Intelligent identification method of generator parameters based on neural network

Abstract

The present application relates to the technical field of generator speed measurement and control, and specifically to a method for intelligent identification of generator parameters based on a neural network, the method comprising: obtaining the generator parameters at the nodes of a power generation grid; obtaining a grid fluctuation power vector according to the difference of the generator parameters at all times; performing wavelet decomposition on the grid fluctuation power vector to obtain each grid fluctuation power decomposition vector; obtaining the grid fluctuation characteristic value at each time according to the local numerical distribution of the elements in all grid fluctuation power decomposition vectors corresponding to each time; and predicting the diesel generator control speed at the next time according to the generator parameters at all times and the grid fluctuation characteristic value. The present application can realize the intelligent identification and control of the generator speed parameters, and improve the identification and control accuracy of the generator speed.

Description

Technical Field

[0001] The present application relates to the technical field of generator speed measurement and control, and in particular to a method for intelligently identifying generator parameters based on a neural network. Background Art

[0002] In the generator power generation system, the core component is the PCS power electronic energy storage power conversion device, also known as the energy storage controller and inverter. In the generator power generation system, PCS participates in auxiliary services such as grid frequency regulation and voltage regulation, peak shaving and valley filling. It stabilizes the load fluctuation of the grid by controlling the output active power and reactive power, thereby effectively adjusting the intelligent parameters of the generator.

[0003] At present, the virtual synchronous machine technology in the voltage source PCS grid connection is often used to achieve the grid connection of the PCS simulated synchronous generator droop characteristics by controlling the output active power and reactive power. However, the virtual synchronous machine technology has poor adaptability to the power grid, especially in the weak power grid dominated by diesel generators. The insufficient output power of the diesel generator may cause the PCS to be unable to effectively regulate the power grid, affecting the stability of the power grid, making the recognition and control accuracy of the speed parameters in the diesel generator parameters poor, and then the diesel engine will run away or overload. Summary of the invention

[0004] In order to solve the above technical problems, the present application provides a generator parameter intelligent identification method based on neural network to solve the existing problems.

[0005] The neural network-based intelligent generator parameter identification method of the present application adopts the following technical solutions:

[0006] An embodiment of the present application provides a method for intelligently identifying generator parameters based on a neural network, the method comprising the following steps:

[0007] S1, obtaining generator parameters at a node of a power generation grid; the generator parameters include: output power, rotation speed, grid power and PCS power of a diesel generator;

[0008] S2, according to the difference between the power grid power and the PCS power at all times, the power grid fluctuation power vector is obtained; the power grid fluctuation power vector is subjected to wavelet decomposition to obtain the decomposition vector of each power grid fluctuation power;

[0009] S3, according to the local numerical distribution of the elements in the power decomposition vector of all power grid fluctuations corresponding to each moment, the power grid fluctuation characteristic value at each moment is obtained, specifically including:

[0010] S31, obtaining a power impact weight according to the numerical distribution of all elements in the power grid fluctuation power decomposition vector;

[0011] S32, obtaining a power generation information association weight according to a correlation between the output power of the diesel generator at all times and the power grid fluctuation power decomposition vector;

[0012] S33, determining a power grid fluctuation power prediction weight according to the power impact weight and the power generation information association weight;

[0013] S34, obtaining the power grid fluctuation characteristic value at each moment according to the numerical distribution of the neighborhood elements of all power grid fluctuation power decomposition vectors at each moment and the power grid fluctuation power prediction weight;

[0014] S4, predicting the diesel generator control speed at the next moment according to the generator parameters and grid fluctuation characteristic values ​​at all moments.

[0015] Furthermore, the grid fluctuation power vector is obtained according to the difference between the grid power and the PCS power at all times, including:

[0016] The difference between the grid power and the PCS power at each moment is calculated and recorded as the fluctuating power at each moment; the vector composed of the fluctuating power at all moments is recorded as the grid fluctuating power vector.

[0017] Furthermore, the power impact weight is obtained according to the numerical distribution of all elements in the power grid fluctuation power decomposition vector, including:

[0018] According to the values ​​of all elements in the neighborhood of each element in the power grid fluctuation power decomposition vector, the power imbalance factor of each element is obtained;

[0019] The mean and standard deviation of the power imbalance factors of all elements in the power grid fluctuation power decomposition vector are calculated; and the forward fusion result of the mean and standard deviation is used as the power impact weight of the power grid fluctuation power decomposition vector.

[0020] Furthermore, the power imbalance factor of each element is obtained according to the values ​​of all elements in the neighborhood of each element in the power grid fluctuation power decomposition vector, including:

[0021] For each power grid fluctuation power decomposition vector, perform fast Fourier transform on the power grid fluctuation power decomposition vector to obtain the center frequency of the power grid fluctuation power decomposition vector; and obtain the average period of the power grid fluctuation power decomposition vector according to the center frequency;

[0022] The value of the average period is used as the length of the preset window, and the sum of all elements in the preset window centered on each element in the power grid fluctuation power decomposition vector is calculated as the power imbalance factor of each element.

[0023] Furthermore, the average period is the ratio of the value 2π to the central frequency of the grid fluctuation power decomposition vector; and π is pi.

[0024] Furthermore, the method for obtaining the power generation information association weight includes:

[0025] The vector composed of the output power of the diesel generator at all times is recorded as the motor power vector;

[0026] The normalized value of the power imbalance factor of each element in the power grid fluctuation power decomposition vector is used as the complementary information weight of each element;

[0027] The element value of each element in the motor power vector is replaced by the opposite number of the element value of each element to obtain the motor power symmetric vector; the complementary information weight of each element in the power grid fluctuation power decomposition vector is used as the weight, and the covariance of each power grid fluctuation power decomposition vector and the motor power symmetric vector is calculated, which is recorded as the power generation information association weight.

[0028] Further, the method of determining the power grid fluctuation power prediction weight according to the power impact weight and the power generation information association weight includes:

[0029] The normalized value of the product of the power impact weight and the power generation information associated weight is used as the grid fluctuation power prediction weight.

[0030] Furthermore, the method for obtaining the power grid fluctuation characteristic value at each moment includes:

[0031] According to the numerical distribution of all elements in the neighborhood of each element in the power fluctuation decomposition vector of each power grid, the signal amplitude of each element at the corresponding time is obtained;

[0032] The ratio of the signal amplitude to the central frequency of the power grid fluctuation power decomposition vector is recorded as the fluctuation amplitude of the power grid fluctuation power decomposition vector at the time corresponding to the target element;

[0033] The grid fluctuation power prediction weight of each grid fluctuation power decomposition vector is used as the weight, and the weighted average value of the fluctuation amplitude of all grid fluctuation power decomposition vectors at each moment is calculated, which is recorded as the grid fluctuation characteristic value at each moment.

[0034] Furthermore, the method for obtaining the signal amplitude of each element at a corresponding moment includes:

[0035] For each power grid fluctuation power decomposition vector, any element in the power grid fluctuation power decomposition vector is recorded as the target element; before calculating the target element, the mean of the absolute values ​​of all elements in the preset window including the target element is recorded as the signal amplitude at the corresponding moment of the target element.

[0036] Further, the method of predicting the diesel generator control speed at the next moment according to the generator parameters and the grid fluctuation characteristic values ​​at all moments includes:

[0037] Obtain the preset diesel generator control speed at each moment; use the generator parameters at all moments and the preset diesel generator control speed as training data sets, input them into neural network training, and obtain a trained neural network model;

[0038] The generator parameters at the current moment are input into the trained neural network model to obtain the diesel generator control speed at the next moment.

[0039] This application has at least the following beneficial effects:

[0040] The present application analyzes the influence of virtual synchronous machine on the stability of power grid, extracts the frequency domain information of power grid fluctuation power by wavelet decomposition algorithm, analyzes the power grid power fluctuation imbalance represented by frequency domain information in different decomposition signals to obtain power imbalance factor, and then obtains power influence weight, which reflects the power grid power fluctuation and quantifies the importance of power grid fluctuation power decomposition vector to predict power grid fluctuation; further, the power imbalance factor is used as weight, and the complementarity between motor power vector and decomposition vector is compared to obtain power generation information association weight, which reflects the complementarity between power grid fluctuation power decomposition vector and motor power vector, and the power grid fluctuation power prediction weight is obtained by combining power influence weight and power generation information association weight, and accurately judges the importance of power grid fluctuation power decomposition vector to generator parameter control; finally, the power grid fluctuation characteristic value is obtained by combining the fluctuation frequency and fluctuation amplitude of the decomposition signal, and the fluctuation characteristics related to generator parameters in the power grid information at the corresponding time are extracted; finally, the fluctuation characteristics and generator parameters are input into the neural network together, and the neural network model is trained to control the generator speed. The present application adjusts the speed of diesel generator in real time according to the fluctuation state of the power grid, and improves the recognition and control accuracy of the generator speed. BRIEF DESCRIPTION OF THE DRAWINGS

[0041] In order to more clearly illustrate the technical solutions and advantages in the embodiments of the present application or the prior art, the drawings required for use in the embodiments or the prior art descriptions are briefly introduced below. Obviously, the drawings described below are only some embodiments of the present application. For ordinary technicians in this field, other drawings can be obtained based on these drawings without creative work.

[0042] Figure 1 A flowchart of the steps of the method for intelligent identification of generator parameters based on neural network provided in this application;

[0043] Figure 2 Schematic diagram of power grid fluctuation power vector provided for this application;

[0044] Figure 3 A schematic diagram of the power grid fluctuation power decomposition vector provided for this application;

[0045] Figure 4 This is the fast Fourier transform image of the power grid fluctuation power decomposition vector provided in this application. DETAILED DESCRIPTION

[0046] The following will be combined with the drawings in the embodiments of the present application to clearly and completely describe the technical solutions in the embodiments of the present application. Obviously, the described embodiments are only part of the embodiments of the present application, not all of the embodiments. In the absence of conflict, the embodiments of the present application and the technical features in the embodiments can be combined with each other. Based on the embodiments in the present application, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of this application.

[0047] Unless defined otherwise, all technical and scientific terms used herein have the same meaning as commonly understood by one of ordinary skill in the art to which this application belongs.

[0048] The specific scheme of the method for intelligent identification of generator parameters based on neural network provided by the present application is described in detail below with reference to the accompanying drawings.

[0049] An embodiment of the present application provides a method for intelligently identifying generator parameters based on a neural network. Specifically, the following method for intelligently identifying generator parameters based on a neural network is provided. Figure 1 , the method comprises the following steps:

[0050] S1, obtaining the parameters of the generator at the node of the power generation grid.

[0051] This embodiment aims to determine the fluctuation of the grid load condition according to the current grid load condition, predict the degree to which the grid needs the PCS to adjust the power, and then the generator speed in the power generation diesel unit, combined with the PCS equipment to store energy for the grid in advance, to avoid the generator running away or overloading due to the excessive power required by the grid, and at the same time improve the stability of the grid.

[0052] Since it is necessary to judge the load condition of the power grid, the load data of the power grid, that is, the power size of the power grid, is obtained in real time at the node of the power generation grid, which is recorded as the power grid.

[0053] The output power of the PCS is recorded as PCS power. PCS power refers to the output power of the PCS to the power grid. The power can be positive or negative. When the power is positive, it means that the PCS provides power support to the power grid. When the power is negative, it means that the PCS extracts energy from the power grid to ensure the stability of the power grid.

[0054] The output power of the diesel generator, the grid power and the PCS power at each moment are collected, and a speed sensor is installed at the diesel generator to obtain the speed of the diesel generator at each moment; the data sampling period is set to 1 second, and a total of N = 10,000 data are obtained.

[0055] S2, according to the difference between the power grid power and the PCS power at all times, the power grid fluctuation power vector is obtained; the power grid fluctuation power vector is subjected to wavelet decomposition to obtain each power grid fluctuation power decomposition vector.

[0056] The difference between the grid power and the PCS power at each moment is calculated and recorded as the fluctuating power at each moment; it represents the actual power size of the grid at each moment without PCS auxiliary control, and can characterize the fluctuation of grid power.

[0057] The historical data of power grid fluctuation power is further processed, and the vector composed of the fluctuation power at all times is recorded as the power grid fluctuation power vector; Figure 2 As shown, the nth element of the power grid fluctuation power vector represents the power grid fluctuation value at time n, and represents the power grid power fluctuation.

[0058] Since this embodiment needs to predict the power required by the power grid, the energy storage unit management of the PCS can be used in advance to ensure that the power grid has sufficient power supply at the peak when the PCS needs to provide power, and the PCS can have sufficient energy storage to provide power for the power grid. In order to predict the power required by the power grid, it is necessary to extract feature information that is conducive to predicting power from the power information of the power grid, so the frequency domain features of the power grid fluctuation power vector are extracted.

[0059] Specifically, the power grid fluctuation power vector is subjected to wavelet decomposition to obtain a preset number of power grid fluctuation power decomposition vectors; in this embodiment, the power grid fluctuation power vector is decomposed by a wavelet decomposition algorithm, and a total of M=7 decompositions are performed to obtain M decomposition signals, wherein the mth signal is recorded as the mth power grid fluctuation power decomposition vector; the wavelet decomposition algorithm is a common technology in the field of electrical signal processing, and the specific decomposition process is not repeated. Figure 3 As shown, it is the result of decomposition of the power grid fluctuation vector, where different images are images of power grid fluctuation vectors in different frequency domains. As the number of signal decompositions increases, the decomposed signal frequency becomes higher and higher. Fluctuation vectors with different frequencies may represent different factors that cause power grid power fluctuations.

[0060] S3, obtaining the power grid fluctuation characteristic value at each moment according to the local numerical distribution of the elements in all power grid fluctuation power decomposition vectors corresponding to each moment.

[0061] Among these M decomposition vectors, the fluctuation of any one of them represents the power imbalance of the power grid. The stronger the fluctuation, the more serious the imbalance. When the power is unbalanced, the PCS needs to adjust the power grid. Therefore, the more information that affects the PCS control is contained in the decomposition vector with a strong imbalance. In the decomposition vector, the more similar the area of ​​the graph above the x-axis and the area of ​​the graph below the x-axis are in the graph formed by the waveform of the vector and the x-axis, the more similar the grid can complete the self-balance of power in a shorter time span, and the less impact it has on the energy storage management of the PCS energy storage unit.

[0062] S31, obtaining a power impact weight according to the numerical distribution of all elements in the power grid fluctuation power decomposition vector.

[0063] The power impact weight of the mth power grid fluctuation power decomposition vector is calculated. The more unbalanced the fluctuation power of the mth power grid fluctuation power decomposition vector in a short period of time, the greater the corresponding power impact weight, and the better the prediction effect of the decomposition vector on the power grid power fluctuation. The power balance degree of the mth power grid fluctuation power decomposition vector is judged by taking the area enclosed by the window calculation curve.

[0064] Specifically, for each power grid fluctuation power decomposition vector, the power grid fluctuation power decomposition vector is subjected to fast Fourier transform to obtain the center frequency of the power grid fluctuation power decomposition vector; the average period of the power grid fluctuation power decomposition vector is obtained according to the center frequency; the average period is the ratio of 2π to the center frequency of the power grid fluctuation power decomposition vector; wherein the product of the center frequency and period obtained by fast Fourier transform is equal to 2π; π is pi. Figure 4 As shown, it is a fast Fourier transform image of the power grid fluctuation decomposition vector, wherein the horizontal axis is frequency and the vertical axis is amplitude, which represents the frequency spectrum range of the decomposition vector, wherein the frequency concentration part is the main frequency of the decomposition vector.

[0065] The value of the averaging period is used as the length of the preset window, and the sum of all elements in the preset window centered on each element in the power grid fluctuation power decomposition vector is calculated as the power imbalance factor of each element; the mean and standard deviation of the power imbalance factors of all elements in the power grid fluctuation power decomposition vector are calculated; and the forward fusion result of the mean and standard deviation is used as the power impact weight of the power grid fluctuation power decomposition vector.

[0066] It should be noted that forward fusion means that the fused data and the fused result have a same-direction changing relationship, where the larger the fused data, the larger the fused result. Forward fusion can specifically be a multiplication relationship, an addition relationship, etc., which is determined by the actual application and this application does not impose any special restrictions.

[0067] As an embodiment of the present application, the power impact weight of the grid fluctuation power decomposition vector is the product of the mean and the standard deviation.

[0068] The power imbalance factor can be approximated by calculating the area of ​​the figure enclosed by the grid fluctuation power decomposition vector curve and the x-axis in the window. The area of ​​the figure above the x-axis is positive, and the area below the x-axis is negative. The sum of the areas of all figures is recorded as the power imbalance factor of the corresponding element. The smaller the value, the better the power balance at the element position. The mean and standard deviation of the power imbalance factors of all elements are further calculated. The larger the absolute value of the mean, the more unbalanced the power of the grid fluctuation power decomposition vector. The larger the standard deviation, the greater the fluctuation of the power imbalance factor of the mth decomposition vector, the longer the time span for the grid to balance power, the more unbalanced the power, and the more it needs to be stabilized by PCS control. Finally, the absolute value of the mean and the standard deviation are multiplied to obtain the power influence weight. The larger the value, the more important the grid fluctuation power decomposition vector is for predicting grid fluctuations.

[0069] S32, obtaining the power generation information association weight according to the correlation between the output power of the diesel generator at all times and the power grid fluctuation power decomposition vector.

[0070] Since the purpose of predicting power grid fluctuations is to control the control parameters of the generator, it is necessary to obtain the characteristic data portion with a high correlation with the diesel generator from the extracted power grid fluctuation prediction information.

[0071] Therefore, the historical data of the output power of the diesel generator is first obtained for processing, and the vector composed of the output power of the diesel generator at all times is recorded as the motor power vector; the nth element represents the output power of the diesel generator at time n.

[0072] The power fluctuation decomposition vectors of each power grid are further compared with the motor power vector. The more complementary the waveforms of the two vectors are, the more it means that when the power grid power fluctuates in the power fluctuation decomposition vectors of each power grid, the more it is necessary to stabilize the power grid through direct power generation of the generator. The higher the correlation between the power grid power fluctuation information and the motor at this time, the more generator control information is contained in the power fluctuation decomposition vectors of each power grid, and the more important it is to control the generator parameters.

[0073] When judging the complementarity of two vector waveforms, if the absolute value of the power imbalance factor of the nth element of the grid fluctuation power decomposition vector is larger, it means that the grid stability is completed by charging and discharging the PCS energy storage unit at the nth element position, rather than using the generator to generate power for grid stability. In this case, the complementarity between the motor power vector and the decomposition vector at the nth moment cannot be used as a basis for judging the degree of correlation between the mth decomposition signal and the generator parameters.

[0074] Therefore, the normalized value of the power imbalance factor of each element in the grid fluctuation power decomposition vector is used as the complementary information weight of each element.

[0075] In order to characterize the complementarity between the mth decomposition vector and the motor power vector, a power generation information association weight is constructed. The larger the weight value, the more significant the complementarity between the mth decomposition signal and the motor power vector, and the more the information of the mth decomposition vector should be used to control the motor parameters.

[0076] The calculation method of the power generation information association weight is as follows: replace the element value of each element in the motor power vector with the opposite number of the element value of each element to obtain the motor power symmetric vector; that is, complete the symmetric processing of the motor power vector about the x-axis; further, take the complementary information weight of each element in the power grid fluctuation power decomposition vector as the weight, calculate the covariance of each power grid fluctuation power decomposition vector and the motor power symmetric vector, and record it as the power generation information association weight;

[0077] Taking the complementary information weight as the weight, the greater the complementary information weight, the greater the contribution of the corresponding data to the covariance. The greater the calculated power generation information association weight, the more significant the complementarity between the grid fluctuation power decomposition vector and the motor power vector.

[0078] S33, determining a power grid fluctuation power prediction weight according to the power impact weight and the power generation information association weight.

[0079] The power impact weight represents the predictive ability of the grid fluctuation power decomposition vector for the grid, and the power generation information association weight represents the degree of association between the grid fluctuation power decomposition vector and the motor control information. The power impact weight and the power generation information association weight are combined to judge the importance of the grid fluctuation power decomposition vector for the pre-control of the grid state.

[0080] The normalized value of the product of the power impact weight and the power generation information associated weight is used as the power grid fluctuation power prediction weight;

[0081] The larger the grid fluctuation power prediction weight is, the more important the information in the grid fluctuation power decomposition vector is for pre-controlling motor parameters through grid status.

[0082] S34, obtaining the power grid fluctuation characteristic value at each moment according to the numerical distribution of the neighborhood elements of all power grid fluctuation power decomposition vectors at each moment and the power grid fluctuation power prediction weight.

[0083] In order to characterize the influence of the grid state on the engine speed adjustment, the amplitude and frequency intensity of the decomposed signal are extracted. The larger the fluctuation amplitude and the smaller the fluctuation frequency, the greater the PCS energy storage resources that the grid itself needs to consume through the PCS in a charging and discharging cycle, and the greater the adjustment range of the engine speed at the corresponding moment.

[0084] Specifically, for each power grid fluctuation power decomposition vector, any element in the power grid fluctuation power decomposition vector is recorded as the target element; before calculating the target element, the mean of the absolute values ​​of all elements in a preset window including the target element is recorded as the signal amplitude at the corresponding time of the target element; the ratio of the signal amplitude to the central frequency of the power grid fluctuation power decomposition vector is recorded as the fluctuation amplitude of the power grid fluctuation power decomposition vector at the corresponding time of the target element;

[0085] The grid fluctuation power prediction weight of each grid fluctuation power decomposition vector is used as the weight, and the weighted average value of the fluctuation amplitude of all grid fluctuation power decomposition vectors at each moment is calculated, which is recorded as the grid fluctuation characteristic value at each moment.

[0086] The smaller the fluctuation frequency or the larger the fluctuation amplitude of the power grid fluctuation power at each moment, the larger the power grid fluctuation characteristic value, and the larger the amplitude of the engine speed adjustment.

[0087] S4, predicting the diesel generator control speed at the next moment according to the generator parameters and grid fluctuation characteristic values ​​at all moments.

[0088] In order to monitor the state of the power grid in real time and adjust the speed of the diesel generator in time according to the state, a training set of the neural network is constructed based on the data in the historical time period.

[0089] Obtain the preset diesel generator control speed at each moment; use the generator parameters at all moments and the preset diesel generator control speed as training data sets, input them into neural network training, and obtain a trained neural network model;

[0090] The neural network model in this embodiment adopts the LSTM neural network model, wherein the activation function of the LSTM neural network model adopts the RLU function, and the optimizer adopts the Adma optimizer; the loss function is the difference between the power grid fluctuation characteristic value at each moment and the previous moment. The training process of the LSTM neural network model is a common technology in the field of electrical signal processing and will not be repeated.

[0091] The generator parameters at the current moment are input into the trained neural network model to obtain the diesel generator control speed at the next moment.

[0092] The speed of the diesel generator is adjusted in real time according to the status of the power grid, so that energy can be stored or released in advance when the power grid needs to provide peak power or when the power grid needs to absorb peak power, to avoid the diesel generator from running out of speed or overloading.

[0093] Through the above description of the implementation method in combination with the accompanying drawings, technical personnel in the relevant field can understand that for the convenience and simplicity of description, only the division of the above-mentioned functional modules is used as an example. In actual applications, the above-mentioned functions can be assigned to different functional modules as needed, that is, the internal structure of the device can be divided into different functional modules to complete all or part of the functions described above.

[0094] The above contents are only specific implementation methods of the present application, but the protection scope of the present application is not limited thereto. Any technician familiar with the technical field can easily think of changes or substitutions within the technical scope disclosed in the present application, which should be included in the protection scope of the present application. Therefore, the protection scope of the present application should be based on the protection scope of the claims.

Claims

1. A method for intelligent identification of generator parameters based on neural network, characterized in that: The method comprises the following steps: S1, obtaining generator parameters at a node of a power generation grid; the generator parameters include: output power, rotation speed, grid power and PCS power of a diesel generator; S2, according to the difference between the power grid power and the PCS power at all times, the power grid fluctuation power vector is obtained; the power grid fluctuation power vector is subjected to wavelet decomposition to obtain the decomposition vector of each power grid fluctuation power; S3, according to the local numerical distribution of the elements in the power decomposition vector of all power grid fluctuations corresponding to each moment, the power grid fluctuation characteristic value at each moment is obtained, specifically including: S31, obtaining a power impact weight according to the numerical distribution of all elements in the power grid fluctuation power decomposition vector; S32, obtaining a power generation information association weight according to a correlation between the output power of the diesel generator at all times and the power grid fluctuation power decomposition vector; S33, determining a power grid fluctuation power prediction weight according to the power impact weight and the power generation information association weight; S34, obtaining the power grid fluctuation characteristic value at each moment according to the numerical distribution of the neighborhood elements of all power grid fluctuation power decomposition vectors at each moment and the power grid fluctuation power prediction weight; S4, predicting the diesel generator control speed at the next moment according to the generator parameters and grid fluctuation characteristic values ​​at all moments; obtaining the power influence weight according to the numerical distribution of all elements in the grid fluctuation power decomposition vector, including: According to the values ​​of all elements in the neighborhood of each element in the power grid fluctuation power decomposition vector, the power imbalance factor of each element is obtained; Calculate the mean and standard deviation of the power imbalance factors of all elements in the power grid fluctuation power decomposition vector; use the forward fusion result of the mean and standard deviation as the power impact weight of the power grid fluctuation power decomposition vector; The method for obtaining the power generation information association weight comprises: The vector composed of the output power of the diesel generator at all times is recorded as the motor power vector; The normalized value of the power imbalance factor of each element in the power grid fluctuation power decomposition vector is used as the complementary information weight of each element; The element value of each element in the motor power vector is replaced by the opposite number of the element value of each element to obtain the motor power symmetric vector; the complementary information weight of each element in the power grid fluctuation power decomposition vector is used as the weight, and the covariance of each power grid fluctuation power decomposition vector and the motor power symmetric vector is calculated, which is recorded as the power generation information association weight; The method of predicting the diesel generator control speed at the next moment according to the generator parameters and the grid fluctuation characteristic values ​​at all moments includes: Obtain the preset diesel generator control speed at each moment; use the generator parameters at all moments and the preset diesel generator control speed as training data sets, input them into neural network training, and obtain a trained neural network model; The generator parameters at the current moment are input into the trained neural network model to obtain the diesel generator control speed at the next moment.

2. The method for intelligent identification of generator parameters based on neural network according to claim 1, characterized in that: The grid fluctuation power vector is obtained according to the difference between the grid power and the PCS power at all times, including: The difference between the grid power and the PCS power at each moment is calculated and recorded as the fluctuating power at each moment; the vector composed of the fluctuating power at all moments is recorded as the grid fluctuating power vector.

3. The method for intelligent identification of generator parameters based on neural network according to claim 1, characterized in that: The power imbalance factor of each element is obtained according to the values ​​of all elements in the neighborhood of each element in the power grid fluctuation power decomposition vector, including: For each power grid fluctuation power decomposition vector, perform fast Fourier transform on the power grid fluctuation power decomposition vector to obtain the center frequency of the power grid fluctuation power decomposition vector; and obtain the average period of the power grid fluctuation power decomposition vector according to the center frequency; The value of the average period is used as the length of the preset window, and the sum of all elements in the preset window centered on each element in the power grid fluctuation power decomposition vector is calculated as the power imbalance factor of each element.

4. The method for intelligent identification of generator parameters based on neural network according to claim 3, characterized in that: The average period is the ratio of the value 2π to the central frequency of the power grid fluctuation power decomposition vector; and π is the ratio of a circle to a circle.

5. The method for intelligent identification of generator parameters based on neural network according to claim 1, characterized in that: The step of determining the power grid fluctuation power prediction weight according to the power impact weight and the power generation information association weight includes: The normalized value of the product of the power impact weight and the power generation information associated weight is used as the grid fluctuation power prediction weight.

6. The method for intelligent identification of generator parameters based on neural network according to claim 3, characterized in that: The method for obtaining the power grid fluctuation characteristic value at each moment includes: According to the numerical distribution of all elements in the neighborhood of each element in the power fluctuation decomposition vector of each power grid, the signal amplitude of each element at the corresponding time is obtained; The ratio of the signal amplitude to the central frequency of the power grid fluctuation power decomposition vector is recorded as the fluctuation amplitude of the power grid fluctuation power decomposition vector at the time corresponding to the target element; The grid fluctuation power prediction weight of each grid fluctuation power decomposition vector is used as the weight, and the weighted average value of the fluctuation amplitude of all grid fluctuation power decomposition vectors at each moment is calculated, which is recorded as the grid fluctuation characteristic value at each moment.

7. The method for intelligently identifying generator parameters based on neural network according to claim 6, characterized in that: The method for obtaining the signal amplitude of each element at a corresponding moment includes: For each power grid fluctuation power decomposition vector, any element in the power grid fluctuation power decomposition vector is recorded as the target element; before calculating the target element, the mean of the absolute values ​​of all elements in the preset window including the target element is recorded as the signal amplitude at the corresponding moment of the target element.

Images