An adaptive filtering method for medium voltage carrier systems

Abstract

The present invention belongs to the technical field of power line communication, and specifically discloses an adaptive filtering method suitable for a medium voltage carrier system, comprising: determining the number of sampling points that the decision system can store based on the computing power of the FPGA; storing a signal containing multiple consecutive sampling points in the FPGA as an input signal (noise floor); providing an objective function of the adaptive filter, and solving the filter coefficients in the form of partial derivatives and inverse matrices based on the input signal. The present invention eliminates the requirement of the adaptive filter for two signals by pre-storing a single-phase noise floor as an input signal. It is completely effective in scenarios such as medium voltage carrier power line communication where there is only one signal. By iteratively calculating the parameters of the pre-stored single-phase noise floor, the next frame or even several frames of data can be effectively filtered to improve the link signal-to-noise ratio. The present invention does not have high requirements for storage time, that is, it does not have high requirements for the computing power of the chip, and has a wide range of applicability.

Description

Technical Field

[0001] The present invention belongs to the technical field of power line communications, and in particular relates to an adaptive filtering method suitable for a medium voltage carrier system. Background Art

[0002] In common filtering scenarios, filtering is performed from a frequency domain perspective. Given the appropriate design parameters, a filter that meets the requirements can be easily designed. However, in more general situations, the operating environment of the filter required is time-varying, which can cause the performance of pre-designed filters to degrade or even become unusable. For example, a weak and unstable useful signal buried in strong background noise requires adaptive noise cancellation technology to extract the signal from the non-stationary and time-varying background noise.

[0003] The core of adaptive noise cancellation technology is the adaptive filter, whose parameters are controlled by an adaptive algorithm to achieve optimal filtering. Different adaptive filtering algorithms have varying convergence speeds, steady-state offsets, and algorithmic complexity. Depending on whether the adaptive algorithm is dependent on the filter output, it can be categorized as either open-loop or closed-loop. Adaptive noise cancellation technology utilizes output feedback and is therefore classified as a closed-loop algorithm. Its advantages lie in its ability to maintain optimal output despite changes in the filter input and to compensate, to some extent, for variations in filter component parameters and operational errors. Its disadvantages lie in its stability and convergence speed.

[0004] Furthermore, adaptive filtering requires simultaneous processing of at least two signals: one phase line containing both signal and noise, and the other containing only noise (or a signal with minimal amplitude). In a medium-voltage carrier system, the three-phase power lines are strongly coupled, effectively treating them as a single signal. This poses significant challenges for adaptive filtering. Summary of the Invention

[0005] To address the shortcomings or defects of the above-mentioned existing technologies, the present invention proposes an adaptive filtering method suitable for medium-voltage carrier systems. This method eliminates the mandatory requirement for adaptive filtering on two signal paths and adopts a single-phase noise extraction method for filtering, ultimately improving the signal-to-noise ratio of the link.

[0006] The technical solution of the present invention is:

[0007] An adaptive filtering method applicable to a medium voltage carrier system comprises the following steps:

[0008] S1: Calculate the number of sampling points that can be stored in the FPGA computing system. Based on the FPGA computing power, calculate the number of sampling points q that the FPGA can handle within a specified time.

[0009] S2: After the carrier machine sends the message, it switches to the receiving state. The FPGA in the carrier machine stores the received background noise signal containing q consecutive sampling points as the input signal x(n). Before the carrier machine receives a new signal, the FPGA will discard the first sampling point of the original q sampling points each time it stores a new sampling point. That is, the FPGA always stores only the latest q sampling points.

[0010] S3: Collect the current carrier signal received as the desired signal d(n); the desired signal d(n) is a superposition of noise and the useful signal, where the noise in d(n) is correlated with x(n) in S2. After passing through the adaptive filter, x(n) will be used to cancel the noise in d(n), obtaining the difference e(n) between the desired signal and the output signal of the model filter;

[0011] S4: At a given time k, the signal vector input to S2 is Get the objective function ξ based on the filter coefficient d (k);

[0012] S5: Calculate the partial derivative of the filter coefficient w(k) of the objective function as the deterministic correlation matrix R of the input signal D (k) and the deterministic cross-correlation vector P between the input signal and the desired signal D (k) in product form;

[0013] S6: Use the matrix inversion lemma to solve R D (k) The inverse matrix replaces the complex inverse operation with multiplication and division operations;

[0014] S7: Solve for P D (k), calculate w(k), and calculate the objective function ξ again d (k);

[0015] S8: Given that the lengths of messages for different services are different, the performance of the filter is further enhanced by fine-tuning q.

[0016] The beneficial effects of the present invention are as follows: by using the form of pre-stored single-phase background noise as the input signal, the requirement of the adaptive filter for two signals is eliminated. Compared with the adaptive filter calculated in real time, the performance is somewhat reduced. However, it is completely effective in scenarios such as medium-voltage carrier power line communication where there is only one signal. By iteratively calculating the parameters of the pre-stored single-phase background noise, the next frame or even several frames of data can be effectively filtered to improve the link signal-to-noise ratio. The present invention does not have high requirements for storage time, that is, it does not have high requirements for the computing power of the chip, and has a wide range of applicability. BRIEF DESCRIPTION OF THE DRAWINGS

[0017] Figure 1This is a diagram of a noise elimination method using an adaptive filtering method applicable to a medium voltage carrier system according to an embodiment of the present invention.

[0018] Figure 2 This is a flow chart of an adaptive filtering method applicable to a medium voltage carrier system according to an embodiment of the present invention. DETAILED DESCRIPTION

[0019] The following is a clear and complete description of the technical solutions in the embodiments of the present invention in conjunction with the accompanying drawings. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without making creative work are within the scope of protection of the present invention.

[0020] like Figure 2 As shown, the specific process of this embodiment is:

[0021] S1: Calculate the number of sample points that the FPGA can store. Calculate the number of sample points q that the FPGA can handle within a specified timeframe based on the FPGA's computing power.

[0022] The amount of calculation required for each sampling point is analyzed as follows: S D The calculation requires 4*n 2 +n multiplications, (3n-2)*n additions; P D The calculation requires 2n multiplications and n additions. The calculation of w(k) requires n 2 multiplications and (n-1) additions.

[0023] S2: After the carrier machine sends the message, it switches to the receiving state, and the FPGA stores the background noise signal containing q consecutive sampling points as the input signal x(n). The storage time is seconds, where Q is the sampling rate in MHz. The storage time must be greater than the time difference between the carrier machine's message transmission and the carrier machine's message reception. This ensures that the stored signal contains no carrier signal noise. If the FPGA stores q new sampling points before the carrier machine detects the carrier signal, the most recently stored signal is retained and the previous one is deleted.

[0024] S3: The carrier machine currently receives the signal as the desired signal d(n), where the noise in d(n) is correlated with x(n) in S2. Figure 1 As shown in Figure 2, after x(n) passes through the adaptive filter, it will be used to offset the noise in d(n) and obtain the difference e(n) between the expected signal and the output signal of the model filter;

[0025] S4: At a given time k, the signal vector input to step S2 Where N is the number of filters. The filter coefficient is w j(k),j=0,1,2,…,N, and adaptively adjust w j (k) is used to minimize the objective function (the sum of the squares of the difference between the expected signal and the output signal of the model filter, e(n)). The objective function is defined as:

[0026]

[0027] Where d(i) is the expected signal at time i, ε(i) = e(n) is the posterior output error at time i, and λ is an exponential weighting factor, also known as the forgetting factor, which ranges from 0 to 1. The larger its value, the smaller the contribution of old data to the coefficient update.

[0028] S5: Find the objective function The partial derivative of Transformed into a deterministic correlation matrix of the input signal and the deterministic cross-correlation vector between the input signal and the desired signal The product form of .

[0029] In order to seek d (k), find the partial derivative of w(k) and set it to 0. The formula is as follows:

[0030]

[0031] The following relationship is obtained:

[0032]

[0033] S6: Using the matrix inversion lemma, we can get the deterministic correlation matrix The inverse matrix formula is:

[0034]

[0035] As can be seen from the above formula, the complex matrix inversion operation is replaced by ordinary multiplication and division calculations, and the serial system is very suitable for this iterative calculation.

[0036] S7: According to the solution Calculate w(k) in S5 again, and obtain the objective function ξ in S4 based on the recalculated w(k) d (k), where

[0037] S8: Given that the lengths of messages for different services are different, the performance of the filter is further enhanced by fine-tuning q.

[0038] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention rather than to limit it. Ordinary technicians in the relevant field can still modify or replace the specific implementation methods of the present invention with equivalents by referring to the above embodiments. Any modifications or equivalent replacements that do not depart from the spirit and scope of the present invention are within the scope of protection of the claims of the present invention to be approved.

Claims

1. An adaptive filtering method suitable for a medium voltage carrier system, characterized in that: The specific steps include: S1: Calculate the number of sampling points that can be stored in the FPGA computing system. Based on the FPGA computing power, calculate the number of sampling points q that the FPGA can handle within a specified time. S2: After the carrier machine sends the message, it switches to the receiving state. The FPGA in the carrier machine stores the received background noise signal containing q consecutive sampling points as the input signal x(n). Before the carrier machine receives a new signal, the FPGA will discard the first sampling point of the original q sampling points each time it stores a new sampling point. That is, the FPGA always stores only the latest q sampling points. S3: The carrier receives the signal and uses it as the desired signal d(n). The desired signal d(n) is a superposition of noise and the useful signal, where the noise in d(n) is correlated with x(n) in S2. After passing through the adaptive filter, x(n) is used to cancel the noise in d(n), resulting in the difference e(n) between the desired signal and the output signal of the model filter. S4: At a given time k, input the signal vector of q sampling points stored in advance Where N is the number of filters; the filter coefficient is w j (k),j=0,1,2,…,N, and adaptively adjust w j (k) is used to minimize the sum of squares of the difference between the desired signal and the model filter output. The objective function is defined as: Where d(i) is the expected signal d(n) at time i, ε(i) = e(n) is the posterior output error at time i, and λ is an exponential weighting factor, also known as the forgetting factor, which ranges from 0 to 1. The larger its value, the smaller the contribution of old data to the coefficient update. S5: Find the filter coefficient for the objective function The partial derivative transformation of Transformed into a deterministic correlation matrix of the input signal and the deterministic cross-correlation vector between the input signal and the desired signal The product form of In order to seek d (k), find the partial derivative of w(k) and set it to 0. The formula is as follows: The following relationship is obtained: S6: Using the matrix inversion lemma, solve The inverse matrix of , replaces the complex inverse operation with multiplication and division operations; S7: Solving Calculate w(k) and calculate the objective function ξ again d (k); S8: Given that the lengths of messages for different services are different, the performance of the filter is further enhanced by fine-tuning q.

2. The adaptive filtering method applicable to a medium voltage carrier system according to claim 1, characterized in that: The amount of calculation required for each sampling point in step S1 is analyzed as follows: D The calculation requires 4*n2+n multiplications and (3n-2)*n additions; P D The calculation requires 2n multiplications and n additions; the calculation of w(k) requires n2 multiplications and (n-1) additions.

3. The adaptive filtering method applicable to a medium voltage carrier system according to claim 1, characterized in that: After the carrier machine sends the message in step S2, it switches to the receiving state, and the FPGA stores the signal x containing q consecutive sampling points as the input signal. The storage time is seconds, where Q is the sampling rate in MHZ. The storage time must be greater than the time difference between the end of the carrier machine sending the message and the time when the carrier machine receives the message to ensure that the stored signal does not contain the noise floor of the carrier signal. If the FPGA stores a new signal of q sampling points before the carrier machine detects the carrier signal, the latest storage is retained and the previous storage is deleted.

4. The adaptive filtering method applicable to a medium voltage carrier system according to claim 1, characterized in that: The step S6 uses the matrix inversion lemma to obtain the deterministic correlation matrix The inverse matrix formula is:

5. The adaptive filtering method applicable to a medium voltage carrier system according to claim 1, characterized in that: In step S7

Images