square error, that is,

[0031]

[0032] It can be shown that the gradient estimate is the true value An unbiased estimate of replace You can get:

[0033]

[0034] In step 3, the iterative formula of the improved variable step-size LMS algorithm is:

[0035]

[0036] W(n+1)=W(n)-2μ(n)e(n)r(n)

[0037] Wherein, parameter α>0 controls the shape of the function; parameter β>0 controls the value range of the function. By selecting appropriate parameters α and β according to the tracking environment and the distribution of the input signal, the algorithm can have a faster convergence speed and convergence accuracy.

[0038] A substation ventilation noise reduction device based on an adaptive filtering algorithm includes a microphone module, a peripheral signal conditioning module, and a processing and control module. The microphone module includes a measurement sensor and a speaker; the peripheral signal conditioning module includes a signal amplifier, a high-frequency filter, an A / D converter, a D / A converter, and a low-frequency power amplifier; and the processing and control module includes a signal processor.

[0039] Beneficial effects of the present invention: The present invention proposes a method and device for substation ventilation noise reduction based on an adaptive filtering algorithm, including data collection of the substation ventilation system noise; establishing a filtering-XLMS algorithm model; constructing a model based on an improved variable step-size LMS algorithm; analyzing and processing the substation ventilation system noise data, and outputting a reverse cancellation signal to cancel the substation ventilation system noise. By utilizing the principle of acoustic wave interference cancellation, a corresponding waveform can be adaptively generated according to the change of fan noise to drive the speaker to form a "negative signal" and cancel the fan noise signal, thereby achieving the purpose of active noise reduction of the substation ventilation system. It fills the gap in substation noise pollution control, solves the problem of noise disturbing residents in urban areas from substations, and improves the quality service level of the power grid. BRIEF DESCRIPTION OF THE DRAWINGS

[0040] Figure 1 This is a block diagram of the FXLMS algorithm control system of the present invention.

[0041] Figure 2 It is a schematic diagram of the overall design of the control device of the present invention.

[0042] Figure 3 This is a comparison chart between the improved variable step-size LMS algorithm of the present invention and other algorithms. DETAILED DESCRIPTION

[0043] The present invention will be further described below with reference to the accompanying drawings.

[0044] Example

[0045] like Figure 1-3 As shown, a method for reducing noise in substation ventilation based on an adaptive filtering algorithm is provided, wherein the method comprises the following steps:

[0046] Step 1: Collect data on the noise of the substation ventilation system.

[0047] Step 2: Establish the filtering-XLMS algorithm model. The specific process is as follows: In step 2,

[0048] In step 2, let the reference signal be x(n), the primary signal be p(n), the output signal of the secondary sound source be y(n), and the error signal after the secondary sound source and the primary noise source are superimposed be e(n);

[0049] The FXLMS algorithm is used, which makes the reference signal pass through a secondary path H S (z) The same additional path is used to update the weight coefficients in the LMS algorithm, solving the problem that the error signal and the reference signal x(n) cannot be correctly "aligned" in time due to the delay of the error signal e(n); (The system control diagram is shown in Figure 2). Figure 1 As shown, Hs (z), W(z) is an adaptive filter that uses a finite impulse response structure.)

[0050] The relationship between the reference signal and the primary signal is:

[0051] x(n)=p(n)*h1(n)

[0052] At the error sensor, the received primary sound field and secondary sound field signals are:

[0053] d(n)=p(n)*h3(n)

[0054] s(n)=y(n)*h2(n)

[0055] If the length of the transversal filter is L, the secondary signal is the output of the filter:

[0056]

[0057] Assume that the primary noise has a local stationary characteristic, so that the adaptive filter weight coefficient can be considered to be basically unchanged within the L period, then

[0058]

[0059] Among them, r(n) is called the filter-X signal, and the column vector composed of it is called the filter-X signal vector.

[0060] r(n)=[r(n),r(n-1),…,r(n-l+1)] T

[0061] r(n)=X(n)*h2(n)

[0062] Therefore, the signal received by the error sensor can be expressed as:

[0063] e(n)=d(n)+s(n)=d(n)+r τ (n)W(n)

[0064] Since the error signal is a random signal, in order to find the optimal weight coefficient of the filter, we should first set a criterion (objective function) to be achieved, and then derive the optimal filter transfer function under this criterion. The most commonly used criterion is the minimum mean square error criterion, which is to set the objective function of the control system to

[0065] J(n)=E[e 2 (n)]=E[d 2 (n)]+2P T W+W T RW

[0066] According to the principle of steepest descent method, the recursive filter weight coefficient is:

[0067]

[0068] Where μ is a parameter that controls the stability and convergence speed of the adaptive process, that is, the step size factor. In order to facilitate the real-time implementation of the system, the square of a single error sample e is taken. 2 The gradient of (n) is used as the gradient estimate of the mean square error, that is,

[0069]

[0070] It can be shown that the gradient estimate is the true value An unbiased estimate of replace You can get:

[0071]

[0072] The iterative process of the FXLMS algorithm is as follows:

[0073] (1) Input reference signal x(n) and error signal e(n);

[0074] (2) Calculate the secondary signal w i (n) is the i-th coefficient of the transversal filter W(z) at time n, and L is the length of the filter;

[0075] (3) Outputting the secondary signal y(n) to the secondary sound source through the power amplifier;

[0076] (4) Calculate the Filter-X signal r(n), r(n) = x(n) * h2(n);

[0077] (5). Adjust the adaptive filter weight coefficient, w i (n+1)=w(n)-2μe(n)r(n)i=0,1,2...L-1;

[0078] (6) Repeat the above process until the error signal e(n) meets the control target.

[0079] The FXLMS algorithm is based on the transient stochastic gradient descent method and uses the square error gradient to approximate the instantaneous gradient as the mean square error gradient. This iterative method always performs one-dimensional optimization along the negative gradient direction in each iteration. As long as μ is selected properly, x(n) will converge to an optimal weight vector W. * .

[0080] Step 3: Construct an improved variable step size LMS algorithm model. The specific process is as follows: In step 3,

[0081] In step 3, the iterative formula of the improved variable step-size LMS algorithm is:

[0082]

[0083] W(n+1)=W(n)-2μ(n)e(n)r(n)

[0084] Wherein, parameter α>0 controls the shape of the function; parameter β>0 controls the value range of the function. By selecting appropriate parameters α and β according to the tracking environment and the distribution of the input signal, the algorithm can have a faster convergence speed and convergence accuracy.

[0085] In this algorithm, e 2 (n) Adjusting the step size factor inevitably introduces some noise into the error signal. If the noise in the error signal has strong autocorrelation, it will cause large fluctuations in the step size factor μ, affecting the algorithm's convergence rate and accuracy. However, adjusting the step size factor using e(n)e(n-1) can effectively reduce the algorithm's sensitivity to noise signals.

[0086] In this way, the iterative formula of the variable step size adaptive filtering algorithm becomes:

[0087] μ(n)=β[1-exp(-α|e(n)e(n-1)|)]

[0088] W(n+1)=W(n)-2μ(n)e(n)r(n)

[0089] Since μ(n) and e 2 (n) is exponentially related, which makes the algorithm not have a very high convergence accuracy. It will make μ(n) fluctuate too much with the change of e(n), and the convergence accuracy of e(n) is reduced. In order to take into account the variable step size and the fluctuation of μ(n), the following adaptive filtering algorithm is proposed. The iterative formula is:

[0090] μ(n)=β[1-exp(-α·sprt(|e(n)e(-1)|))]

[0091] W(n+1)=W(n)-2μ(n)e(n)r(n)

[0092] From the purpose of the variable step size algorithm, we can know that the principles that should be followed when selecting the values of α and β are: when e(n) is large, the corresponding μ(n) should be large to ensure that the algorithm has a faster convergence speed; when e(n) is very small until it reaches the minimum value, that is, when the algorithm enters a steady state, the corresponding μ(n) should also reach the minimum value to ensure that the algorithm has a high convergence accuracy. According to the above formula, the maximum value of μ is β, so it can be seen that when β < 1 / λ maxWhen , the algorithm must converge.

[0093] Step 4: Analyze and process the noise data of the substation ventilation system, and output a reverse cancellation signal to cancel the noise of the substation ventilation system.

[0094] The following is a simulation experiment based on the improved variable step size LMS algorithm

[0095] In order to verify the effectiveness of the algorithm of the present invention, the differences before and after the improvement are compared and simulation analysis is performed using Matlab.First, the input signal x(n)=s(n)+v(n) is constructed.

[0096] Where v(n) = 0.2·randn(1,N), which is a non-stationary random signal; s(n) = cos(2π·100t) + cos(2π·200t), which is a stationary noise signal with 100Hz and 200Hz components. Then, a variable step-size adaptive control algorithm is constructed with x(n) as the reference signal. During simulation, parameters α = 5 and β = 0.04. The improved algorithm of the present invention is compared with other algorithms as follows: Figure 3 shown.

[0097] from Figure 3 As can be seen, other algorithms can quickly respond to signal changes and converge quickly, but their convergence accuracy is not very high, which does not meet the target noise reduction requirements for substation ventilation systems. However, the improved variable step-size adaptive filtering algorithm of the present invention has relatively high convergence accuracy and a fast convergence rate. As time increases, the residual error signal gradually approaches zero. Therefore, the improved algorithm of the present invention is more suitable for the low-frequency, stable, periodic signal noise of substations.

[0098] A substation ventilation noise reduction device based on an adaptive filtering algorithm includes a microphone module, a peripheral signal conditioning module, and a processing and control module. The microphone module includes a measurement sensor and a speaker; the peripheral signal conditioning module includes a signal amplifier, a high-frequency filter, an A / D converter, a D / A converter, and a low-frequency power amplifier; and the processing and control module includes a signal processor.

[0099] Depend on Figure 2 As shown in the figure, a substation ventilation noise reduction device based on an adaptive filtering algorithm works as follows: a preamplifier amplifies the noise signal collected by the primary sensor and the error signal from the error sensor. The weak signals are then amplified by a conditioning circuit, and then converted to digital signals via an A / D converter. The DSP reads the converted data and performs adaptive calculation control, sending the result to the D / A converter, which then outputs the analog signal to the speaker. This cycle repeats, ultimately reducing the noise in the substation ventilation system.

[0100] This invention analyzes the measured noise from substation ventilation systems and concludes that the high-energy frequency components of low-frequency noise in substations are relatively well-defined, primarily low-frequency components below 500 Hz. These characteristic frequency signals are relatively stable over time and do not undergo sudden changes due to external interference. This invention proposes an improved variable-step-size LMS adaptive filtering algorithm that significantly reduces noise signals while simultaneously reducing the algorithm's sensitivity to noise, effectively improving the algorithm's convergence speed and accuracy. In the practical application of substation ventilation system noise suppression, various complex on-site conditions must be considered, and noise reduction devices must be rationally arranged to achieve satisfactory noise reduction results for the substation ventilation system.

[0101] It will be apparent to those skilled in the art that the present invention is not limited to the details of the exemplary embodiments described above and that the invention can be embodied in other specific forms without departing from the spirit or essential characteristics of the invention. Therefore, the embodiments should be considered in all respects as illustrative and non-restrictive, and the scope of the invention is defined by the appended claims, not the foregoing description, and it is intended that all variations within the meaning and scope of the appended claims be encompassed. Any reference sign in a claim should not be construed as limiting the claim to which it relates.

Claims

1. A method for reducing noise in substation ventilation based on an adaptive filtering algorithm, characterized by: The method comprises the following steps: Step 1: Collect data on the noise of the substation ventilation system; Step 2: Establish the filtering-XLMS algorithm model; Step 3: Construct a model based on the improved variable step-size LMS algorithm; Step 4: Analyze and process the noise data of the substation ventilation system, and output a reverse cancellation signal to cancel the noise of the substation ventilation system.

2. The method for reducing noise in substation ventilation based on an adaptive filtering algorithm according to claim 1, wherein: In step 2, let the reference signal be x(n), the primary signal be p(n), the output signal of the secondary sound source be y(n), and the error signal after the secondary sound source and the primary noise source are superimposed be e(n); The FXLMS algorithm is used, which makes the reference signal pass through a secondary path H S (z) The same additional path is used to update the weight coefficients in the LMS algorithm, solving the problem of incorrect timing alignment between the error signal and the reference signal x(n) caused by the delay of the error signal e(n); The relationship between the reference signal and the primary signal is: x(n)=p(n)*h1(n) At the error sensor, the received primary sound field and secondary sound field signals are: d(n)=p(n)*h3(n) s(n)=y(n)*h2(n) If the length of the transversal filter is L, the secondary signal is the output of the filter: Assume that the primary noise has a local stationary characteristic, so that the adaptive filter weight coefficient can be considered to be basically unchanged within the L period, then Among them, r(n) is called the filter-X signal, and the column vector composed of it is called the filter-X signal vector. r(n)=[r(n),r(n-1),…,r(n-l+1)] T r(n)=X(n)*h2(n) Therefore, the signal received by the error sensor can be expressed as: e(n)=d(n)+s(n)=d(n)+r τ (n)W(n) Since the error signal is a random signal, in order to find the optimal weight coefficient of the filter, we should first set a criterion (objective function) to be achieved, and then derive the optimal filter transfer function under this criterion. The most commonly used criterion is the minimum mean square error criterion, which is to set the objective function of the control system to J(n)=E[e 2 (n)]=E[d 2 (n)]+2P T W+W T RW According to the principle of steepest descent method, the recursive filter weight coefficient is: Where μ is a parameter that controls the stability and convergence speed of the adaptive process, that is, the step size factor. In order to facilitate the real-time implementation of the system, the square of a single error sample e is taken. 2 The gradient of (n) is used as the gradient estimate of the mean square error, that is, It can be shown that the gradient estimate is the true value An unbiased estimate of replace You can get:

3. The method for reducing noise in substation ventilation based on an adaptive filtering algorithm according to claim 1, wherein: In step 3, the iterative formula of the improved variable step-size LMS algorithm is: μ(n)=β[1-exp(-α|e(n)| 2 )] W(n+1)=W(n)-2μ(n)e(n)r(n) Wherein, parameter α>0 controls the shape of the function; parameter β>0 controls the value range of the function. By selecting appropriate parameters α and β according to the tracking environment and the distribution of the input signal, the algorithm can have a faster convergence speed and convergence accuracy.

4. A substation ventilation noise reduction device based on an adaptive filtering algorithm, comprising a microphone module, a peripheral signal conditioning module, and a processing and control module, characterized in that: The microphone module includes a measurement sensor and a speaker; the peripheral signal conditioning module includes a signal amplifier, a high-frequency filter, an A / D converter, a D / A converter and a low-frequency power amplifier; and the processing control module includes a signal processor.

Images