CN113761793B - Inverter output impedance detection device and method and inverter operation control method - Google Patents
Inverter output impedance detection device and method and inverter operation control method Download PDFInfo
- Publication number
- CN113761793B CN113761793B CN202110935878.4A CN202110935878A CN113761793B CN 113761793 B CN113761793 B CN 113761793B CN 202110935878 A CN202110935878 A CN 202110935878A CN 113761793 B CN113761793 B CN 113761793B
- Authority
- CN
- China
- Prior art keywords
- inverter
- output impedance
- neural network
- network model
- training
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Active
Links
- 238000001514 detection method Methods 0.000 title claims abstract description 83
- 238000000034 method Methods 0.000 title claims abstract description 33
- 238000012549 training Methods 0.000 claims abstract description 101
- 238000003062 neural network model Methods 0.000 claims abstract description 72
- 239000003990 capacitor Substances 0.000 claims abstract description 24
- 230000004044 response Effects 0.000 claims abstract description 8
- 238000011176 pooling Methods 0.000 claims description 30
- 238000012360 testing method Methods 0.000 claims description 22
- 238000005070 sampling Methods 0.000 claims description 18
- 238000013527 convolutional neural network Methods 0.000 claims description 15
- 230000006870 function Effects 0.000 claims description 13
- 238000012795 verification Methods 0.000 claims description 11
- 230000004913 activation Effects 0.000 claims description 9
- 230000008569 process Effects 0.000 claims description 8
- 238000007781 pre-processing Methods 0.000 claims description 7
- 238000004458 analytical method Methods 0.000 claims description 6
- 238000013135 deep learning Methods 0.000 claims description 6
- 230000003044 adaptive effect Effects 0.000 claims description 4
- 230000001131 transforming effect Effects 0.000 claims description 3
- 230000003321 amplification Effects 0.000 claims description 2
- 238000003199 nucleic acid amplification method Methods 0.000 claims description 2
- 238000011897 real-time detection Methods 0.000 abstract description 2
- 238000013528 artificial neural network Methods 0.000 description 6
- 239000011159 matrix material Substances 0.000 description 4
- OKTJSMMVPCPJKN-UHFFFAOYSA-N Carbon Chemical compound [C] OKTJSMMVPCPJKN-UHFFFAOYSA-N 0.000 description 3
- 230000009471 action Effects 0.000 description 3
- 238000013473 artificial intelligence Methods 0.000 description 3
- 238000004364 calculation method Methods 0.000 description 3
- 229910052799 carbon Inorganic materials 0.000 description 3
- 238000004891 communication Methods 0.000 description 3
- 238000013461 design Methods 0.000 description 3
- 238000010586 diagram Methods 0.000 description 3
- 238000004146 energy storage Methods 0.000 description 3
- 238000002474 experimental method Methods 0.000 description 3
- 238000006243 chemical reaction Methods 0.000 description 2
- 230000008878 coupling Effects 0.000 description 2
- 238000010168 coupling process Methods 0.000 description 2
- 238000005859 coupling reaction Methods 0.000 description 2
- 238000005516 engineering process Methods 0.000 description 2
- 238000001914 filtration Methods 0.000 description 2
- 230000006872 improvement Effects 0.000 description 2
- 230000035699 permeability Effects 0.000 description 2
- 238000007493 shaping process Methods 0.000 description 2
- 230000001629 suppression Effects 0.000 description 2
- ORILYTVJVMAKLC-UHFFFAOYSA-N Adamantane Natural products C1C(C2)CC3CC1CC2C3 ORILYTVJVMAKLC-UHFFFAOYSA-N 0.000 description 1
- 238000012356 Product development Methods 0.000 description 1
- 230000006978 adaptation Effects 0.000 description 1
- 230000009286 beneficial effect Effects 0.000 description 1
- 230000005540 biological transmission Effects 0.000 description 1
- 238000002847 impedance measurement Methods 0.000 description 1
- 238000009434 installation Methods 0.000 description 1
- 238000011545 laboratory measurement Methods 0.000 description 1
- 238000010801 machine learning Methods 0.000 description 1
- 238000013178 mathematical model Methods 0.000 description 1
- 238000012986 modification Methods 0.000 description 1
- 230000004048 modification Effects 0.000 description 1
- 230000008450 motivation Effects 0.000 description 1
- 238000006386 neutralization reaction Methods 0.000 description 1
- 238000005457 optimization Methods 0.000 description 1
- 238000012545 processing Methods 0.000 description 1
- 238000011160 research Methods 0.000 description 1
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F30/00—Computer-aided design [CAD]
- G06F30/20—Design optimisation, verification or simulation
- G06F30/27—Design optimisation, verification or simulation using machine learning, e.g. artificial intelligence, neural networks, support vector machines [SVM] or training a model
-
- G—PHYSICS
- G01—MEASURING; TESTING
- G01R—MEASURING ELECTRIC VARIABLES; MEASURING MAGNETIC VARIABLES
- G01R27/00—Arrangements for measuring resistance, reactance, impedance, or electric characteristics derived therefrom
- G01R27/02—Measuring real or complex resistance, reactance, impedance, or other two-pole characteristics derived therefrom, e.g. time constant
- G01R27/08—Measuring resistance by measuring both voltage and current
-
- H—ELECTRICITY
- H02—GENERATION; CONVERSION OR DISTRIBUTION OF ELECTRIC POWER
- H02M—APPARATUS FOR CONVERSION BETWEEN AC AND AC, BETWEEN AC AND DC, OR BETWEEN DC AND DC, AND FOR USE WITH MAINS OR SIMILAR POWER SUPPLY SYSTEMS; CONVERSION OF DC OR AC INPUT POWER INTO SURGE OUTPUT POWER; CONTROL OR REGULATION THEREOF
- H02M7/00—Conversion of ac power input into dc power output; Conversion of dc power input into ac power output
- H02M7/42—Conversion of dc power input into ac power output without possibility of reversal
- H02M7/44—Conversion of dc power input into ac power output without possibility of reversal by static converters
- H02M7/48—Conversion of dc power input into ac power output without possibility of reversal by static converters using discharge tubes with control electrode or semiconductor devices with control electrode
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F2119/00—Details relating to the type or aim of the analysis or the optimisation
- G06F2119/02—Reliability analysis or reliability optimisation; Failure analysis, e.g. worst case scenario performance, failure mode and effects analysis [FMEA]
Abstract
The invention discloses an inverter output impedance detection device and method and an inverter operation control method, and provides a new inverter output impedance detection scheme on the premise of verifying the feasibility of calculating the inverter output impedance through the voltage of a filter capacitor and the current of a filter inductor at the inverter side, namely, the sampled voltage of the filter capacitor and the current of the filter inductor at the inverter side are input into a neural network model to predict the output impedance of the inverter, and the establishment and training steps of the neural network model are provided; the neural network model is downloaded into an intelligent chip, and the CPU modulates pulse signals through a self-adaptive algorithm according to a real-time detection result output by the intelligent chip, so that the real-time control of the running state of the inverter is realized. The inverter output impedance detection device does not need a voltage source or a current source as disturbance quantity or a frequency response analyzer, and can accurately predict the output impedance of the inverter in real time through a deep neural network model so as to realize the self-adaptive control of the inverter.
Description
Technical Field
The invention relates to the field of inverters, in particular to an inverter output impedance detection device and method and an inverter operation control method.
Background
Under the goals of carbon neutralization and carbon peak reaching of various countries in the world, the increment of new energy installation including photovoltaic, wind power and energy storage is greatly increased year by year, but the continuous improvement of the permeability of new energy reduces carbon emission and simultaneously reduces the strength of a power grid, and the power grid with lower strength is generally called a weak power grid.
The weak power grid can cause instability of the new energy inverter, and a great challenge is brought to safe and stable operation of the power system. The current impedance analysis method is widely used for stability analysis of the inverter, the stability of the system can be judged by calculating the output impedance of the inverter and the impedance of a power grid, and the output impedance of the inverter is changed by a design algorithm so as to improve the stability of the inverter. However, the method is based on mathematical model deduction, and is only an approximation of the actual model, and the errors are larger because the parameters of software and hardware in the inverter can be changed along with the running state.
In the prior art, control technology for an inverter is mainly focused on resonance suppression, such as chinese patent application with patent publication number CN110247427a, and discloses a grid-connected inverter resonance intelligent suppression method for online identification of grid parameters, and also in the prior art for detecting performance of an inverter, such as chinese patent application with patent publication number CN106291107a, an output impedance angle detection device and a detection method of an inverter are disclosed, but in the prior art, less research is required for detecting output impedance of the inverter, and in the current common inverter output impedance measurement method, disturbance voltage or disturbance current is generally applied at a common coupling Point (PCC), and voltage and current of the PCC point are measured by a frequency response analyzer to calculate output impedance. The method needs more instruments, and can be used for optimizing the design parameters of the inverter only in the product development and test stage, but cannot be applied to the actual working condition of the inverter.
The prior art has at least the following drawbacks:
in the first and industrial application, the output impedance of the actual working condition of the inverter is difficult to detect in real time, because the application of disturbance signals in real time in the actual working condition or the application of more laboratory instruments in the working condition detection is impractical;
second, the real-time detection value of the inverter output impedance is not known at present, which results in no motivation for those skilled in the art to study the detection technology of the inverter output impedance.
Disclosure of Invention
In order to solve the problems in the prior art, the invention provides an inverter output impedance detection device and method and an inverter operation control method, wherein the output impedance of an inverter is calculated in real time based on a neural network model by utilizing filter capacitor voltage and filter inductor current at the inverter side, and a control command is generated by the inverter output impedance value which is output in real time so as to control the operation state of the inverter in real time. Compared with the prior art that the disturbance voltage or disturbance current is applied at the public coupling point, the frequency response analyzer is used for measuring the voltage and the current of the PCC point to calculate the output impedance, and the invention provides a brand-new method for detecting the output impedance of the inverter. The verification process of the feasibility of calculating the output impedance of the inverter (i.e., the detection principle of the output impedance of the inverter) by using the filter capacitor voltage and the filter inductor current is as follows:
see the equivalent circuit diagram of the inverter and the grid of fig. 2, wherein i s 、z o1 And C is the equivalent circuit of the inverter, i s I is the output current of the equivalent current source L1 For inverter side inductor current, z o1 For the equivalent output impedance of the inverter without considering the filter capacitance, C is the filter capacitance, z o1 And C together form the output impedance z of the inverter o 。z g For the equivalent impedance of the power grid, v g Is the grid voltage. From the equivalent current, it is possible to:
i L1 ＝i s -v c /z o1 (1)
by collecting n variable values at different moments, and m different operating states (including active power P and reactive power Q), the following equation can be obtained:
when the inverter is stably operated in one state, it can be considered that
z o1_0 ＝z o1_1 ＝…＝z o1_n-1 ，i s_0 ＝i s_1 ＝…＝i s_n-1 (3)
Then z can be estimated according to equations 2 and 3 o1 And i s 。
At the same time, since the filter capacitor voltage and the filter inductor current contain abundant subharmonics, z under different frequencies can be obtained in principle o1 And i s Further, an inverter output impedance z is obtained according to the following equation (4) o Is that
Wherein z is c Is a capacitive impedance, and z c ＝1/C。
Thus far, the feasibility of the scheme of calculating the inverter output impedance according to the inverter side filter inductance current and the filter capacitance voltage is verified.
However, although the output impedance of the inverter can be theoretically calculated by this calculation method, in an actual system, the collected voltage and current signals have electromagnetic interference and are simultaneously interfered by the voltage of the power grid, so that it is difficult to accurately and stably detect the impedance by using the calculation method.
On the premise of verifying the feasibility of the calculation scheme (under the condition of not considering interference), the invention provides an inverter output impedance detection algorithm based on artificial intelligence. Compared with the traditional machine learning algorithm, the deep neural network algorithm does not need to manually design a feature extractor, is obtained by self-learning of the network, is particularly suitable for data with complex changes, and has very good generalization capability and robustness. The excellent characteristics enable the output impedance of the inverter to be accurately and stably detected (predicted) by adopting the deep neural network, and the specific technical scheme is as follows:
in one aspect, the present invention provides an intelligent detection device for detecting an output impedance of an inverter outside the detection device, where the detection device does not need a voltage source or a current source as a disturbance amount, and the intelligent detection device includes the following modules:
the sampling module is used for sampling the voltage of the filter capacitor and the current of the filter inductor at the side of the inverter;
the intelligent chip is used for receiving the voltage signals and the current signals sampled by the sampling module and inputting the voltage signals and the current signals into an internal inverter output impedance detection neural network model, and the inverter output impedance detection neural network model is built in advance and training is completed so as to output inverter output impedance according to the input voltage signals and the current signals.
Further, a training sample for training the inverter output impedance detection neural network model is obtained by:
s1, collecting data: under the condition of building two identical inverter working platforms, collecting filter capacitor voltage and filter inductor current data of a working circuit of an inverter at different moments on a first inverter working platform to obtain a plurality of samples, and measuring actual impedance values of the inverter at corresponding different moments on a second inverter working platform;
s2, data preprocessing: and (3) converting the sample data collected in the step (S1) into frequency domain data to obtain training samples, and respectively taking each actual impedance value measured in the step (S1) as a label of the training samples after sample data acquired at the same moment are converted.
Further, the inverter output impedance detection neural network model is established in advance by the following steps:
e1, building a convolutional neural network model by using a deep learning platform, configuring a network structure to comprise a convolutional layer, a pooling layer and a full-connection layer, and defining a network;
e2, configuring a training environment and creating an actuator for the built convolutional neural network, and defining a network output data dimension according to the dimension of the training sample;
and E3, training the neural network model by the executor by using training samples, wherein each training sample is provided with a corresponding label.
Further, in step S2, after transforming the sample data into frequency domain data, training samples and test samples are obtained from the frequency domain data according to a preset ratio respectively;
e3, after training the neural network model by using a training sample for a preset round, verifying the neural network model by using the test sample to obtain a loss value and accuracy; if the loss value and/or the accuracy rate pass the verification, the current neural network model is used as the inverter output impedance detection neural network model, otherwise, the training sample is used for carrying out iterative training on the neural network model until the loss value and/or the accuracy rate of the neural network model obtained through training passes the verification;
and downloading the inverter output impedance detection neural network model into the intelligent chip.
Further, the intelligent detection device of the inverter output impedance further comprises a CPU module and a PWM module, wherein the CPU module is electrically connected with the intelligent chip and the PWM module, and the PWM module is used for being electrically connected with the inverter;
the sampling module is used for collecting and transmitting voltage signals and current signals in real time, the intelligent chip outputs estimated inverter output impedance values to the CPU module in real time, and the CPU module is used for generating control instructions according to the current inverter output impedance values and transmitting the control instructions to the PWM module;
the PWM module is used for outputting an adaptive pulse signal according to the control instruction so as to control the running state of the inverter in real time.
On the other hand, the invention provides an intelligent detection method of the output impedance of an inverter, which does not need a voltage source to apply disturbance voltage and does not need a current source to apply disturbance current, and comprises the following steps:
collecting filter capacitor voltage and filter inductor current at the inverter side;
inputting the data acquisition results of the filter capacitor voltage and the filter inductor current into a pre-generated inverter output impedance detection neural network model, wherein the inverter output impedance detection neural network model outputs the detection results of the corresponding inverter output impedance; the inverter output impedance detection neural network model is pre-generated through the following steps:
s1, collecting data: under the condition of building two identical inverter working platforms, collecting the data of the filter capacitor voltage and the filter inductor current of the inverter side of a working circuit of the inverter at different moments on a first inverter working platform to obtain a plurality of samples, and measuring the actual impedance value of the inverter at corresponding different moments on a second inverter working platform to obtain corresponding labels;
s2, data preprocessing: transforming the sample data collected in the step S1 into frequency domain data to obtain training samples;
s3, model training: and (2) training the neural network model by utilizing the training sample obtained by preprocessing in the step (S2) and the corresponding label thereof to obtain the inverter output impedance detection neural network model.
Further, training the neural network model in step S3 includes:
s31, constructing a convolutional neural network model by using a deep learning platform, configuring a network structure to comprise a first convolutional layer, a first pooling layer, a second convolutional layer, a second pooling layer and a full-connection layer, and defining a network;
s32, configuring a training environment and creating an actuator for the built convolutional neural network, and defining the dimension of network output data to be equal to the dimension of a training sample;
and S33, training the neural network model by using the actuator according to a preset training round.
Further, in step S2, after the sample data is transformed into frequency domain data, a training sample is obtained from the transformed frequency domain data, and a test sample is obtained; or, constructing a third same inverter working platform to obtain a test sample;
step S33 further includes:
s34, after training of preset rounds is completed each time, verifying the neural network model by using the test sample book to obtain a loss value and accuracy;
s35, judging whether the loss value and/or the accuracy obtained through verification meet preset conditions, and if so, taking the current neural network model as the inverter output impedance detection neural network model; otherwise, the process returns to S33-S35.
Further, the network definition in step S31 includes:
defining the convolution kernel size of the first convolution layer and the activation function of the corresponding layer;
defining the size of a pooling core of the first pooling layer and the pooling type of the corresponding layer as max, wherein the pooling step length is 2;
defining the convolution kernel size of the second convolution layer and the activation function of the corresponding layer;
defining the pooling core size of the second pooling layer and the pooling type of the corresponding layer;
and defining the size of the full connection layer and the activation function of the corresponding layer.
Further, in step S1, the actual impedance value of the inverter is measured in real time by using a frequency response analyzer and a linear amplification device, and the operation parameters are synchronously adjusted for the first inverter working platform and the second inverter working platform, so as to obtain sample data under different working conditions and labels corresponding to the same moment.
Further, the collected sample data is subjected to a fast fourier transform in step S2, and the frequency range is selected to be 50Hz to 6000Hz.
In still another aspect, the present invention provides an intelligent control method for an inverter operation state, including the steps of:
detecting the output impedance of the current inverter by using the intelligent detection method of the output impedance of the inverter;
and judging the stability of the current inverter according to an impedance analysis method, and modulating a pulse signal for controlling the running state of the inverter through a self-adaptive algorithm.
The technical scheme provided by the invention has the following beneficial effects:
a. the output impedance of the inverter can be accurately output based on the intelligent chip running artificial neural network algorithm, voltage disturbance or current disturbance is not required to be applied, and the output impedance of the inverter can be detected without a frequency response analyzer or other measuring equipment;
b. the output impedance of the inverter in the actual working condition can be detected in real time;
c. on the premise of detecting output impedance in real time in actual working conditions, the working stability of the inverter is judged according to an impedance analysis method, and then control parameters are improved through a self-adaptive algorithm, so that the stability of the inverter is improved.
Drawings
In order to more clearly illustrate the embodiments of the present application or the technical solutions in the prior art, the drawings that are required to be used in the embodiments or the description of the prior art will be briefly described below, and it is obvious that the drawings in the following description are only some embodiments described in the present application, and other drawings may be obtained according to these drawings without inventive effort for a person skilled in the art.
Fig. 1 is a schematic structural diagram of an intelligent detection device for output impedance of an inverter according to an embodiment of the present invention;
fig. 2 is an equivalent circuit diagram of the inverter and the grid of fig. 1;
fig. 3 is a flowchart of an intelligent detection method for output impedance of an inverter according to an embodiment of the present invention;
Detailed Description
In order that those skilled in the art will better understand the present invention, a technical solution in the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings in which it is apparent that the described embodiments are only some embodiments of the present invention, not all embodiments. All other embodiments, which can be made by those skilled in the art based on the embodiments of the present invention without making any inventive effort, shall fall within the scope of the present invention.
It should be noted that the terms "first," "second," and the like in the description and the claims of the present invention and the above figures are used for distinguishing between similar objects and not necessarily for describing a particular sequential or chronological order. It is to be understood that the data so used may be interchanged where appropriate such that the embodiments of the invention described herein may be implemented in sequences other than those illustrated or otherwise described herein. Furthermore, the terms "comprises," "comprising," and "having," and any variations thereof, are intended to cover a non-exclusive inclusion, such that a process, method, apparatus, article, or device that comprises a list of steps or elements is not necessarily limited to those steps or elements expressly listed or inherent to such process, method, article, or device.
With the improvement of the permeability of new energy, the running power grid environment of the photovoltaic/energy storage inverter is increasingly complex, and the intelligent requirement on the inverter is also increasingly increased, so that the intelligent requirement of the inverter cannot be realized only by using the existing control chip. In this context, the present invention proposes an engineering-implementable, real-time, high-precision inverter output impedance detection technique for photovoltaic/energy storage inverters.
In one embodiment of the present invention, an intelligent detection device for an output impedance of an inverter is provided, which is used for detecting the output impedance of the inverter outside the detection device, the detection device does not need a voltage source or a current source as a disturbance amount, referring to fig. 1, and the intelligent detection device for the output impedance of the inverter includes the following modules:
the sampling module is used for sampling the voltage of the filter capacitor and the current of the filter inductor at the side of the inverter;
the intelligent chip is used for receiving the voltage signals and the current signals sampled by the sampling module and inputting the voltage signals and the current signals into an internal inverter output impedance detection neural network model, and the inverter output impedance detection neural network model is pre-established and training is completed so as to output a predicted value of the inverter output impedance according to the input voltage signals and the current signals;
the CPU module is electrically connected with the intelligent chip and is used for receiving the predicted value of the output impedance of the inverter and generating a control instruction according to the current value of the output impedance of the inverter;
the input end of the PWM module is electrically connected with the CPU module, the output end of the PWM module is electrically connected with the inverter, and the PWM module is used for receiving the control instruction and outputting an adaptive pulse signal according to the control instruction so as to control the running state of the inverter in real time.
In addition, the detection device further comprises a communication module and a storage module, wherein the communication module is used for realizing communication between the CPU inside the device and equipment (such as a controller or a mobile terminal) outside the device, and the storage module is used for storing processing information (such as CPU) inside the detection device.
The application scenario of the intelligent detection device for the output impedance of the inverter is as shown in fig. 1: the inverter connected with the power grid is a target inverter to be detected, and a filter inductor and a filter capacitor are arranged on the inverter side and used for filtering a transmission signal of the inverter to the power grid.
Based on the configuration (the filter capacitor and the filter inductor), the sampling module of the detection device samples the terminal voltage of the filter capacitor and the current flowing through the filter inductor respectively as input parameters of a neural network model in the intelligent chip; the sampling module can collect and send the voltage signal and the current signal in real time, so that the intelligent chip can output the estimated inverter output impedance value to the CPU module in real time, and finally, the real-time self-adaptive control of the inverter is realized.
Namely, an artificial intelligent chip is additionally arranged outside the existing inverter control chip (CPU module), which can be TPU, ARM, STM series chips or other chips suitable for improving intelligent algorithms, is specially used for operating the artificial intelligent algorithm, improves the operation speed of the algorithm and ensures the real-time performance of the algorithm. The smart chip is used for inverter output impedance detection (prediction), but does not preclude running other artificial intelligence algorithms in other scenarios.
The invention utilizes the following working principle to detectInverter output impedance: the sampling module samples the voltage v of the filter capacitor at the side of the inverter in real time c And filtering inductor current i L1 The sampled value is input into a CPU and an intelligent chip, and the intelligent chip estimates the output impedance of the inverter in real time through an intelligent impedance detection algorithm, wherein the algorithm can be an artificial neural network algorithm or other intelligent algorithms. The intelligent chip inputs the estimated output impedance of the inverter into the CPU, the CPU uses the received impedance value in the control program of the inverter, and outputs a control signal to the PWM module, and the PWM module outputs a pulse signal to control the operation of the inverter.
As verified above, it is possible to calculate the inverter output impedance from the filter inductor current and the filter capacitor voltage on the inverter side, and therefore, the embodiment of the present invention uses the inductor current and the capacitor terminal voltage as input parameters of the neural network model and uses the inverter output impedance as output parameters of the neural network model.
After the input parameter type and the output parameter type are determined, a neural network model is established through the following steps:
and E1, constructing a convolutional neural network model by using a deep learning platform, configuring a network structure to comprise a convolutional layer, a pooling layer and a full-connection layer, and performing network definition on the convolutional neural network model.
Specifically, in this embodiment, a Convolutional Neural Network (CNN) may be built using a flying-oar deep learning platform, and the network structure is configured to include a first convolutional layer, a first pooling layer, a second convolutional layer, a second pooling layer, and a full-connection layer, where the size of a convolutional kernel of the first convolutional layer is defined as 10, the number is 5, and the activation function is relu; defining the size of a pooling core of the first pooling layer as 2, the pooling type as max and the pooling step length as 2; defining the convolution kernel of the second convolution layer to be 25 in size and 5 in number, and the activation function to be relu; defining the size of a pooling core of the second pooling layer as 2, the pooling type as max and the pooling step length as 2; defining the size of the full connection layer as 10, and defining the activation function as sigmoid.
Furthermore, the network definition operation further includes:
the loss function is selected to be any one of the following:
log-dash loss functionOr (b)
Mean square error loss functionOr (b)
Average absolute value error loss functionWherein (1)>For the loss value verified after each training, N is the number of test samples verified each time, +.>For the predicted value of the ith test sample, y i The label value corresponding to the ith test sample;
defining the optimization method of the convolutional neural network model as Adam, adaMax or Nadam;
the learning rate is defined to be in the range of 0.0001 to 0.002, preferably 0.001.
E2, configuring a training environment and a creation Executor (Executor) for the built convolutional neural network, and defining a network output data dimension according to the dimension of the training sample.
Specifically, after the CNN network is built, a training environment is configured, an execcoter is created, then an output data dimension is defined to be equal to a sample dimension, and training rounds are set.
And E3, training the neural network model by the executor by using training samples, wherein each training sample is provided with a corresponding label, and finishing (finishing) training until the loss value and/or the accuracy of the trained neural network model meet the requirements, namely deploying the neural network model which is currently finished training into an intelligent chip as the inverter output impedance detection neural network model.
The training samples required for training the neural network model in the present invention, see fig. 3, are obtained in one embodiment of the present invention by:
s1, collecting data: under the condition that two identical inverter working platforms are built, the filter capacitor voltage and filter inductor current data of the inverter side of a working circuit of an inverter at different moments are collected on a first inverter working platform so as to obtain a plurality of samples, and the actual impedance value of the inverter is measured at corresponding different moments on a second inverter working platform.
Specifically, two identical inverter experiment platforms are firstly built in a laboratory, and the two inverters are operated in identical states, wherein one inverter experiment platform is called an A platform, data of filter capacitor voltage and filter inductor current are collected to be used as training samples and test samples, the other inverter experiment platform is called a B platform, a frequency response analyzer and linear amplifying equipment are adopted to measure actual impedance values of the inverters in real time to be used as labels of the training samples and the test samples, and the sampling time of the training samples/the test samples is the same as the measuring time of the corresponding labels (actual impedance values).
Taking voltage and current sampling values with a sampling frequency of 16KHz and 20ms for each sample as an example, each sample is a matrix of 2×320; the samples under different working conditions are acquired by changing the output modes (such as active power, reactive power, direct current bus voltage, grid equivalent impedance and the like (one or any combination can be selected), and the number of the samples acquired in the embodiment exceeds 10 5 The training round is set to 200 (obviously, the number of samples and the training round can be correspondingly adjusted). The inverter output impedance value is measured synchronously on the B-stage with the same frequency and time as the label of the sample.
S2, data preprocessing: and (3) converting the sample data collected in the step (S1) into frequency domain data to obtain training samples, and respectively taking each actual impedance value measured in the step (S1) as a label of the training samples after sample data acquired at the same moment are converted.
Specifically, all the collected data of step S1 are collected together and preprocessed. Since impedance detection is performed in the frequency domain, in this embodiment, the collected sample data is subjected to Fast Fourier Transform (FFT), for example, the frequency range is selected to be (50 Hz,6000 Hz), the frequency point interval is 50Hz, and each converted sample becomes a matrix of 2×120;
the sample data may be shaped as desired, such as by shaping a 2×120 matrix sample into a 12×20 or 16×15 matrix, so that the shaped sample can better train the neural network model.
Taking 80% of samples after FFT conversion or FFT conversion and shaping as training samples and the other 20% of samples as test samples, wherein the sample proportion is only taken as an example, and dividing the training samples and the test samples from the total samples is only one embodiment.
After the training sample set and the test sample set are acquired, the neural network model is trained as follows:
inputting a training sample set with a label into a neural network model, and learning the neural network model according to each training sample in the training sample set until the training samples in the training sample set are trained once, completing one round of learning, if the training round is set to be 200, then starting a new round of learning until 200 rounds of learning are completed, and completing one round of training.
After each training is completed, inputting a test sample set without labels into a neural network model, outputting corresponding predicted values according to the neural network model, comparing each label, and calculating a corresponding loss value Cost and an accuracy acc, wherein a certain tolerance range can be set for calculating the accuracy, for example, the absolute value of the difference between the predicted values and the label values is considered to be accurate in a certain range, if the absolute value is out of range, the absolute value is considered to be inaccurate, and the ratio of the number of the accurate predicted values to the total number of the predicted values is the accuracy.
The loss value is calculated according to a loss function, taking a log-dash loss function as an example,
and carrying the loss function according to the predicted values and the label values of the N test samples to obtain corresponding loss values.
The smaller the loss value or the higher the accuracy rate, the higher the accuracy of the neural network model after the current training is indicated, and the lower the accuracy of the neural network model after the current training is indicated. Setting a training target, wherein the loss value is smaller than a loss threshold value, or the accuracy is higher than an accuracy threshold value, or both are satisfied; and when the verification loss value and/or the accuracy rate meet the training target, the current trained model is considered to meet the requirement.
If the training target is reached, stopping training, and downloading the current neural network model into the intelligent chip;
if the training target is not reached, performing iterative training, namely inputting the training sample set into the neural network model which is trained at present for a plurality of times again until the next training is completed, and performing the verification loss value and/or the accuracy rate as described above to judge whether the training target is met or not, and if the training loss value and/or the accuracy rate are not met, performing iterative training and verification again until the training target is reached.
The inverter output impedance detection scheme of the invention can be applied to single-phase inverters and is also applicable to three-phase inverters, and the invention is not limited to the type or specific model of the applied inverter.
After the neural network model reaching the training target (passing verification) is downloaded into an intelligent chip, the trained model is used for estimating an inverter output impedance value according to voltage signals and current signals obtained by actual sampling under the working condition of an inverter, and the method is used for detecting the inverter output impedance by the inverter output impedance detection device;
after the output impedance of the current inverter is detected, the detection result is output to a CPU, the stability of the current inverter is judged according to an impedance analysis method, and the CPU modulates pulse signals through a self-adaptive algorithm to control the running state of the inverter, so that the method for controlling the running state of the inverter by the inverter output impedance detection device in the embodiment of the invention effectively improves the robustness of a system where the inverter is located.
The invention provides an inverter output impedance detection algorithm based on artificial intelligence, breaks through the limitation of laboratory measurement, verifies the feasibility and scientificity of the algorithm, increases a special intelligent chip, can realize a complex artificial neural network algorithm, improves the detection accuracy and robustness, and enables the inverter output impedance to be accurately and stably predicted by adopting a deep neural network. The inverter output impedance detection device does not need a voltage source or a current source as disturbance quantity or a frequency response analyzer, and can accurately predict the output impedance of the inverter in real time through a deep neural network model so as to realize the self-adaptive control of the inverter.
It is noted that relational terms such as first and second, and the like are used solely to distinguish one entity or action from another entity or action without necessarily requiring or implying any actual such relationship or order between such entities or actions. Moreover, the terms "comprises," "comprising," or any other variation thereof, are intended to cover a non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements does not include only those elements but may include other elements not expressly listed or inherent to such process, method, article, or apparatus. Without further limitation, an element defined by the phrase "comprising one … …" does not exclude the presence of other like elements in a process, method, article, or apparatus that comprises the element.
The foregoing is merely exemplary of the application and it should be noted that modifications and adaptations to those skilled in the art may be made without departing from the principles of the application and are intended to be comprehended within the scope of the application.
Claims (12)
1. An intelligent detection device for the output impedance of an inverter, which is used for detecting the output impedance of the inverter outside the detection device, is characterized in that the detection device does not need a voltage source or a current source as disturbance quantity, and comprises the following modules:
the sampling module is used for sampling the voltage of the filter capacitor and the current of the filter inductor at the side of the inverter;
the intelligent chip is used for receiving the voltage signals and the current signals sampled by the sampling module and inputting the voltage signals and the current signals into an internal inverter output impedance detection neural network model, and the inverter output impedance detection neural network model is pre-established and training is completed so as to output inverter output impedance according to the input voltage signals and the current signals;
the device also comprises a CPU module and a PWM module, wherein the CPU module is used for generating a control instruction according to the current output impedance value of the inverter and sending the control instruction to the PWM module;
the PWM module is used for outputting an adaptive pulse signal according to the control instruction so as to control the running state of the inverter in real time.
2. The intelligent detection apparatus of inverter output impedance according to claim 1, wherein a training sample for training the inverter output impedance detection neural network model is obtained by:
s1, collecting data: under the condition of building two identical inverter working platforms, collecting the data of the filter capacitor voltage and the filter inductor current of the inverter side of a working circuit of the inverter at different moments on a first inverter working platform so as to obtain a plurality of samples, and measuring the actual impedance value of the inverter at corresponding different moments on a second inverter working platform;
s2, data preprocessing: and (3) converting the sample data collected in the step (S1) into frequency domain data to obtain training samples, and respectively taking each actual impedance value measured in the step (S1) as a label of the training samples after sample data acquired at the same moment are converted.
3. The intelligent detection device for inverter output impedance according to claim 2, wherein the inverter output impedance detection neural network model is previously established by:
e1, building a convolutional neural network model by using a deep learning platform, configuring a network structure to comprise a convolutional layer, a pooling layer and a full-connection layer, and defining a network;
e2, configuring a training environment and creating an actuator for the built convolutional neural network, and defining a network output data dimension according to the dimension of the training sample;
and E3, training the neural network model by the executor by using training samples, wherein each training sample is provided with a corresponding label.
4. The intelligent detection device for inverter output impedance according to claim 3, wherein after the sample data is converted into frequency domain data in step S2, training samples and test samples are obtained from the frequency domain data according to a preset ratio, respectively;
e3, after training the neural network model by using a training sample for a preset round, verifying the neural network model by using the test sample to obtain a loss value and accuracy; if the loss value and/or the accuracy rate pass the verification, the current neural network model is used as the inverter output impedance detection neural network model, otherwise, the training sample is used for carrying out iterative training on the neural network model until the loss value and/or the accuracy rate of the neural network model obtained through training passes the verification;
and downloading the inverter output impedance detection neural network model into the intelligent chip.
5. The intelligent detection device for the output impedance of the inverter according to claim 1, wherein the CPU module is electrically connected to both the intelligent chip and a PWM module, and the PWM module is electrically connected to the inverter;
the sampling module is used for collecting and transmitting voltage signals and current signals in real time, and the intelligent chip outputs estimated inverter output impedance values to the CPU module in real time.
6. An intelligent detection method for the output impedance of an inverter, wherein a disturbance voltage is not required to be applied by a voltage source and a disturbance current is not required to be applied by a current source, the detection method comprising the following steps:
collecting the voltage of a filter capacitor and the current of a filter inductor at the side of an inverter;
inputting the data acquisition results of the filter capacitor voltage and the filter inductor current into a pre-generated inverter output impedance detection neural network model, wherein the inverter output impedance detection neural network model outputs the detection results of the corresponding inverter output impedance; generating a control command according to the current output impedance value of the inverter, and outputting an adaptive pulse signal according to the control command to control the running state of the inverter in real time; the inverter output impedance detection neural network model is pre-generated through the following steps:
s1, collecting data: under the condition of building two identical inverter working platforms, collecting the data of the filter capacitor voltage and the filter inductor current of the inverter side of a working circuit of the inverter at different moments on a first inverter working platform to obtain a plurality of samples, and measuring the actual impedance value of the inverter at corresponding different moments on a second inverter working platform to obtain corresponding labels;
s2, data preprocessing: transforming the sample data collected in the step S1 into frequency domain data to obtain training samples;
s3, model training: and (2) training the neural network model by utilizing the training sample obtained by preprocessing in the step (S2) and the corresponding label thereof to obtain the inverter output impedance detection neural network model.
7. The intelligent detection method according to claim 6, wherein training the neural network model in step S3 includes:
s31, constructing a convolutional neural network model by using a deep learning platform, configuring a network structure to comprise a first convolutional layer, a first pooling layer, a second convolutional layer, a second pooling layer and a full-connection layer, and defining a network;
s32, configuring a training environment and creating an actuator for the built convolutional neural network, and defining the dimension of network output data to be equal to the dimension of a training sample;
and S33, training the neural network model by using the actuator according to a preset training round.
8. The intelligent detection method of inverter output impedance according to claim 7, wherein after converting the sample data into frequency domain data in step S2, obtaining a training sample and a test sample from the converted frequency domain data; or, constructing a third same inverter working platform to obtain a test sample;
step S33 further includes:
s34, after training of preset rounds is completed each time, verifying the neural network model by using the test sample book to obtain a loss value and accuracy;
s35, judging whether the loss value and/or the accuracy obtained through verification meet preset conditions, and if so, taking the current neural network model as the inverter output impedance detection neural network model; otherwise, the process returns to S33-S35.
9. The intelligent detection method according to claim 7, wherein the network defining in step S31 includes: defining the convolution kernel size of the first convolution layer and the activation function of the corresponding layer;
defining the pooling core size of the first pooling layer and the pooling type of the corresponding layer;
defining the convolution kernel size of the second convolution layer and the activation function of the corresponding layer;
defining the pooling core size of the second pooling layer and the pooling type of the corresponding layer;
and defining the size of the full connection layer and the activation function of the corresponding layer.
10. The intelligent detection method of inverter output impedance according to claim 6, wherein in step S1, the actual impedance value of the inverter is measured in real time by using a frequency response analyzer and a linear amplification device, and the operation parameters are synchronously adjusted for the first inverter working platform and the second inverter working platform, so as to obtain sample data under different working conditions and labels corresponding to the same moment.
11. The intelligent detection method according to claim 6, wherein the collected sample data is subjected to fast fourier transform in step S2, and the frequency range is selected to be 50Hz to 6000Hz.
12. An intelligent control method for the running state of an inverter is characterized by comprising the following steps:
detecting the output impedance of the present inverter using the intelligent detection method of the output impedance of the inverter according to any one of claims 6 to 11;
and judging the stability of the current inverter according to an impedance analysis method, and modulating a pulse signal for controlling the running state of the inverter through a self-adaptive algorithm.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202110935878.4A CN113761793B (en) | 2021-08-16 | 2021-08-16 | Inverter output impedance detection device and method and inverter operation control method |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202110935878.4A CN113761793B (en) | 2021-08-16 | 2021-08-16 | Inverter output impedance detection device and method and inverter operation control method |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN113761793A CN113761793A (en) | 2021-12-07 |
| CN113761793B true CN113761793B (en) | 2024-02-27 |
Family
ID=78789505
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202110935878.4A Active CN113761793B (en) | 2021-08-16 | 2021-08-16 | Inverter output impedance detection device and method and inverter operation control method |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN113761793B (en) |
Citations (6)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| BE1020014A5 (en) * | 2012-01-27 | 2013-03-05 | Mindcet Bvba | METHODS AND SYSTEM FOR CALCULATING MODEL PARAMETERS FOR A SPOOL TO BE MODELED. |
| CN105680474A (en) * | 2016-02-22 | 2016-06-15 | 中国电力科学研究院 | Control method for restraining rapid power change of photovoltaic station based on energy storage system |
| CN111157798A (en) * | 2020-01-03 | 2020-05-15 | 国网吉林省电力有限公司 | Impedance measurement system based on real-time simulator and object controller |
| WO2020106894A1 (en) * | 2018-11-20 | 2020-05-28 | Georgia Tech Research Corporation | Systems and methods for enhanced engineering design and optimization |
| CN111262238A (en) * | 2020-02-12 | 2020-06-09 | 浙江大学 | Machine learning-based method for predicting short-circuit current of power distribution network containing IIDG |
| CN112946363A (en) * | 2019-12-11 | 2021-06-11 | 国网湖北省电力有限公司电力科学研究院 | Neural network-based power grid impedance online identification method |
Family Cites Families (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US9140731B2 (en) * | 2011-09-16 | 2015-09-22 | Virginia Tech Intellectual Properties, Inc. | Algorithm and implementation system for measuring impedance in the D-Q domain |
| US9429610B2 (en) * | 2014-01-16 | 2016-08-30 | Qualcomm Incorporated | Voltage dependent die RC modeling for system level power distribution networks |
| US11393546B2 (en) * | 2019-07-19 | 2022-07-19 | Silicon Storage Technology, Inc. | Testing circuitry and methods for analog neural memory in artificial neural network |
-
2021
- 2021-08-16 CN CN202110935878.4A patent/CN113761793B/en active Active
Patent Citations (6)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| BE1020014A5 (en) * | 2012-01-27 | 2013-03-05 | Mindcet Bvba | METHODS AND SYSTEM FOR CALCULATING MODEL PARAMETERS FOR A SPOOL TO BE MODELED. |
| CN105680474A (en) * | 2016-02-22 | 2016-06-15 | 中国电力科学研究院 | Control method for restraining rapid power change of photovoltaic station based on energy storage system |
| WO2020106894A1 (en) * | 2018-11-20 | 2020-05-28 | Georgia Tech Research Corporation | Systems and methods for enhanced engineering design and optimization |
| CN112946363A (en) * | 2019-12-11 | 2021-06-11 | 国网湖北省电力有限公司电力科学研究院 | Neural network-based power grid impedance online identification method |
| CN111157798A (en) * | 2020-01-03 | 2020-05-15 | 国网吉林省电力有限公司 | Impedance measurement system based on real-time simulator and object controller |
| CN111262238A (en) * | 2020-02-12 | 2020-06-09 | 浙江大学 | Machine learning-based method for predicting short-circuit current of power distribution network containing IIDG |
Non-Patent Citations (2)
| Title |
|---|
| Non-communication and artificial neural network based photovoltaic monitoring using the existing impedance relay;Heba M. Abdullah等;Sustainable Energy, Grids and Networks;第22卷;1-13 * |
| 基于BP神经网络的SSSC的建模与仿真;林俊波;蒋程;;电气开关(第04期);72-75 * |
Also Published As
| Publication number | Publication date |
|---|---|
| CN113761793A (en) | 2021-12-07 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN103728535A (en) | Extra-high-voltage direct-current transmission line fault location method based on wavelet transformation transient state energy spectrum | |
| CN111027408A (en) | Load identification method based on support vector machine and V-I curve characteristics | |
| CN108155648A (en) | Method for estimating state based on the infinite Extended Kalman filter of adaptive H | |
| CN107219447B (en) | A kind of direct-current arc detection method and system based on impedance characteristic | |
| CN104616061B (en) | Island detection method based on wavelet packet logarithmic energy entropy and genetic algorithm optimization | |
| CN105938578A (en) | Large-scale photovoltaic power station equivalent modeling method based on clustering analysis | |
| CN110363334B (en) | Grid line loss prediction method of photovoltaic grid connection based on gray neural network model | |
| CN104410360A (en) | Safe operation method of photovoltaic power generation system, training method for artificial neural network and real-time detection method in safe operation method, and real-time detection device | |
| CN105182141A (en) | Transformer fault detection method and system | |
| CN110045317B (en) | Mutual inductor metering error online detection method and system | |
| Soliman et al. | Capacitance estimation for dc-link capacitors in a back-to-back converter based on artificial neural network algorithm | |
| CN106250904A (en) | Based on Power Disturbance analyser and the sorting technique of improving S-transformation | |
| Azad et al. | Parameter estimation of doubly fed induction generator driven by wind turbine | |
| CN109193635A (en) | A kind of power distribution network topological structure method for reconstructing based on adaptive sparse homing method | |
| CN104638671A (en) | Island detection method based on introducing particle swarm optimization and phase disturbance | |
| CN108108672A (en) | A kind of weak information identifying method of accidental resonance electric current based on linear search strategy | |
| CN113158781B (en) | Lightning trip type identification method | |
| CN106093516B (en) | Replay method based on the transmission line malfunction current traveling wave waveform that multiband decomposes | |
| CN113761793B (en) | Inverter output impedance detection device and method and inverter operation control method | |
| CN110210152A (en) | A kind of ultra harmonics source modeling method | |
| CN104332984B (en) | A kind of node voltage based on noise like influences each other the on-line identification method of the factor | |
| CN114236234A (en) | Electrical appliance characteristic identification method based on fundamental wave and harmonic wave mixed criterion | |
| CN113746132B (en) | Photovoltaic power station based on cloud edge cooperation and control method thereof | |
| CN104730384A (en) | Power disturbance identification and localization method based on incomplete S transformation | |
| CN107069710B (en) | Power system state estimation method considering new energy space-time correlation |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| CB02 | Change of applicant information |
Address after: 215000 No. 90, Zijin Road, high tech Zone, Suzhou, Jiangsu Applicant after: Goodway Technology Co.,Ltd. Address before: 215163 No.90 Zijin Road, high tech Zone, Suzhou City, Jiangsu Province Applicant before: JIANGSU GOODWE POWER SUPPLY TECHNOLOGY Co.,Ltd. |
|
| CB02 | Change of applicant information | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| GR01 | Patent grant | ||
| GR01 | Patent grant |