Calibration method based on pcs parameter representation

Abstract

The application discloses a calibration method based on PCS parameter representation, comprising the following steps: constructing a calibration network, data preprocessing, model training and data generation, and calibration processing; the calibration network is composed of a detection interface, a preprocessing unit, a model generator and a calibration unit; electrical quantity data collected by an electric meter is taken as a training sample set, and electrical quantity data collected by a PCS is taken as a measured sample set, and the training sample set and the measured sample set are respectively marked with unique determined labels; the training sample set and the measured sample set are trained respectively, so that a calibration target value of the electric meter and an actual measurement value of the PCS are obtained; the calibration unit calculates a calibration coefficient and a bias of the PCS based on the calibration target value of the electric meter corresponding to each label, and calibrates the actual measurement value of the corresponding PCS according to the calculated bias. Compared with an artificial calibration method, the calibration method greatly reduces artificial input cost and experimental cost, and significantly improves calibration efficiency.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of power electronic devices, in particular to a calibration method based on PCS parameter representation. BACKGROUND

[0002] For energy storage systems and power conversion, energy storage converters (PCS) are indispensable. The bi-directional conversion of energy storage system and grid power is based on the controllable charging and discharging process of PCS, which converts AC to DC and vice versa. The PCS can convert the DC power of the energy storage system into AC power and deliver it to the grid or AC load. It can also convert the AC power of the grid into DC power and charge the energy storage system. Under grid-connected conditions, the energy storage system can be controlled to maintain constant power or current to charge or discharge the battery, and smooth the output of fluctuating power sources such as wind and solar power. Under microgrid conditions, SVPWM or SPWM with closed-loop control can accurately and quickly regulate output voltage, frequency, active and reactive power. With the development of installed system capacity, the PCS needs to be updated and developed to meet the demand.

[0003] However, for PCS, multiple PCSs may be used in series or parallel. To ensure that the PCS parameters meet the requirements of any operating condition before shipment, the parameters of each PCS need to be accurately calibrated. The previous PCS calibration method requires manual measurement and comparison using experimental instruments, which not only requires a certain amount of labor cost, but also increases the experimental cost and has low calibration efficiency. SUMMARY

[0004] Therefore, the purpose of the present application is to provide a calibration method based on PCS parameter representation to solve the technical problems mentioned in the prior art.

[0005] The calibration method based on PCS parameter representation comprises the following steps:

[0006] S1, constructing a calibration network: the calibration network is composed of a detection interface, a preprocessing unit, a model generator and a calibration unit; connecting the data acquisition end of an electric meter and a PCS to a to-be-tested circuit to collect electrical quantity data in real time, and connecting the detection interface to the communication interface end of the electric meter and the PCS through a communication bus to read the electrical quantity data collected by the electric meter and the PCS;

[0007] S2, data preprocessing: the preprocessing unit selects the electrical quantity data collected by the electric meter in any detection cycle as a training sample set, and selects the electrical quantity data collected by the PCS as a measured sample set, and labels the training sample set and the measured sample set with unique labels;

[0008] S3, model training and data generation: input the training sample set into the model generator for training to obtain the calibration target value of the electric meter; input the measured sample set into the model generator for training to obtain the actual measurement value of the PCS;

[0009] S4, calibration processing: the calibration unit calculates the calibration coefficient and bias of the PCS based on the calibration target value of the electric meter corresponding to each label, and calibrates the actual measurement value of the corresponding PCS according to the calculated bias, wherein;

[0010] ;

[0011] .

[0012] Optionally, the preprocessing unit labels the training sample set or the measured sample set with a uniquely determined label according to a detection path;

[0013] The label information of the detection path is set as any one or more of the voltage and / or current of the A-phase AC bus, the voltage and / or current of the B-phase AC bus, the voltage and / or current of the C-phase AC bus, and the voltage and / or current of the DC bus.

[0014] Optionally, the preprocessing unit discards the electrical quantity data collected by the electric meter beyond the preset detection threshold range, and takes the electrical quantity data within the preset detection threshold range as the training sample set.

[0015] Optionally, the method for setting the preset detection threshold range comprises:

[0016] Based on the detection path of the electric meter corresponding to each kind of label, the average value of all electrical quantity data collected within the detection period is calculated, and a preset error allowance value is added to the average value to generate a preset detection threshold range corresponding to the label.

[0017] Optionally, the model generator is sequentially provided with an input layer, a hidden layer, a summation layer and an output layer connected to each other along the direction of data flow;

[0018] The input layer is used to identify the label information of the sample data;

[0019] The hidden layer is configured with a first neuron corresponding to each kind of label, and the first neuron is used to capture and store the sample data labeled with the corresponding label;

[0020] The summation layer is configured with a second neuron for any number of the first neurons, and the second neuron is provided with a summation node for synchronously summing all the sample data stored in the interrelated number of the first neurons to obtain an average value.

[0021] The output layer outputs the calculated average value as an output result.

[0022] Optionally, before calculating the calibration coefficient and bias of the PCS, the calibration unit further comprises:

[0023] The actual measurement value of the PCS is classified into the calibration target value category of the electric meter closest to it, and the calibration coefficient and bias of the PCS are calculated based on the closest calibration target value of the electric meter.

[0024] The present application can produce beneficial effects, including:

[0025] The calibration method based on PCS parameter representation provided by the present application constructs a calibration network to automatically and real-time collect electrical quantity data of the electric meter and the PCS, and pre-processes the electrical quantity data collected by the electric meter to eliminate data exceeding the preset detection threshold range, so as to ensure the quality of the training sample set and improve the accuracy of subsequent model training and calibration; and the model generator synchronously processes different label data, which greatly reduces the labor input cost and experimental cost compared with the manual calibration method, and significantly improves the calibration efficiency. BRIEF DESCRIPTION OF DRAWINGS

[0026] Figure 1 The figure is a flowchart of the calibration method based on PCS parameter representation of the present application;

[0027] Figure 2 The figure is an architecture diagram of the model generator in the calibration method of the present application;

[0028] Figure 3 The figure is a line connection diagram of the circuit to be tested in the calibration method of the present application. DETAILED DESCRIPTION

[0029] The technical solutions in the embodiments of the present application will be described clearly and completely below with reference to the drawings in the embodiments of the present application. Obviously, the described embodiments are only part of the embodiments of the present application, rather than all the embodiments. Based on the embodiments in the present application, all other embodiments obtained by those skilled in the art without creative labor fall within the scope of protection of the present application.

[0030] Please refer to Figure 1 The present application provides a calibration method based on PCS parameter representation, and the calibration method comprises the following steps:

[0031] Step one, construct the calibration network: the calibration network is composed of detection interface, preprocessing unit, model generator and calibration unit; as shown in the figure, the ammeter and the data acquisition end of the PCS are connected to the designated key node position of the circuit to be tested, so as to realize the real-time and accurate collection of various dynamic electrical quantity data in the circuit to be tested, wherein the ammeter can use AD conversion chip to collect electrical quantity data, and the PCS can use analog-to-digital converter (ADC) to collect electrical quantity data; and the communication interface end (such as RS485, RS232, Ethernet, etc.) of the ammeter and the PCS is connected to the detection interface through the communication bus, so as to realize reading the electrical quantity data collected by the ammeter and the PCS, and provide effective data support for subsequent data processing and analysis. Figure 3

[0032] Step two, data preprocessing: the preprocessing unit selects the electrical quantity data collected by the ammeter in any one detection cycle as the training sample set, and the electrical quantity data collected by the PCS as the measured sample set, and labels the training sample set and the measured sample set with unique labels, which are labeled according to the detection path. The label information of the detection path is specifically set as the voltage and / or current of phase A AC bus, the voltage and / or current of phase B AC bus, the voltage and / or current of phase C AC bus, and the voltage and / or current of DC bus. In addition, the preprocessing unit will eliminate the electrical quantity data collected by the ammeter that exceeds the preset detection threshold range, and only include the electrical quantity data within the preset detection threshold range into the training sample set. The detection threshold range is set as follows: for each detection path corresponding to the ammeter of each label, calculate the average value of all electrical quantity data collected in the detection cycle, and then add a set error allowance value based on the average value to generate the detection threshold range corresponding to the label.

[0033] Step three, model training and data generation: input the training sample set into the model generator for training, and finally obtain the calibration target value of the ammeter; input the measured sample set into the model generator for training, and obtain the actual measurement value of the PCS. Specifically, the model generator is sequentially provided with an input layer, a hidden layer, a summation layer and an output layer connected to each other according to the data flow. Among them, the input layer is responsible for identifying the label information of the sample data; the hidden layer is configured with a first neuron corresponding to each label, which is used to capture and store sample data labeled with the corresponding label; the summation layer is configured with a second neuron for any number of first neurons, and the second neuron is provided with a summation node which can synchronously sum all sample data stored in the first neurons associated with each other to obtain the average value; the output layer outputs the calculated average value as the output result, which is specifically the calibration target value of the ammeter or the actual measurement value of the PCS.​

[0034] Step four, calibration process: before calculating the calibration coefficient and bias of the PCS, the actual measurement value of the PCS needs to be classified into the closest calibration target value category of the electric meter, and then the calibration coefficient and bias of the PCS are calculated according to the closest calibration target value of the electric meter, wherein the calibration coefficient is kept to three decimal places; and then the actual measurement value of the corresponding PCS is calibrated according to the calculated bias;

[0035] ;

[0036] .

[0037] For example, 670-685V is taken as the training label 1 of the electric meter, and 705V-730V is taken as the training label 2 of the electric meter; if the actual measurement voltage value collected by the PCS at this time is 700V, which is closer to the training label 2, it is classified into the training label 2. Further, if the real-time calibration target voltage collected by the electric meter corresponding to the training label 2 is 705V, at this time,

[0038] ;

[0039] ;

[0040] Therefore, the actual measurement value of the PCS is adjusted up by 0.001V by the calibration unit, and the final electrical quantity data of the PCS is 700.001V.

[0041] In the above, when the electric meter and the PCS are connected to the to-be-tested circuit, first, the line connection test of the electric meter and the PCS is performed respectively, wherein:

[0042] 1) Function test of the electric meter: a standard signal source is used to input known standard voltage and current signals to the electric meter to simulate the electrical quantity parameters in the actual circuit. The display interface of the electric meter is observed to confirm whether it can accurately display the input voltage and current values, and whether the error is within the accuracy allowed range of the electric meter. At the same time, various measurement functions of the electric meter are tested, such as active power, reactive power, power factor, etc., to check the accuracy of the measurement results. For example, by inputting a test signal with a specific power factor to the electric meter, the power factor value displayed by the electric meter is compared with the standard value to judge the reliability of the electric meter in this function.

[0043] 2) Functional Testing of the PCS: Targeted tests are conducted on the core functions of the PCS, such as power conversion and data acquisition. For power conversion testing, the PCS is connected to a simulated power supply to observe its ability to stably convert input electrical energy into the required output form according to the set conversion mode, such as converting DC power to AC power, and ensuring that the output power quality meets relevant standards, including voltage fluctuations and frequency deviations within allowable ranges. For the PCS's data acquisition function, its internal self-test program is triggered upon initial power-on or connection to the circuit under test to collect simulated electrical quantity data. The accuracy and completeness of the data acquisition are then checked to ensure accurate acquisition of various electrical parameters without data loss or erroneous acquisition.

[0044] For example, part of the PCS self-test procedure is as follows:

[0045] "switch(st_FlagSelfTestState.bit.u2_InvVoltSampleState)

[0046] {

[0047] case UNCHECK:

[0048] i16_SelfTest1msCnt1 = 0;

[0049] i16_SelfTest1msCnt2 = 0;

[0050] st_FlagSelfTestState.bit.u2_InvVoltSampleState = CHECKING;

[0051] break

[0052] case CHECKING:

[0053] f_TempAx = fabs(st_Uinv3S2R_S.f_DirectAxis);

[0054] f_TempBx = fabs(st_Uinv3S2R_S.f_QuadAxis);

[0055] f_TempCx = fabs(f_InvCapUab + f_InvCapUbc + f_InvCapUca); / / The sum of the three-phase inverter voltages is theoretically 0

[0056] if ((f_TempCx < 76.0f) && (f_TempAx < 1.63f * f_UNIT_RATED_AC_VOLTAGE) / / sampling range is ±1842V, 1.628 times of rated value

[0057] && (f_TempBx < 1.63f * f_UNIT_RATED_AC_VOLTAGE))

[0058] {

[0059] i16_SelfTest1msCnt1 ++;

[0060] }

[0061] if (++i16_SelfTest1msCnt2 == 100)

[0062] {

[0063] st_FlagSelfTestState.bit.u2_InvVoltSampleState = (i16_SelfTest1msCnt1 > 90)? CHECKPASS”。

[0064] 3) Verify the device compatibility of the electric meter and the PCS: Considering that the to-be-tested circuit may have complex situations of multiple electrical devices working together, the compatibility of the PCS and the electric meter with other devices needs to be verified. The PCS and the electric meter are connected to a simulated circuit environment containing common devices in the to-be-tested circuit, and it is observed whether they will interfere with each other when working together. For example, it is checked whether the measurement accuracy of the electric meter will be affected by the electromagnetic radiation of the PCS, or whether the working state of the PCS will abnormally fluctuate due to the connection of the electric meter, to ensure that in the actual application environment, they can coexist harmoniously with other devices and run stably.

[0065] In the present application, the traditional power system operation and maintenance relies on a large number of manual inspection and experience judgment, which has high cost and limited accuracy. The calibration method realizes real-time monitoring and fault warning of the operation state of power equipment through continuous and accurate electrical quantity monitoring and calibration. For example, through long-term calibration analysis of the PCS data of the substation, potential overheat and insulation aging faults of the transformer can be found in advance, and the maintenance personnel can arrange maintenance in advance to avoid high maintenance cost and large-scale power outage loss caused by equipment sudden failure. According to actual case statistics, line faults occur from time to time when the smart grid is running. The calibration method can quickly capture the abnormal fluctuation of electrical quantities such as fault instantaneous current and voltage due to its real-time acquisition of electrical quantity data. For example, when a line short-circuit fault occurs in a regional power grid due to lightning strike, the data collected by the electric meter and PCS in real time can be quickly analyzed by the calibration network to accurately locate the fault position of the A-phase AC bus or other specific line position. According to the accurate data after calibration, the maintenance personnel can quickly judge the fault type and severity, greatly shorten the fault troubleshooting time from several hours to within half an hour, greatly improve the power grid repair efficiency, reduce the economic loss and social influence caused by power outage, and reduce the overall operation and maintenance cost of the power system by 20%-30%, greatly improving the economic benefit of the power enterprise.

Claims

1. Calibration method based on the characterization of PCS parameters, characterized in that, The calibration method comprises the following steps: S1, constructing a calibration network: the calibration network is composed of a detection interface, a preprocessing unit, a model generator and a calibration unit; the data acquisition end of an electric meter and a PCS are respectively connected to a to-be-tested circuit to collect electrical quantity data in real time, and the detection interface is connected to the communication interface end of the electric meter and the PCS through a communication bus to read the electrical quantity data collected by the electric meter and the PCS; S2, data preprocessing: the preprocessing unit selects the electrical quantity data collected by the electric meter in any one detection cycle as a training sample set, and selects the electrical quantity data collected by the PCS as a measured sample set, and labels the training sample set and the measured sample set with unique and determined labels respectively; S3, model training and data generation: inputting the training sample set into the model generator for training to obtain the calibration target value of the electric meter; inputting the measured sample set into the model generator for training to obtain the actual measurement value of the PCS; wherein the model generator is sequentially provided with an input layer, a hidden layer, a summation layer and an output layer connected with each other along the data flow direction; the input layer is used for identifying the label information of sample data; the hidden layer is configured with a first neuron corresponding to each label, and the first neuron is used for capturing and storing the sample data labeled with the corresponding label; the summation layer is configured with a second neuron for any number of first neurons, and the second neuron is provided with a summation node, which is used for synchronously summing all sample data stored in the mutually associated first neurons to obtain an average value; and the output layer outputs the calculated average value as an output result; S4, calibration processing: the calibration unit calculates the calibration coefficient and bias of the PCS based on the calibration target value of the electric meter corresponding to each label, and calibrates the actual measurement value of the corresponding PCS according to the calculated bias, wherein; ； 。 2. The calibration method based on PCS parameter characterization of claim 1, wherein, The preprocessing unit labels the training sample set or the measured sample set with a unique and determined label according to a detection path; The label information of the detection path is set as any one or more of the voltage and / or current of phase A AC bus, the voltage and / or current of phase B AC bus, the voltage and / or current of phase C AC bus and the voltage and / or current of DC bus.

3. The calibration method based on PCS parameter characterization of claim 2, wherein, The preprocessing unit excludes the electrical quantity data collected by the electric meter that exceeds the preset detection threshold range, and takes the electrical quantity data within the preset detection threshold range as the training sample set.

4. The calibration method based on PCS parameter characterization of claim 3, wherein, The setting method of the preset detection threshold range comprises: Based on the detection path of the electric meter corresponding to each label, the average value of all electrical quantity data collected in the detection cycle is calculated, and a preset detection threshold range corresponding to the label is generated based on the average value plus a set error allowance.

5. The calibration method based on PCS parameter characterization of claim 1, wherein, Before calculating the calibration coefficient and bias of the PCS, the calibration unit further comprises: The actual measured value of the PCS is classified into the calibration target value category of the electric meter which is closest to it, and the calibration coefficient and bias of the PCS are calculated based on the calibration target value of the closest electric meter.

Images