Temperature compensation method and system for all-fiber current transformer based on BP (Back Propagation) network

Abstract

The invention provides an all-fiber current transformer temperature compensation method and system based on a BP network. The method comprises the following steps: obtaining all-fiber current transformer operation data through a physical test; processing the initial operation data of the all-fiber current transformer, and extracting sample data; and establishing an all-fiber current transformer temperature compensation model based on a BP neural network algorithm. The technical problem that the measurement accuracy of the all-fiber current transformer is reduced when the ambient temperature changes due to the fact that the influence of uneven temperature is not fully considered is solved.

Description

Technical Field

[0001] The invention relates to the technical field of all-optical current transformers, and in particular to a temperature compensation method and system for an all-optical current transformer based on a BP network. Background Art

[0002] Due to the increasing demand for electricity and China's energy distribution problems, the power system has gradually developed from ultra-high voltage to ultra-high voltage. For ultra-high voltage transmission systems, accurate acquisition of primary-side current signals is the basis for ensuring the safe, stable and economical operation of ultra-high voltage transmission systems, and plays a vital role in the accuracy of power information data measurement and the reliability of relay protection. Traditional electromagnetic current transformers have shortcomings such as magnetic saturation, high insulation requirements, and lack of information interconnection functions. They cannot meet the needs of ultra-high voltage transmission systems and digital and intelligent construction of power grids. All-fiber current transformers have no magnetic saturation, simple insulation structure, wide dynamic range, and can output digital and analog interfaces. They are more adaptable to the needs of power system construction and have been increasingly widely used in the ultra-high voltage field in recent years.

[0003] However, in actual applications, temperature is always the main factor restricting the large-scale application of all-fiber current transformers. The measurement signal will produce a ratio error offset due to the influence of temperature, which will cause the measurement accuracy of the all-fiber current transformer to decrease. The temperature fluctuation will also directly affect the long-term operation stability of the all-fiber current transformer. In addition, the polarization-maintaining transmission optical fiber between the sensing ring and the acquisition unit will be arranged in the insulator in practice. It will show uneven temperature distribution due to the heating of the wire, the surrounding wind speed and the ambient temperature, which will further affect the measurement error. Considering only the influence of temperature, when the ambient temperature is in the range of 10℃ to 60℃, the influence of temperature change on the Verdet constant and linear birefringence of the sensing optical fiber makes the FOCT ratio error between -0.14% and 0.38%. In this temperature range, it can no longer meet the accuracy requirement of level 0.2. In extreme environments, the performance of the equipment will become worse.

[0004] Therefore, researchers at home and abroad have proposed various methods to solve the problem of temperature sensitivity of all-fiber current transformers. Specifically, they can be divided into two directions. One direction is to compensate through hardware, which can greatly reduce the temperature sensitivity of all-fiber current transformers. However, the manufacturing process is demanding and the cost is high, which is not conducive to large-scale applications. The other direction is to compensate through software, which can make the all-fiber current transformer meet the 0.2 level accuracy requirement in a wide temperature range of -10 to 70°C without too much cost.

[0005] In the existing related technologies, the Chinese invention patent application document with publication number CN112986650A discloses an all-fiber current transformer and a temperature compensation method thereof, which includes: (1) arranging a temperature conductive material similar to the metal shell of the sensing ring near the temperature sensor to simulate a temperature environment similar to that inside the sensing ring; (2) the temperature sensor is at a certain distance from the sensing ring, but the temperature environment near the temperature sensor is consistent with that of the sensing ring. This method simulates the same temperature environment as that around the sensing ring for the temperature sensor in order to measure the actual temperature of the optical fiber inside the sensing ring more accurately. However, in practice, it is difficult to keep the temperature environments of the two consistent, that is, the temperature environment of the temperature sensor cover itself is different from the temperature inside the sensing ring, which will further increase the temperature measurement error of the temperature sensor, and the influence of the polarization-maintaining transmission optical fiber is not considered.

[0006] The Chinese invention patent application document with publication number CN116773890A discloses an all-optical current transformer and a temperature compensation method thereof. In this method, a curve fitting method is used to establish a correspondence between the error of the all-optical current transformer and the temperature data of the optical fiber temperature sensor, so as to calculate the compensation coefficient to perform temperature compensation on the all-optical current transformer. However, the compensation accuracy of the curve fitting method is limited, and in the face of complex data, it will significantly increase the complexity of the model and the calculation cost.

[0007] In summary, the prior art has a technical problem in that the influence of uneven temperature is not sufficiently considered, resulting in a decrease in the measurement accuracy of the all-optical current transformer when the ambient temperature changes. Summary of the invention

[0008] The technical problem to be solved by the present invention is how to solve the technical problem that the influence of uneven temperature is not sufficiently considered in the prior art, resulting in a decrease in the measurement accuracy of the all-optical current transformer when the ambient temperature changes.

[0009] Existing studies on temperature compensation of all-fiber current transformers all regard the environment in which the all-fiber current transformer is located as uniform temperature, and use the temperature of the temperature control box and the temperature change rate or the temperature of the temperature control box and the output data of the all-fiber current transformer as the input of the neural network, and perform temperature compensation after network training. However, the use of a temperature compensation model that ignores the influence of uneven temperature will lead to a decrease in the measurement accuracy of the all-fiber current transformer. The temperature compensation method of the all-fiber current transformer based on the BP network proposed in the present invention uses the output signal of the all-fiber current transformer, the multi-point real-time temperature of the temperature sensor, the humidity around the all-fiber current transformer, the surrounding wind speed, and the historical conductor current carrying capacity as multiple network inputs, fully considering the uneven temperature distribution characteristics of the environment in which the all-fiber current transformer is located in actual applications. The present invention effectively solves the problem of decreased measurement accuracy of the all-fiber current transformer under the influence of uneven temperature.

[0010] The present invention adopts the following technical solutions to solve the above technical problems: the temperature compensation method of all-optical current transformer based on BP network includes:

[0011] S1. Using a collection unit to obtain the operating data of the all-optical current transformer in real time, in the collection unit, the operating data of the all-optical current transformer is transmitted to the host computer in real time through the serial port and the optical port line;

[0012] S2, downsampling the operating data of the all-optical current transformer to obtain measurement data with consistent sampling rate as input samples of the BP neural network;

[0013] S3. Establish a BP neural network model, train the BP neural network model, obtain a temperature compensation model for an all-optical current transformer, perform model prediction based on it, and obtain a suitable temperature compensation result to perform temperature compensation on the all-optical current transformer.

[0014] The present invention performs temperature compensation on an all-fiber current transformer based on a BP neural network and considers the influence of uneven temperature, thereby solving the problem of decreased measurement accuracy of the all-fiber current transformer when the ambient temperature changes. The present invention performs temperature compensation on the all-fiber current transformer based on the prediction results of the BP neural network, and verifies that the compensated all-fiber current transformer meets the 0.2 level accuracy requirement. This method has important reference significance for improving the long-term working stability of the all-fiber current transformer.

[0015] In a more specific technical solution, in S1, the operating data of the all-optical current transformer includes: the output signal X of the all-optical current transformer, the real-time temperature T (T1, T2, T3) of the temperature sensor, the humidity H around the all-optical current transformer, the surrounding wind speed W and the historical conductor current carrying capacity I, which are used as the input of the subsequent BP neural network.

[0016] Specifically, the operating data can be expressed as:

[0017] Input={X,T(T1,T2,T3),H,W,I}

[0018] The present invention performs temperature compensation on an all-fiber current transformer based on a BP neural network and takes the influence of uneven temperature into consideration, thereby solving the problem of reduced measurement accuracy of the all-fiber current transformer when the ambient temperature changes.

[0019] In a more specific technical solution, the real-time temperature of the temperature sensor includes: the fiber ring temperature T1, the polarization-maintaining transmission fiber temperature T2 and the ambient temperature T3.

[0020] In a more specific technical solution, in S1, when measuring the temperature of the transmission optical fiber, no less than two temperature sensors are used to measure the temperature of multiple points on the optical fiber inside the optical fiber ring and the polarization-maintaining transmission optical fiber respectively;

[0021] Use humidity sensor to measure humidity around all-fiber current transformer;

[0022] Use an anemometer to measure the surrounding wind speed;

[0023] According to the historical measurement data of the all-optical current transformer, the historical conductor current carrying capacity is obtained by processing.

[0024] Compared with the prior art, the present invention further improves the temperature compensation accuracy of the all-optical fiber current transformer, so that the measurement accuracy and stability of the all-optical fiber current transformer are further improved.

[0025] In a more specific technical solution, a temperature sensor is used to measure the temperature of the optical fiber inside the optical fiber ring, and the remaining temperature sensors are used to measure the temperature in the middle of the polarization-maintaining transmission optical fiber and the temperature at the end of the polarization-maintaining transmission optical fiber, and the temperature distribution curve of the sensing optical fiber is obtained by fitting.

[0026] The present invention fully considers the influence of uneven temperature distribution of polarization-maintaining transmission optical fiber on the output signal of the all-optical fiber current transformer, and can obtain a more accurate temperature distribution curve of the sensing optical fiber according to the temperature measurement data.

[0027] In a more specific technical solution, in S2, the obtained all-optical current transformer operation data is processed and cleaned, and the abnormal point is replaced by the average value of the sampling point data before and after the abnormal data point, so as to achieve the purpose of removing the data abnormal point, wherein the abnormal data point includes a mutation point and a missing point; data downsampling is performed to unify the sampling rates of different parameters; for a set of data D = {…, d1, d, d2, …}, when d is an abnormal data point, the new value d after data processing and cleaning can be expressed as:

[0028]

[0029] The data after data processing and cleaning is:

[0030]

[0031] Data downsampling, that is, unifying the sampling rates of different parameters;

[0032] Specifically: when the output signal X of the all-fiber current transformer collects four data points per second, that is, X = {x1, x2, x3, x4}, and the fiber ring temperature T1 collects one data point per second, that is, T1 = {t1}, the output signal of the all-fiber current transformer is averaged by every four data points to replace the original data, that is, Thus, the data sampling rate is unified, and finally the input data of the BP neural network is obtained;

[0033] The measurement data with the same sampling rate are randomly divided into training set Input1 and test set Input2, and the input samples are input into the BP neural network to obtain the prediction results after training;

[0034] The BP neural network is trained and the prediction result is used as the compensation value to perform temperature compensation on the all-optical current transformer.

[0035] In more specific technical solutions, S3 includes:

[0036] S31, importing the input sample into the BP neural network model;

[0037] S32, dividing the input samples into a training set Input1 and a test set Input2, setting the number of training set samples, and normalizing the input samples;

[0038] Specifically, in S32, the input samples are randomly divided according to the ratio of 3:1 between the number of training set samples and the number of test set samples. The random function randperm(n) can be used to set a random sequence for the input samples to achieve randomness, where n is the total number of samples. Then, the training set and the test set are normalized, and the normalization formula is:

[0039]

[0040] In the formula: x represents the normalized data; x0 represents the original data; x max 、x min are the maximum and minimum values ​​in this group of data respectively;

[0041] S33, setting model parameters and training parameters of the BP neural network model, and initializing weights and thresholds of the BP neural network model;

[0042] S34. Using the BP neural network model, perform model training based on input samples, solve weights and thresholds until the preset target error is met, and obtain applicable prediction output results of the all-optical current transformer temperature compensation model and the BP neural network model.

[0043] S35, using the applicable prediction output result as a compensation parameter to compensate for the temperature error of the all-optical current transformer;

[0044] S36. Compare the applicable prediction output result with the true value and calculate the correlation index to verify the temperature compensation effect of the BP neural network model, wherein the correlation index includes: determination coefficient R, absolute error MAE and root mean square error RMSE.

[0045] The multi-parameter input adopted by the present invention can greatly improve the stability and accuracy of the model. After the neural network training, the prediction result is used as the compensation amount, thereby realizing the temperature compensation of the all-optical current transformer based on the BP neural network algorithm, and greatly improving the temperature stability of the all-optical current transformer. The allocated data set is normalized to improve the training efficiency and stability of the neural network.

[0046] In a more specific technical solution, in S33, the model parameters include: the number M of input layer nodes, the number L of hidden layer nodes and the number N of output layer nodes.

[0047] Specifically, the number of input layer nodes M is the number of parameters of network input, and the input parameters are the output signal X of the all-optical current transformer, the real-time temperature T (T1, T2, T3) of the temperature sensor, the humidity H around the all-optical current transformer, the surrounding wind speed W and the historical conductor current carrying capacity I, that is, M=5; the number of output layer nodes N is the number of parameters of the network output, and the present invention uses the output ratio difference of the all-optical current transformer as the output, that is, N=1; and the number of hidden layer nodes L is determined by an empirical formula, which can be expressed as:

[0048]

[0049] Wherein: a is an integer between [1, 10].

[0050] In a more specific technical solution, in S34, when the preset target error is met, the applicable training parameters are saved and the temperature compensation model of the all-optical current transformer is set accordingly.

[0051] Specifically, the formula for solving the weight and threshold is:

[0052]

[0053] Where: Q i , Y k are the hidden layer output and the output layer output respectively; f is the transfer function of the hidden layer and the output layer. The transfer function of the hidden layer is generally the tansig function, and the transfer function of the output layer is generally the purelin function; ω ij ,ω jk are the weights from the input layer to the hidden layer and from the hidden layer to the output layer respectively; x i is the input of the neural network; b j is the bias of the jth neuron in the hidden layer; c k is the bias of the kth neuron in the output layer.

[0054] The preset target error function error is the sum of squares of the errors between the predicted output value of the neural network and the actual output value, which can be expressed as:

[0055]

[0056] Where: y i Predict the output value for the i-th neural network model; is the true output value of the ith neural network model; n is the number of samples; when error<10 -6 Or the training ends after reaching the maximum number of iterations.

[0057] After the training is completed, the predicted data is denormalized. The denormalization formula is:

[0058]

[0059] Where: y is the denormalized data; y0 is the original data; y max ,y min are the maximum and minimum values ​​in the original data respectively.

[0060] In a more specific technical solution, in S35, the predicted output ratio difference data of the all-optical current transformer temperature compensation model after training is used to compensate the actual output ratio difference data, so as to realize the temperature error compensation of the all-optical current transformer due to ambient temperature changes.

[0061] In a more specific technical solution, the all-optical current transformer temperature compensation system based on BP network includes:

[0062] An operation data acquisition module is used to acquire the operation data of the all-optical current transformer in real time by using the acquisition unit. In the acquisition unit, the operation data of the all-optical current transformer is transmitted to the host computer in real time through the serial port and the optical port line;

[0063] The sample extraction module is used to perform downsampling operation on the operation data of the all-optical current transformer to obtain measurement data with consistent sampling rate as input samples of the BP neural network. The sample extraction module is connected to the operation data acquisition module;

[0064] The temperature supplement module is used to establish a BP neural network model, train the BP neural network model, obtain the temperature compensation model of the all-optical current transformer, perform model prediction based on it, and obtain the applicable temperature compensation result to perform temperature compensation on the all-optical current transformer. The temperature supplement module is connected to the sample extraction module.

[0065] Compared with the prior art, the present invention has the following advantages:

[0066] The present invention performs temperature compensation on an all-fiber current transformer based on a BP neural network and considers the influence of uneven temperature, thereby solving the problem of decreased measurement accuracy of the all-fiber current transformer when the ambient temperature changes. The present invention performs temperature compensation on the all-fiber current transformer based on the prediction results of the BP neural network, and verifies that the compensated all-fiber current transformer meets the 0.2 level accuracy requirement. This method has important reference significance for improving the long-term working stability of the all-fiber current transformer.

[0067] The present invention performs temperature compensation on an all-fiber current transformer based on a BP neural network and takes the influence of uneven temperature into consideration, thereby solving the problem of reduced measurement accuracy of the all-fiber current transformer when the ambient temperature changes.

[0068] Compared with the prior art, the present invention further improves the temperature compensation accuracy of the all-optical fiber current transformer, so that the measurement accuracy and stability of the all-optical fiber current transformer are further improved.

[0069] The present invention fully considers the influence of uneven temperature distribution of polarization-maintaining transmission optical fiber on the output signal of the all-optical fiber current transformer, and can obtain a more accurate temperature distribution curve of the sensing optical fiber according to the temperature measurement data.

[0070] The multi-parameter input adopted by the present invention can greatly improve the stability and accuracy of the model. After the neural network training, the prediction result is used as the compensation amount, thereby realizing the temperature compensation of the all-optical current transformer based on the BP neural network algorithm, and greatly improving the temperature stability of the all-optical current transformer. The allocated data set is normalized to improve the training efficiency and stability of the neural network.

[0071] The present invention solves the technical problem in the prior art that insufficient consideration is given to the influence of uneven temperature, resulting in a decrease in the measurement accuracy of the all-optical current transformer when the ambient temperature changes. BRIEF DESCRIPTION OF THE DRAWINGS

[0072] Figure 1 This is a schematic diagram of the basic steps of the temperature compensation method for an all-optical current transformer based on a BP network according to Embodiment 1 of the present invention;

[0073] Figure 2 This is a schematic diagram of the physical structure of an all-optical current transformer according to Embodiment 1 of the present invention;

[0074] Figure 3 This is a schematic diagram of specific steps for training a BP neural network according to Embodiment 1 of the present invention;

[0075] Figure 4 This is a diagram showing the temperature compensation effect of an all-optical current transformer based on a BP neural network algorithm according to Example 1 of the present invention. DETAILED DESCRIPTION

[0076] In order to make the purpose, technical solution and advantages of the embodiments of the present invention clearer, the technical solution in the embodiments of the present invention will be clearly and completely described below in combination with the embodiments of the present invention. Obviously, the described embodiments are part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0077] Example 1

[0078] like Figure 1 As shown, the temperature compensation method of the all-optical current transformer based on the BP network provided by the present invention includes the following basic steps:

[0079] S1. Obtaining the operating data of the all-optical current transformer through physical tests;

[0080] In this embodiment, the operating data of the all-optical current transformer is obtained in real time, and each data sample corresponds to the output signal of the all-optical current transformer, the real-time temperature of the temperature sensor, the humidity around the all-optical current transformer, the surrounding wind speed, and the historical conductor current carrying capacity measured at the same time.

[0081] In this embodiment, the operating data of the all-optical current transformer is collected and the temperature compensation of the all-optical current transformer is performed using the BP neural network algorithm. The operating data of the all-optical current transformer includes but is not limited to: the output signal of the all-optical current transformer, the real-time temperature of the temperature sensor, the humidity around the all-optical current transformer, the surrounding wind speed, and the historical conductor current carrying capacity; the real-time temperature data of the temperature sensor is multi-point temperature measurement data, including but not limited to: the fiber ring temperature, the polarization-maintaining transmission fiber temperature, and the ambient temperature; the multi-point temperature measurement data is used to establish the temperature distribution curve of the sensing fiber; the BP neural network algorithm is used to perform network training on the operating data; the network training data set is divided into a training set and a test set, the test set is used to detect the neural network training effect, and the training result of the BP neural network is used to perform real-time temperature compensation on the all-optical current transformer.

[0082] In this embodiment, the operation data of the all-fiber current transformer is obtained, and the output signal of the all-fiber current transformer can be transmitted to the host computer in real time through the serial port or optical port line in the acquisition unit. The sensing optical fiber can be divided into the optical fiber inside the optical fiber ring and the polarization-maintaining transmission optical fiber. In the actual application scenario of this embodiment, the optical fiber inside the optical fiber ring is mainly affected by the heating of the conductor and the ambient temperature. The conductor is generally arranged in the central part of the optical fiber ring. The ambient temperature in a small local area is approximately the same. Therefore, it can be approximately considered that the temperature of the optical fiber inside the optical fiber ring is the same, and the polarization-maintaining transmission optical fiber is located between the optical fiber ring and the acquisition unit, and is generally arranged in a vertical insulator. Affected by the heating of the conductor, the heating of the surrounding electrical equipment and the ambient temperature, the temperature of the polarization-maintaining transmission optical fiber is unevenly distributed. When measuring the temperature of the sensing optical fiber, multiple temperature sensors are used to measure the temperature of multiple points on the optical fiber inside the optical fiber ring and the polarization-maintaining transmission optical fiber, respectively, so that the optical fiber temperature state can be measured more accurately.

[0083] In this embodiment, multiple temperature sensors measure the temperature of multiple points of the optical fiber. One temperature sensor can be used to measure the temperature of the optical fiber inside the optical fiber ring, and two temperature sensors can be used to measure the temperature of the middle part of the polarization-maintaining transmission optical fiber and the temperature of the end of the polarization-maintaining transmission optical fiber (the part of the optical fiber connected to the acquisition unit), respectively. The influence of the uneven temperature distribution of the polarization-maintaining transmission optical fiber on the output signal of the all-optical current transformer is fully considered, and the temperature distribution curve of the sensing optical fiber is obtained based on the temperature measurement data.

[0084] like Figure 2 As shown, in this embodiment, a physical structure diagram of an all-fiber current transformer is provided, wherein the optical fiber 1 inside the optical fiber ring and the polarization-maintaining transmission optical fiber 5 are most affected by temperature, and the optical fiber 1 inside the optical fiber ring is protected by a shell. When the ambient temperature changes and the conductor heats up, it can be regarded that the temperature of the optical fiber 1 inside the optical fiber ring is uniformly distributed, and the temperature of the optical fiber 1 inside the optical fiber ring is higher than the ambient temperature. The polarization-maintaining transmission optical fiber 5 is placed in a columnar insulator. When the ambient temperature changes, the conductor heats up, and the surrounding electrical equipment heats up, if the temperature of the polarization-maintaining transmission optical fiber 5 is regarded as being consistent with the temperature of the optical fiber 1 inside the optical fiber ring, it will deviate from the actual temperature and further increase the output error of the all-fiber current transformer. Therefore, at this time, the temperature of the polarization-maintaining transmission optical fiber 5 is a law of gradually attenuating from the 1 / 4 wave plate 3, and finally the temperature of the polarization-maintaining transmission optical fiber at the bottom of the columnar insulator 4 is approximately the same as the ambient temperature. By measuring the temperature of multiple positions of the polarization-maintaining transmission optical fiber 5, the temperature distribution curve of the polarization-maintaining transmission optical fiber 5 can be fitted, so as to more accurately perform the temperature compensation of the all-fiber current transformer.

[0085] S2, processing the initial operation data of the all-optical current transformer and extracting sample data;

[0086] In this embodiment, due to the high sampling rate of the all-fiber current transformer, if the output signal is directly used as the input of the BP neural network, the number of samples will be too large, which greatly increases the operation time of the BP neural network. Therefore, it is necessary to downsample the operating data of the all-fiber current transformer so that the sampling rates of different parameters remain consistent, and use the processed measurement data as the input samples of the BP neural network.

[0087] In this embodiment, the data imported into the BP neural network is divided into a training set and a test set, wherein the output signal of the all-optical current transformer, the real-time temperature of the temperature sensor, the humidity around the all-optical current transformer, the surrounding wind speed, and the historical conductor current carrying capacity are used as the input of the neural network, and the all-optical current transformer ratio difference is used as the output. Multi-parameter input can greatly improve the stability and accuracy of the model. After the neural network training, the prediction result is used as the compensation amount, thereby realizing the temperature compensation of the all-optical current transformer based on the BP neural network algorithm, greatly improving the temperature stability of the all-optical current transformer.

[0088] S3. Establish a temperature compensation model of all-optical current transformer based on BP neural network algorithm.

[0089] like Figure 3 As shown, in this embodiment, step S3 of establishing a temperature compensation model of an all-optical current transformer based on a BP neural network algorithm further includes the following specific steps:

[0090] S31, importing the processed operating data as a sample data set into the BP neural network model;

[0091] S32, dividing the sample data into a training set and a test set, the number of samples in the training set: the number of samples in the test set = 3:1, and normalizing the sample data;

[0092] S33, determining the number of nodes in the input layer, the number of nodes in the hidden layer, and the number of nodes in the output layer of the BP neural network, setting training parameters, and initializing the weights and thresholds in the BP neural network;

[0093] In this embodiment, the number of input layer nodes, hidden layer nodes, and output layer nodes of the BP neural network are determined by the input sample set; the aforementioned training parameters include the number of training times, target error, and learning rate. The appropriateness of the parameters will directly affect the prediction results of the BP neural network. Therefore, it is necessary to continuously adjust the training parameter values ​​to achieve the expected training effect; the aforementioned BP neural network weight threshold is initialized by the algorithm itself.

[0094] S34, creating a BP neural network, and training it with the input sample data, by solving the expected weights and thresholds to meet the set target error, and obtaining the optimal prediction output result of the BP neural network;

[0095] In this embodiment, the target error can be set to, for example: 10 -6 , after achieving the expected training effect, the final training parameters are saved and the temperature compensation model of the all-optical current transformer is obtained.

[0096] S35, using the BP neural network prediction output result as a compensation parameter to compensate for the temperature error of the all-optical current transformer;

[0097] S36. Compare the predicted output result with the true value, and verify the temperature compensation effect of the BP neural network through calculation of relevant indicators.

[0098] In this embodiment, the related indicators generally include determination coefficient R, absolute error MAE, and root mean square error RMSE. The training effect of the BP neural network can be intuitively viewed by calculating the indicators.

[0099] The determination coefficient expression is:

[0100]

[0101] Among them, R is the coefficient of determination, y i is the true value, is the predicted value, is the average value, and the determination coefficient ranges from (0, 1). The closer the determination coefficient of the training result is to 1, the better the BP neural network training fitting effect is; on the contrary, the closer the determination coefficient of the training result is to 0, the worse the fitting effect is.

[0102] In this embodiment, the absolute error expression is:

[0103]

[0104] In the formula, MAE is the absolute error, m is the number of samples, and the absolute error is used to reflect the deviation of the predicted value from the true value.

[0105] The root mean square error expression is:

[0106]

[0107] Among them, RMSE is the root mean square error, which is used to reflect the degree to which the predicted value deviates from the true value.

[0108] The present embodiment proposes a temperature compensation method for an all-optical current transformer based on a BP neural network algorithm, which can effectively compensate for the temperature error of the all-optical current transformer. Figure 4The compensated all-fiber current transformer meets the accuracy requirement of 0.2 level or even higher, which further improves the long-term working stability of the all-fiber current transformer and has very good economic benefits.

[0109] In summary, the present invention performs temperature compensation on the all-fiber current transformer based on the BP neural network, and considers the influence of uneven temperature, thereby solving the problem of decreased measurement accuracy of the all-fiber current transformer when the ambient temperature changes. The present invention performs temperature compensation on the all-fiber current transformer according to the prediction results of the BP neural network, and verifies that the compensated all-fiber current transformer meets the 0.2 level accuracy requirement. This method has important reference significance for improving the long-term working stability of the all-fiber current transformer.

[0110] The present invention performs temperature compensation on an all-fiber current transformer based on a BP neural network and takes the influence of uneven temperature into consideration, thereby solving the problem of reduced measurement accuracy of the all-fiber current transformer when the ambient temperature changes.

[0111] Compared with the prior art, the present invention further improves the temperature compensation accuracy of the all-optical fiber current transformer, so that the measurement accuracy and stability of the all-optical fiber current transformer are further improved.

[0112] The present invention fully considers the influence of uneven temperature distribution of polarization-maintaining transmission optical fiber on the output signal of the all-optical fiber current transformer, and can obtain a more accurate temperature distribution curve of the sensing optical fiber according to the temperature measurement data.

[0113] The multi-parameter input adopted by the present invention can greatly improve the stability and accuracy of the model. After the neural network training, the prediction result is used as the compensation amount, thereby realizing the temperature compensation of the all-optical current transformer based on the BP neural network algorithm, and greatly improving the temperature stability of the all-optical current transformer. The allocated data set is normalized to improve the training efficiency and stability of the neural network.

[0114] The present invention solves the technical problem in the prior art that insufficient consideration is given to the influence of uneven temperature, resulting in a decrease in the measurement accuracy of the all-optical current transformer when the ambient temperature changes.

[0115] The above embodiments are only used to illustrate the technical solutions of the present invention, rather than to limit the same. Although the present invention has been described in detail with reference to the aforementioned embodiments, those skilled in the art should understand that the technical solutions described in the aforementioned embodiments may still be modified, or some of the technical features may be replaced by equivalents. However, these modifications or replacements do not deviate the essence of the corresponding technical solutions from the spirit and scope of the technical solutions of the embodiments of the present invention.

Claims

1. A temperature compensation method for all-optical current transformer based on BP network, characterized in that: The method comprises: S1. Using a collection unit to obtain the operating data of the all-optical current transformer in real time, in the collection unit, the operating data of the all-optical current transformer is transmitted to the host computer in real time through the serial port and the optical port line; S2, performing a downsampling operation on the operating data of the all-optical current transformer to obtain measurement data with a consistent sampling rate as an input sample of the BP neural network; S3, establishing a BP neural network model, training the BP neural network model, obtaining a temperature compensation model of the all-optical current transformer, performing model prediction based on the model, obtaining a suitable temperature compensation result, and performing temperature compensation on the all-optical current transformer.

2. The temperature compensation method for all-optical current transformer based on BP network according to claim 1, characterized in that: In S1, the operation data of the all-optical current transformer includes: the output signal of the all-optical current transformer, the real-time temperature of the temperature sensor, the humidity around the all-optical current transformer, the surrounding wind speed and the historical conductor current carrying capacity.

3. The temperature compensation method for all-optical current transformer based on BP network according to claim 2, characterized in that: The real-time temperature of the temperature sensor includes: fiber ring temperature, polarization-maintaining transmission fiber temperature and ambient temperature.

4. The temperature compensation method for all-optical current transformer based on BP network according to claim 1, characterized in that: In S1, when measuring the temperature of the transmission optical fiber, at least two temperature sensors are used to measure the temperature of multiple points on the optical fiber inside the optical fiber ring and the polarization-maintaining transmission optical fiber respectively; Using a humidity sensor to measure the humidity around the all-optical current transformer; Using an anemometer to measure the surrounding wind speed; The historical conductor current carrying capacity is obtained by processing the historical measurement data of the all-optical current transformer.

5. The temperature compensation method for all-optical current transformer based on BP network according to claim 1, characterized in that: The temperature sensor is used to measure the temperature of the optical fiber inside the optical fiber ring, and the remaining temperature sensors are used to measure the temperature in the middle of the polarization-maintaining transmission optical fiber and the temperature at the end of the polarization-maintaining transmission optical fiber, and the temperature distribution curve of the sensing optical fiber is obtained by fitting processing.

6. The temperature compensation method for all-optical current transformer based on BP network according to claim 1, characterized in that: In S2, the operation data of the all-optical current transformer is processed and cleaned, and the abnormal data point is replaced by the average value of the sampling point data before and after the abnormal data point to remove the abnormal data point, wherein the abnormal data point includes: mutation point and missing point; data downsampling is performed to unify the sampling rates of different parameters; for a set of data D={…,d1,d,d2,…}, when d is the abnormal data point, the following logic is used to express the new value d after data processing and cleaning: The data is processed and cleaned using the following logic: Unifying the sampling rates of the different parameters by performing the data downsampling; The sampling rate consistent measurement data is randomly divided into a training set and a test set, and the input samples are input into the BP neural network to obtain prediction results through training; The BP neural network is trained, and the prediction result is used as a compensation amount to perform temperature compensation on the all-optical current transformer.

7. The temperature compensation method for all-optical current transformer based on BP network according to claim 1, characterized in that: The S3 includes: S31, importing the input sample into the BP neural network model; S32, dividing the input samples into a training set and a test set, setting the number of samples in the training set, and performing normalization processing on the input samples; S33, setting model parameters and training parameters of the BP neural network model, and initializing weights and thresholds of the BP neural network model; S34, using the BP neural network model, performing model training according to the input samples, solving the weights and the thresholds until a preset target error is met, and obtaining an all-optical current transformer temperature compensation model and an applicable prediction output result of the BP neural network model. S35, using the applicable prediction output result as a compensation parameter to compensate for the temperature error of the all-optical current transformer; S36. Compare the applicable prediction output result with the true value and calculate the correlation index to verify the temperature compensation effect of the BP neural network model, wherein the correlation index includes: determination coefficient R, absolute error MAE and root mean square error RMSE.

8. The temperature compensation method for all-optical current transformer based on BP network according to claim 7, characterized in that: In S33, the model parameters include: the number of input layer nodes, the number of hidden layer nodes and the number of output layer nodes.

9. The temperature compensation method for all-optical current transformer based on BP network according to claim 7, characterized in that: In the S34, when the preset target error is met, the applicable training parameters are saved and the temperature compensation model of the all-optical current transformer is set accordingly.

10. The all-optical current transformer temperature compensation system based on BP network is characterized in that: The system comprises: An operation data acquisition module is used to acquire the operation data of the all-optical current transformer in real time by using a collection unit, and in the collection unit, the operation data of the all-optical current transformer is transmitted to the host computer in real time through the serial port and the optical port line; A sample extraction module, used for downsampling the operating data of the all-optical current transformer to obtain measurement data with consistent sampling rate as input samples of the BP neural network, and the sample extraction module is connected to the operating data acquisition module; The temperature supplement module is used to establish a BP neural network model, train the BP neural network model, obtain a temperature compensation model of the all-optical current transformer, perform model prediction based on it, and obtain a suitable temperature compensation result to perform temperature compensation on the all-optical current transformer. The temperature supplement module is connected to the sample extraction module.

Images