Distributed photovoltaic system non-stationary signal feature extraction method

Abstract

A non-stationary signal feature extraction method for a distributed photovoltaic system relates to the technical field of photovoltaic power generation, and comprises the following steps: acquiring a section of original signal in the distributed photovoltaic system; setting initial parameters of a decomposition layer number and a penalty factor in variational mode decomposition; variational mode decomposition is carried out on the original signal to obtain an intrinsic mode function, and a fitness function is constructed; calculating an optimal decomposition layer number and an optimal penalty factor by combining the decomposition layer number, the penalty factor and the fitness function based on a sheep flock migration optimization algorithm; performing variational mode decomposition on the original signal by using the optimal decomposition layer number and the optimal penalty factor to obtain an optimal intrinsic mode function; removing the optimal intrinsic mode function with the maximum composite entropy value, and defining the remaining optimal intrinsic mode functions as effective intrinsic mode functions; and respectively extracting the composite entropy, the time domain kurtosis, the time domain skewness and the frequency domain main frequency as signal feature parameters, so as to solve the problem of poor accuracy and stability of the non-stationary features extracted traditionally.

Description

Technical Field

[0001] The present invention relates to the technical field of photovoltaic power generation, and in particular to a method for extracting features of non-stationary signals of a distributed photovoltaic system. Background Art

[0002] In distributed photovoltaic power generation systems, since environmental conditions (such as light intensity, temperature, and shading) and grid loads are dynamically changing, the signals generated by the system often exhibit non-stationary characteristics. By extracting the characteristics of these non-stationary signals, it is helpful to classify and identify signals such as voltage and current, and improve the system's operating stability and measurement accuracy; Usually, the non-stationary signals generated by distributed photovoltaic power generation systems contain a lot of noise, including power frequency interference, electromagnetic interference, random noise, etc., which affect the accuracy of photovoltaic system feature extraction. At present, the commonly used time-frequency analysis methods and denoising methods for non-stationary signals mainly include short-time Fourier transform (STFT), empirical mode decomposition (EMD), variational mode decomposition (VMD), etc. However, these methods are susceptible to mode aliasing when facing the complex dynamic environment of photovoltaic systems, and rely on manually set parameters and lack robustness.

[0003] The Chinese patent with publication number: CN117150701A discloses a virtual power plant optimization scheduling method based on parameter-optimized variational modal decomposition. By introducing virtual power plant technology, regional wind power, photovoltaics, fluctuating loads, gas turbines and energy storage equipment are integrated into a virtual power plant. The whale algorithm is used to optimize the number of VMD modal components and penalty factors to reduce signal reconstruction errors. Based on the parameter-optimized VMD algorithm, the original power of wind power, photovoltaics, loads, etc. is decomposed into high-frequency components and low-frequency components, and then the battery energy storage and gas turbines in the VPP are differentiated for different frequency bands. Control, achieve optimal utilization of resources; however, the optimization process of this method is relatively cumbersome, and is not practical for distributed photovoltaic systems with complex dynamic environments.

[0004] Therefore, we propose a method that can stably and accurately obtain distributed photovoltaic non-stationary signals. Summary of the invention

[0005] The object of the present invention is to provide a method for extracting features of non-stationary signals of a distributed photovoltaic system, which is used to solve the problem that the accuracy and stability of the non-stationary features extracted traditionally are poor.

[0006] The present invention is achieved through the following technical solutions: A method for extracting features of non-stationary signals of distributed photovoltaic systems, specifically: Get a raw signal from a distributed photovoltaic system ; Set the initial parameters of the number of decomposition levels and penalty factors in variational mode decomposition; Perform variational mode decomposition on the original signal to obtain the intrinsic mode function and construct the fitness function; Based on the flock migration optimization algorithm, combined with the decomposition layer , Penalty Factor And fitness function, calculate the optimal number of decomposition layers and the optimal penalty factor ; Using the optimal number of decomposition levels and the optimal penalty factor For the original signal Perform variational mode decomposition and obtain The optimal intrinsic mode function; Eliminate the optimal intrinsic mode function with the largest composite entropy value, and define the remaining optimal intrinsic mode functions as effective intrinsic mode functions; Four parameters, namely, composite entropy, time domain kurtosis, time domain skewness and frequency domain main frequency, are extracted from the effective intrinsic mode function as signal characteristic parameters.

[0007] Furthermore, the steps of constructing the fitness function are specifically as follows: Original signal After variational mode decomposition, K eigenmode functions are obtained. IMF Divided into Segments, find the envelope entropy of each segment and permutation entropy , , n; Calculate the discretization index of envelope entropy and permutation entropy; Count the joint frequency distribution of all discretized indices of envelope entropy and permutation entropy; Calculate the composite entropy from the joint frequency distribution; The fitness function is calculated based on the compound entropy.

[0008] Further, the steps of calculating the discretization index of the envelope entropy and the permutation entropy are specifically as follows: Calculate the interval interval of envelope entropy and get envelope entropy intervals; The entropy values ​​of each envelope , ,…, Map to the corresponding envelope entropy interval and determine the Envelope Entropy Discretization index of The value of The number of the mapped envelope entropy interval; Calculate the interval interval of permutation entropy and obtain N intervals of permutation entropy; The entropy value of each arrangement , ,…, Map to the corresponding interval and determine the jth permutation entropy value Discretization index of The value of The number of the mapped permutation entropy interval.

[0009] Furthermore, the interval calculation formula of the envelope entropy is specifically: ; Where: is the number of intervals; It is IMF The maximum envelope entropy value; It is IMF Minimum envelope entropy value.

[0010] Furthermore, the envelope entropy The intervals are: .

[0011] Furthermore, the interval calculation formula of the permutation entropy is specifically: ; Where: is the number of intervals; It is IMF The maximum permutation entropy value; It is IMF The minimum permutation entropy value.

[0012] Furthermore, the permutation entropy The intervals are: .

[0013] Furthermore, the calculation formula of the joint frequency distribution is: ; In the formula, is the ith eigenmode function Value is , and any of the i-th eigenmode functions Value is A combination of is the number of times the combination occurs, is the probability of a combination occurring.

[0014] Furthermore, the calculation formula of the composite entropy is: .

[0015] Furthermore, the calculation formula of the fitness function is: .

[0016] The technical solution of the present invention has at least the following advantages and beneficial effects: The invention discloses a method for extracting features of non-stationary signals of a distributed photovoltaic system. Based on envelope entropy and permutation entropy, composite entropy is used as a fitness function for optimizing a flock migration optimization algorithm, which helps to improve the optimization effect and the stability of non-stationary feature extraction.

[0017] The key parameters of variational mode decomposition are optimized using the flock migration optimization algorithm, which avoids the limitations brought by manually setting parameters, improves the accuracy of variational mode decomposition in generating intrinsic mode function components, and helps to increase the effectiveness of signal feature extraction. BRIEF DESCRIPTION OF THE DRAWINGS

[0018] Figure 1 A method flow chart of the present invention; Figure 2 Schematic diagram of original signal according to an embodiment of the present invention; Figure 3 It is a schematic diagram of the intrinsic mode function result of an embodiment of the present invention; Figure 4 It is a schematic diagram of the optimal intrinsic mode function result of an embodiment of the present invention. DETAILED DESCRIPTION

[0019] In order to make the purpose, technical solutions and advantages of the embodiments of the present invention clearer, the technical solutions in the embodiments of the present invention will be clearly and completely described below in conjunction with the drawings in the embodiments of the present invention. Obviously, the described embodiments are part of the embodiments of the present invention, not all of the embodiments. Generally, the components of the embodiments of the present invention described and shown in the drawings here can be arranged and designed in various different configurations.

[0020] Example 1 like Figure 1 A method for extracting non-stationary signal features of a distributed photovoltaic system is shown in FIG. , specifically: Get a raw signal from a distributed photovoltaic system ; Set the initial parameters of the number of decomposition levels and penalty factors in variational mode decomposition; Construct fitness function; Based on the flock migration optimization algorithm, combined with the decomposition layer , Penalty Factor And fitness function, calculate the optimal number of decomposition layers and the optimal penalty factor ; The number of decomposition layers and penalty factor It is the input variable of the flock migration optimization algorithm, and the fitness function is the optimization fitness of the flock migration optimization algorithm; The sheep migration optimization algorithm is used to automatically optimize the key parameters of variational mode decomposition to improve the accuracy of signal decomposition; the sheep migration optimization algorithm is a meta-heuristic algorithm that simulates the behavior of sheep following the leader, avoiding danger and looking for resources during migration. Through the combination of global search and local adjustment, the algorithm can efficiently find the best in the parameter space; in this method, the fitness function based on compound entropy is used as the optimization target, and the algorithm iteratively updates the parameter combination to finally find the optimal number of decomposition layers and the optimal penalty factor that minimizes the compound entropy; it avoids the limitations of manually setting parameters, solves the problem of traditional methods relying on empirical parameter adjustment, and improves the robustness of variational mode decomposition through global optimization, ensuring that the generated intrinsic mode function can more accurately reflect the true characteristics of the signal.

[0021] Using the optimal number of decomposition levels and the optimal penalty factor For the original signal Perform variational mode decomposition and obtain The optimal eigenmode functions are , 、···、 ; The optimal intrinsic mode function is used to obtain the intrinsic mode function components of the original signal through variational mode decomposition under the optimized parameters, providing a basis for subsequent feature extraction; and variational mode decomposition decomposes the signal into multiple narrowband intrinsic mode functions through constrained variational problems, each of which has a center frequency and a limited bandwidth in the frequency domain; and the obtained The optimal intrinsic mode function avoids over-decomposition or under-decomposition, ensuring that the signal characteristics are fully captured. The penalty factor The optimization balances the bandwidth constraint and the reconstruction error, and improves the physical meaning and interpretability of the intrinsic mode function.

[0022] Eliminate the optimal intrinsic mode function with the largest composite entropy value, and define the remaining optimal intrinsic mode functions as effective intrinsic mode functions; These effective intrinsic mode functions are the selected intrinsic mode functions with smaller composite entropy values, which can remove noise components and retain effective signal components. Since composite entropy combines envelope entropy and permutation entropy, the smaller the entropy value of composite entropy, the more stable the signal and the less noise. By selecting the IMF with the lowest entropy value, the noise-dominated mode can be eliminated; it has the effect of reducing the interference of noise on feature extraction, improving the reliability of subsequent parameter calculation, reducing the processing of redundant modes, and improving the overall efficiency of the algorithm.

[0023] The four parameters of composite entropy, time domain kurtosis, time domain skewness and frequency domain main frequency are extracted from the effective intrinsic mode function as signal characteristic parameters; Through the comprehensive extraction of multi-dimensional features (including composite entropy, time domain kurtosis, time domain skewness and frequency domain main frequency), the characteristics of non-stationary signals can be fully described, thereby providing highly discriminative feature inputs for subsequent signal classification, fault diagnosis or system optimization, and improving the accuracy and stability of photovoltaic system monitoring; The functions of each feature are as follows: composite entropy is used to quantify the non-stationarity of the signal and distinguish noise from effective components; time domain kurtosis is used to detect whether there are shocks or mutations in the signal, reflecting transient characteristics; time domain skewness is used to determine whether the signal tends to positive / negative amplitudes, revealing asymmetric fluctuation patterns; frequency domain main frequency is used to capture the main periodic characteristics of the signal and assist in analyzing power grid harmonics or environmental interference.

[0024] Example 2 As an embodiment, the steps of constructing the fitness function are specifically as follows: Original signal After variational mode decomposition, K eigenmode functions are obtained. IMF Divided into Segments, find the envelope entropy of each segment and permutation entropy , , n; And the envelope entropy reflects the complexity of signal amplitude, and the permutation entropy reflects the randomness of signal timing; Calculate the discretization index of envelope entropy and permutation entropy; Count the joint frequency distribution of all discretized indices of envelope entropy and permutation entropy; That is, by quantifying the distribution law of the discrete index combination of envelope entropy and permutation entropy, the correlation between the two entropies in the signal segmentation is described, providing a data basis for calculating the composite entropy; and revealing the coordinated change law of envelope entropy and permutation entropy in different signal segments, avoiding the limitations of a single entropy indicator, providing joint distribution information for the composite entropy, and enhancing the comprehensive characterization ability of the features.

[0025] Calculate the composite entropy from the joint frequency distribution; The composite entropy is the comprehensive entropy value of the signal segment calculated based on the joint frequency distribution. It can measure the complexity and randomness of the signal and is used to screen effective intrinsic mode functions and optimize variational mode decomposition parameters. It improves the noise suppression capability by combining the dual information of envelope entropy and permutation entropy. As the core indicator of the fitness function, it directly drives parameter optimization and ensures that the decomposed intrinsic mode function is more physically meaningful.

[0026] Calculate the fitness function based on the composite entropy; This fitness function takes the composite entropy as the optimization target, and then uses the flock migration algorithm to find the variational mode decomposition parameters that minimize the composite entropy; and by minimizing the composite entropy, it automatically screens out the intrinsic mode function with the least noise and the highest stability, improves the feature extraction accuracy, and combines the global optimization algorithm to avoid local optimality, ensuring that the parameters adapt to complex dynamic environments.

[0027] In addition, the steps for calculating the discretization index of the envelope entropy and the permutation entropy are specifically: Calculate the interval interval of envelope entropy and get envelope entropy intervals; The entropy values ​​of each envelope , ,…, Map to the corresponding envelope entropy interval and determine the Envelope Entropy Discretization index of The value of The number of the mapped envelope entropy interval; Calculate the interval interval of permutation entropy and obtain N intervals of permutation entropy; The entropy value of each arrangement , ,…, Map to the corresponding interval and determine the jth permutation entropy value Discretization index of The value of The number of the mapped permutation entropy interval.

[0028] In addition, the interval calculation formula of the envelope entropy is specifically: ; Where: is the number of intervals; It is IMF The maximum envelope entropy value; It is IMF Minimum envelope entropy value.

[0029] The envelope entropy The intervals are: .

[0030] According to the needs, the interval calculation formula of the permutation entropy is specifically: ; Where: is the number of intervals; It is IMF The maximum permutation entropy value; It is IMF The minimum permutation entropy value.

[0031] The permutation entropy The intervals are: .

[0032] In addition, the calculation formula of the joint frequency distribution is: ; In the formula, is any of the ith eigenmode functions Value is , and any of the i-th eigenmode functions Value is A combination of is the number of times the combination occurs, is the probability of a combination occurring.

[0033] The calculation formula of the composite entropy is: .

[0034] The calculation formula of the fitness function is: .

[0035] Example 3 According to a specific embodiment of this method: Original voltage signal like Figure 2 As shown, set the number of decomposition levels of variational mode decomposition and penalty factor The initial parameters of are 4 and 1000 respectively; The decomposed eigenmode function is as follows: Figure 3 As shown in Figure 3-1 , 3-2 is , 3-3 is , 3-4 is , by For example, It is divided into 6 segments, and the envelope entropy value of each segment is 3.14387, 4.29828, 4.24469, 4.27760, 4.26891, and 2.93384, and the permutation entropy value of each segment is 0.84349, 0.84574, 0.84383, 0.84316, 0.84003, and 0.83864.

[0036] The number of envelope entropy intervals is set to 7, so the interval interval of envelope entropy is: (4.29828-2.93384) / 7=0.19492; Therefore, the 7 intervals of envelope entropy are: [2.93384, 3.12876), [3.12876, 3.32368), [3.32368, 3.51860), [3.51860, 3.71352), [3.71352, 3.90844), [3.90844, 4.10336), [4.10336, 4.29828]; The 6 envelope entropy values ​​are in the 2nd, 7th, 7th, 7th, 7th, and 1st envelope entropy intervals respectively, so the discretization indexes of the 6 envelope entropy values ​​are: 2, 7, 7, 7, 7th, and 1st respectively; The number of permutation entropy intervals is set to 7, so the interval interval of permutation entropy is: (0.84574-0.83864) / 7=0.00101 Therefore, the 7 intervals of permutation entropy are: [0.83864 0.83965), [0.83965, 0.84067), [0.84067, 0.84168), [0.84168, 0.84270), [0.84270, 0.84371), [0.84371, 0.84473), [0.84473, 0.84574].

[0037] The 6 permutation entropy values ​​are in the 5th, 7th, 6th, 5th, 2nd, and 1st permutation entropy intervals respectively, so the discretization indexes of the 6 permutation entropy values ​​are: 5, 7, 6, 5, 2, and 1 respectively.

[0038] Taking the case where the discretization index of the envelope entropy value is 7 and the discretization index of the permutation entropy value is 5 as an example: the discretization index of the 2nd, 3rd, 4th, and 5th envelope entropy values ​​is 7, and the discretization index of the 1st and 4th permutation entropy values ​​is 5, so (I EE,1 =1,I PE,1 =5) is 4×2=8, so its probability is: P(I EE,1=1,I PE,1 =5)=8 / (6 2 )=0.222; The joint frequency distribution is shown in the following table:

[0039] but The composite entropy is:

[0040] Similarly, we can get , , The compound entropies are 2.5721, 2.6593, and 3.1214 respectively, so the fitness function value at this time is:

[0041] Repeat the above steps using the flock migration optimization algorithm to adjust the number of decomposition layers. and penalty factor Optimize and get the optimal number of decomposition layers and the optimal penalty factor 3 and 671 respectively.

[0042] Optimal number of decomposition levels and the optimal penalty factor The optimal eigenmode function is obtained as Figure 4 As shown, 4-1, 4-2 and 4-3 are respectively , and ; After calculation The composite entropy of is the largest, so it is removed and retained and As the effective intrinsic mode function, the non-stationary signal characteristic parameters are further obtained as shown in the following table.

[0043]

[0044] The above are only preferred embodiments of the present invention and are not intended to limit the present invention. For those skilled in the art, the present invention may have various modifications and variations. Any modification, equivalent replacement, improvement, etc. made within the spirit and principle of the present invention shall be included in the protection scope of the present invention.

Claims

1. A method for extracting non-stationary signal features of a distributed photovoltaic system, characterized in that: Specifically: Obtain a section of original signal in a distributed photovoltaic system; Set the initial parameters of the number of decomposition levels and penalty factors in variational mode decomposition; Perform variational mode decomposition on the original signal to obtain the intrinsic mode function and construct the fitness function; Based on the flock migration optimization algorithm, the optimal number of decomposition layers and the optimal penalty factor are calculated by combining the number of decomposition layers, penalty factor and fitness function; Use the optimal number of decomposition layers and the optimal penalty factor to decompose the original signal Perform variational mode decomposition to obtain multiple optimal eigenmode functions; Eliminate the optimal intrinsic mode function with the largest composite entropy value, and define the remaining optimal intrinsic mode functions as effective intrinsic mode functions; Four parameters, namely, composite entropy, time domain kurtosis, time domain skewness and frequency domain main frequency, are extracted from the effective intrinsic mode function as signal characteristic parameters.

2. The method for extracting non-stationary signal features of a distributed photovoltaic system according to claim 1, characterized in that: The steps for constructing the fitness function are specifically as follows: Original signal After variational mode decomposition, K eigenmode functions are obtained. IMF Divided into Segments, find the envelope entropy of each segment and permutation entropy , , n; Calculate the discretization index of envelope entropy and permutation entropy; Count the joint frequency distribution of all discretized indices of envelope entropy and permutation entropy; Calculate the composite entropy from the joint frequency distribution; The fitness function is calculated based on the compound entropy.

3. The method for extracting non-stationary signal features of a distributed photovoltaic system according to claim 2, characterized in that: The steps for calculating the discretization index of envelope entropy and permutation entropy are as follows: Calculate the interval interval of envelope entropy and get envelope entropy intervals; The entropy values ​​of each envelope , ,…, Map to the corresponding envelope entropy interval and determine the Envelope Entropy Discretization index of The value of The number of the mapped envelope entropy interval; Calculate the interval interval of permutation entropy and obtain N intervals of permutation entropy; The entropy value of each arrangement , ,…, Map to the corresponding interval and determine the jth permutation entropy value Discretization index of The value of The number of the mapped permutation entropy interval.

4. The method for extracting non-stationary signal features of a distributed photovoltaic system according to claim 3, characterized in that: The envelope entropy interval calculation formula is specifically: ； Where: is the number of intervals; It is IMF The maximum envelope entropy value; It is IMF Minimum envelope entropy value.

5. The method for extracting non-stationary signal features of a distributed photovoltaic system according to claim 4, characterized in that: The envelope entropy The intervals are: 。 6. The method for extracting non-stationary signal features of a distributed photovoltaic system according to claim 3, characterized in that: The interval calculation formula of the permutation entropy is specifically: ； Where: is the number of intervals; It is IMF The maximum permutation entropy value; It is IMF The minimum permutation entropy value.

7. The method for extracting non-stationary signal features of a distributed photovoltaic system according to claim 6, characterized in that: The permutation entropy The intervals are: 。 8. The method for extracting non-stationary signal features of a distributed photovoltaic system according to claim 3, characterized in that: The calculation formula of the joint frequency distribution is: ； In the formula, is any of the ith eigenmode functions Value is , and any of the i-th eigenmode functions Value is A combination of is the number of times the combination occurs, is the probability of a combination occurring.

9. The method for extracting non-stationary signal features of a distributed photovoltaic system according to claim 8, characterized in that: The calculation formula of the composite entropy is: 。 10. The method for extracting non-stationary signal features of a distributed photovoltaic system according to claim 9, characterized in that: The calculation formula of the fitness function is: 。

Images