Intelligent method for data fault-tolerant mechanism in loop verification of secondary system of transformer substation

Abstract

The invention relates to the technical field of transformer substation verification, and discloses a data fault-tolerant mechanism intelligent method in transformer substation secondary system loop verification, and the method comprises the steps: training a Ridge Polynomial neural network through the primary loop data of a power supply end, so as to obtain an optimal weight and ideal error precision; secondly, a fault-tolerant mechanism is constructed, primary loop data transmitted by the wireless communication network and received by the handheld terminal is used as network input, a network actual output value is trained, and then error judgment is carried out; compared with a traditional neural network method, the method has the advantages that a high-order Ridge Polynomial neural network model is introduced, and high data processing and data analysis accuracy can be guaranteed due to super-high nonlinear processing capacity of a network structure; compared with an existing transformer substation secondary system loop verification method, the method has the advantages that a data fault-tolerant mechanism is established through the Ridge Polynomial neural network, the problem of false alarm or missing report in the data transmission process of a wireless communication network is avoided, and the quality of primary loop data transmitted to a handheld end is guaranteed.

Description

Technical Field

[0001] This invention relates to the field of substation verification technology, specifically to an intelligent method for data fault tolerance mechanism in substation secondary system circuit verification. Background Technology

[0002] Existing substation secondary system circuit verification, such as Figure 1 As shown: Primary circuit data such as current and voltage from the power supply are wirelessly transmitted to the handheld device. The handheld device can measure the corresponding secondary circuit data such as current and voltage. Based on the amplitude and phase values ​​transmitted from the power supply, and combined with its own measurements, conversion formulas, and diagnostic algorithms, the handheld device verifies wiring faults such as wiring, turns ratio, polarity, and open circuits in the secondary AC system. By measuring the amplitude and phase of the primary and secondary currents, the correctness of the wiring in each circuit is determined based on the phase and amplitude of the primary and secondary currents. However, the primary circuit data is transmitted to the handheld device via a wireless communication network and compared with the secondary circuit data collected by the handheld device to determine whether the wiring is normal. During this process, noise interference may cause false alarms in the primary circuit data transmitted to the handheld device; signal instability may cause missed alarms, making it difficult to guarantee the data quality of the primary circuit. False alarms or missed alarms will directly affect the final judgment result. Summary of the Invention

[0003] To address the shortcomings of existing technologies, this invention provides an intelligent method for data fault tolerance mechanism in substation secondary system circuit verification, which has advantages such as avoiding false alarms and solves the aforementioned technical problems.

[0004] To achieve the above objectives, the present invention provides the following technical solution: an intelligent method for data fault tolerance mechanism in substation secondary system circuit verification, comprising the following steps: S1: Use primary circuit data from the power supply to train a Ridge Polynomial neural network; S2: A pre-trained Ridge Polynomial neural network is used to construct a data fault tolerance mechanism. Specifically, the primary loop data received by the handheld terminal from the wireless communication network is used as the input of the Ridge Polynomial neural network to obtain the actual output value. S3: Perform error judgment: When the error between the actual output value and the ideal output value is greater than the ideal error accuracy, the received primary loop data has false alarms or omissions, and is retransmitted; When the error between the actual output value and the ideal output value is less than or equal to the ideal error accuracy, there are no false alarms or omissions in the received primary loop data; S4: carry out the first, second current phase, amplitude comparison, to determine the correctness of each loop wiring.

[0005] As a preferred technical scheme of the present application, the Ridge Polynomial neural network comprises an input layer, a hidden layer and an output layer, wherein the hidden layer comprises k PSNN nodes, and the output of the hidden layer is: wherein each PSNN node is composed of a Pi-Sigma neural network, the output of the i-th PSNN node is , wherein i represents the i-th PSNN node, and k represents the total number of PSNN nodes.

[0006] As a preferred technical scheme of the present application, the Ridge Polynomial neural network model is trained by using a gradient algorithm, and when the m-th iteration is performed in the training process, the weight update is: wherein m represents the iteration number, represents the weight between the i-th node of the input layer and the j-th node of the hidden layer at the m-th iteration, represents the weight gradient between the i-th node of the input layer and the j-th node of the hidden layer at the m-th iteration. Compared with the prior art, the present application provides an intelligent data fault-tolerant mechanism method in substation secondary system loop verification, which has the following beneficial effects: Compared with the traditional neural network method (only containing summation neurons

[0007] ), the present application introduces a high-order Ridge Polynomial neural network model (containing summation neurons and product neurons ), which has super nonlinear processing capacity and can ensure high data processing and data analysis accuracy; compared with the existing substation secondary system loop verification method, the Ridge Polynomial neural network is used to establish a data fault-tolerant mechanism, which can avoid false positives or false negatives in the process of transmitting data by a wireless communication network, and ensure the quality of the primary loop data transmitted to the handheld terminal. BRIEF DESCRIPTION OF DRAWINGS

[0008] Figure 1 ​​​​​​​It is a schematic diagram of the existing substation secondary system loop verification method; Figure 2 It is a schematic diagram of Pi-Sigma neural network structure; Figure 3 It is a schematic diagram of Ridge Polynomial neural network structure; Figure 4 It is a schematic diagram of the process of the present application. DETAILED DESCRIPTION

[0009] The technical solutions in the embodiments of the present application will be clearly and completely described below with reference to the drawings in the embodiments of the present application. Obviously, the described embodiments are only a 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 work fall within the protection scope of the present application.

[0010] Please refer to Figures 1-4 The intelligent method of data fault-tolerant mechanism in substation secondary system loop verification comprises the following steps: S1: training Ridge Polynomial neural network by using primary loop data of power supply end; S2: constructing data fault-tolerant mechanism by using the trained Ridge Polynomial neural network, specifically, taking the primary loop data transmitted by the received wireless communication network of handheld end as the input of Ridge Polynomial neural network, and obtaining actual output value; S3: error judgment: When the error between the actual output value and the ideal output value is greater than the ideal error precision, the received primary loop data has false or missing report, and is retransmitted; When the error between the actual output value and the ideal output value is less than or equal to the ideal error precision, the received primary loop data has no false or missing report; S4: comparing primary and secondary current phase, amplitude, etc. to determine the correctness of each loop wiring; Firstly, the Ridge Polynomial neural network is trained by the primary loop data of the power supply end to obtain the optimal weight and ideal error precision. Secondly, the trained Ridge Polynomial neural network is used to construct the data fault-tolerant mechanism. The primary loop data received by the handheld end from the wireless communication network is taken as the network input, and the actual output value of the network is trained, and then the error is judged. When the error between the actual output value and the ideal output value is greater than the ideal error precision, the received primary loop data has the false or missing report condition, and is retransmitted. When the error between the actual output value and the ideal output value is less than or equal to the ideal error precision, the received primary loop data does not have the false or missing report condition. Thirdly, the primary and secondary current phases, amplitudes and the like are compared to determine the correctness of the loop wiring.

[0011] Training the Ridge Polynomial neural network: The primary loop data directly collected from the power supply end (without being transmitted through the wireless communication network) is defined as data x , wherein, n is the number of elements contained in data x , x i is the i-th element in data x . i

[0012] The traditional neural network structure only contains summation neurons ( ), and has poor nonlinear ability. However, in actual life, many problems to be solved are nonlinear, which greatly limits the application of the network. In order to overcome the shortcomings of this model, scholars put forward another network structure, that is, a high-order neural network containing summation neurons ( ) and product neurons ( ), which has strong nonlinear ability and is more suitable for actual problems.

[0013] The Ridge Polynomial neural network (RPNN) is a high-order neural network formed by combining the Pi-sigma neural network (PSNN). It inherits the fast learning ability of the Pi-Sigma neural network, can approximate any continuous function, and has been researched and applied in a certain range. The number of neuron nodes of each layer of the Pi-Sigma neural network is N, K and 1 respectively, the weight value between the hidden layer and the output layer is 1, the input sample is , represents the threshold value, represents the weight value between the k-th node of the input layer and the j-th node of the hidden layer,​ , denote the weights of the connections to the threshold nodes. Then The output of the unit is: (1) The final output of the PSNN is: (2) where is an activation function; Each node in the hidden layer of the Ridge Polynomial neural network is a PSNN. For each , the subscript indicates that it contains i hidden nodes. Assume that the number of input layer nodes of the RPNN is d, the number of hidden layer nodes is k, and the number of output layer nodes is 1. Denote as the weight between the kth hidden layer node in the ith PSNN and the jth input node, denote the output of each , then the output of the RPNN is: (3) where g is an activation function .

[0014] The gradient algorithm is used to train the Ridge Polynomial neural network model. Let the data directly collected at the source end (without being transmitted through the wireless communication network) be denoted as a sample set , and label it. By default, all O a = 1, i.e., all data is directly collected at the power supply end without being transmitted through the wireless communication network, and there is no false negative or false positive situation.

[0015] The error function of the RPNN output is: (4) Denote (5)From the above two equations (4) and (5), we have (6)The ultimate goal of using the gradient algorithm to train the network is to find a such that , Take the gradient of : (7) During the training process, when the mth iteration is performed, the weight update is: (8) (8) weight update variable is: (9) (10) is the learning rate.

[0016] By training, the optimal weight and the ideal error precision E * .

[0017] Constructing data fault-tolerant intelligent mechanism: Fix the weight of the Ridge Polynomial neural network , and the received wireless communication network transmitted loop data is X 1, which is brought into equation (3) to obtain its output value Y 1. Through equation (6), the corresponding error is E 1. Then make error judgment: When E 1> E * , X 1 exists false alarm or false alarm situation, retransmit; When E 1<= E * , X 1 exists false alarm or false alarm situation, can be compared.

[0018] Data comparison: Compare the first and second current phase, amplitude, etc. to determine the correctness of each loop wiring.

[0019] Although embodiments of the present application have been shown and described, it will be understood by those having ordinary skill in the art that various changes, modifications, substitutions and alterations can be made therein without departing from the principles and spirit of the present application, the scope of which is defined by the appended claims and their equivalents.

Claims

1. An intelligent method of data fault tolerance mechanism in substation secondary system loop verification, characterized in that: The method comprises the following steps: S1: training a Ridge Polynomial neural network by using primary loop data of a power supply end; S2: constructing a data fault-tolerant mechanism by using the trained Ridge Polynomial neural network, specifically, taking the primary loop data transmitted by a wireless communication network and received by a handheld end as the input of the Ridge Polynomial neural network, and obtaining an actual output value; S3: performing error judgment: When the error between the actual output value and an ideal output value is greater than an ideal error precision, the received primary loop data has a false alarm or a missed alarm, and is retransmitted; When the error between the actual output value and the ideal output value is less than or equal to the ideal error precision, the received primary loop data has no false alarm or missed alarm; S4: performing comparison of primary and secondary current phases, amplitudes and the like to determine the correctness of each loop wiring.

2. The intelligent method for data fault-tolerant mechanism in substation secondary system loop verification according to claim 1, characterized in that: The Ridge Polynomial neural network comprises an input layer, a hidden layer and an output layer, wherein the hidden layer comprises k PSNN nodes, and the output of the hidden layer is: where each PSNN node is composed of a Pi-Sigma neural network, the output of the i-th PSNN node is given by , where denotes the i-th PSNN node, k denotes the total number of PSNN nodes, g is an activation function, and g= , e denotes the natural constant.

3. The intelligent method for data fault-tolerance mechanism in substation secondary system loop verification according to claim 1, characterized in that: The Ridge Polynomial neural network model is trained by using a gradient algorithm, and when the mth iteration is performed in the training process, the weight value is updated as: wherein, denotes the iteration number, denotes the weight value between the th node of the input layer and the th node of the hidden layer at the th iteration, denotes the gradient of the weight value between the th node of the input layer and the th node of the hidden layer at the th iteration.

Images