A method and system for predicting optical fiber flow transmission rate for power stations
By decomposing and processing the fiber transmission data and linear superposition, and filtering historical data with evaluation indicators, the problems of low accuracy and insufficient robustness of fiber transmission rate prediction are solved, and efficient flow control and network optimization are achieved.
Patent Information
- Application Number
- CN202111528919.4
- Authority / Receiving Office
- CN · China
- Patent Type
- Patents(China)
- Current Assignee / Owner
- Filing Date
- 2021-12-14
- Publication Date
- 2024-11-26
- Estimated Expiration
- 2041-12-14
AI Technical Summary
The existing fiber transmission rate prediction has problems such as low prediction accuracy, high volatility, long time, poor prediction efficiency and robustness.
By decomposing the original data, the influence of modal aliasing is reduced, the decomposed subcomponents are predicted separately, and linearly superimposed by weight coefficients is established to filter historical data, and the prediction accuracy and robustness are improved.
It significantly improves the prediction accuracy and robust performance of fiber traffic transmission rates, providing more accurate flow control and network optimization support.
Smart Images

Figure CN114186745B_ABST
Abstract
Description
Technical Field
[0001] The present invention relates to the field of power station communication technology, and in particular to a method and system for predicting optical fiber flow transmission rate for power stations. Background Art
[0002] The advent of the information age has made the network penetrate into all aspects of social life. From the initial 2G network to the current 5G era, the speed of the network has increasingly affected people's surfing experience. Optical fiber has become the main transmission method for current data communication and information networks because its transmission speed is much higher than other transmission media. However, due to the various intrusion behaviors on the network, optical fiber transmission is not safe. The security problems of optical fiber networks are becoming increasingly serious due to the strong network aggression and low awareness of prevention. By predicting the flow rate in optical fiber transmission, real-time monitoring of optical fiber transmission is completed, and real and effective data is provided for flow control, fault location, routing basis and bandwidth allocation in optical fiber networks. At the same time, it also provides technical support for operators to regulate optical fiber traffic. By predicting network traffic, problems such as network overload can be effectively prevented, countermeasures can be taken in advance, the stable operation of the network can be ensured, and network congestion can be avoided. In the existing technology, the prediction of optical fiber transmission rate has problems such as low prediction accuracy, large prediction volatility, long time consumption, low prediction efficiency and robustness. Therefore, further research is needed on data processing and prediction of optical fiber transmission rate. Summary of the invention
[0003] To solve the above problems, the present invention aims to propose a method and system for predicting the optical fiber flow transmission rate for power stations. By decomposing the original data, the influence of modal aliasing on the prediction result is reduced, and then the decomposed sub-components are predicted separately, so that the prediction effect is significantly improved. Finally, the prediction result is obtained by superposition, which has high prediction accuracy and better robust performance.
[0004] To achieve the above object, the technical solution of the present invention is achieved as follows:
[0005] A method for predicting optical fiber flow transmission rate for a power station comprises the following steps:
[0006] S1, sending marked original information data at the beginning of the optical fiber within a unit time, and receiving the original information data sequence at the other end of the optical fiber;
[0007] S2, decomposing the original information data sequence to obtain multiple groups of data sequences with different characteristics, and obtaining weight coefficients of each group of data by calculation;
[0008] S3. Predict the multiple groups of data sequences with different characteristics; linearly superimpose the prediction results obtained from the multiple groups of data sequences with different characteristics according to the calculated weight coefficients, and use the superimposed result as the prediction result of the traffic transmission rate at the next moment.
[0009] Furthermore, it also includes step S4, establishing an evaluation index system, evaluating and analyzing the prediction result values obtained from predictions in different time periods, and saving the information data that meets the evaluation index range requirements as historical information data, which is used as the historical value for predicting the traffic transmission rate at later moments; the information data that does not meet the requirements is discarded, that is, the historical data of the previous moment is used to predict the traffic transmission rate value at this moment.
[0010] Furthermore, in step S2, multiple groups of data sequences with different characteristics are first decomposed into multiple data sequences with simple characteristics, and the method is as follows: first find all the maximum points and minimum points in the original data Y(t) or the signal, connect all the maximum points and minimum points respectively, obtain the upper and lower limits and range of the data Y(t), and calculate the mean n(t) of the upper and lower limits; let g 1 (t) = Y(t) - n(t), check g 1 (t) Whether the following two conditions are met:
[0011] (1) The number of extreme points (including maximum and minimum points) in the signal is equal to or differs by 1 from the number of zero-crossing points;
[0012] (2) At any point in time, the mean of the upper and lower limits determined by the local extreme value of the signal is zero;
[0013] If the above conditions are not met, repeat the above steps. If they are met, record f 1 (t) = g 1 (t), as the first decomposition sequence; repeat the above steps until the data sequence is a monotonic function or a constant value, and obtain a linear data sequence.
[0014] Furthermore, in step S2, the signal Y(t) of the original data is expressed as
[0015] Y(t)=f 1 (t)+f 2 (t)+f 3 (t)+…+f N (t)+i(t) (1)
[0016] Where t is the number of elements in the time series, N is the number of decomposed components, and i(t) is the component remaining after the original signal cannot be decomposed, i.e., the remainder;
[0017] Through the above n sets of original data input variables, that is, N sets of different sample training data x j , j = 1, ..., N, to obtain the setting function g(a i , b i , x j ), i = 1, 2, ..., T, thereby obtaining the optimized output matrix function Z(t) based on the sample data; the model expression is as follows:
[0018] Z=Hβ
[0019]
[0020] Where H is the hidden layer output matrix; Z is the output matrix; β is the weight matrix between the hidden layer and the output layer; a i is the matrix of corresponding weights between all nodes in the input layer and the i-th node in the hidden layer; b i is the threshold of the i-th hidden layer node;
[0021] In the matrix H, calculate the optimal weights for
[0022]
[0023] In the formula, H + is the generalized inverse matrix of the matrix H.
[0024] Furthermore, in step S3, the following two formulas are used to predict the positions of the multiple groups of data sequences with different characteristics obtained by decomposition within the value range, where A and C are random weight coefficients:
[0025] D=|CX rand (t)-X(t)| (4)
[0026] X(t+1)=X rand (t)-AD (5)
[0027] Where D is the distance from the data value at this moment to the target optimization value; X(t) and X(t+1) are the current position and the position after the next iteration respectively; t is the current iteration time; X rand It is a random value generated by a random function within the value range;
[0028] The following two formulas are used to search for the weight coefficient of each group of data and the optimal value of the data at this moment:
[0029] D=|CX * (t)-X(t)| (6)
[0030] X * (t+1)=X *(t)-AD (7)
[0031] Where, X * (T) and X * (t+1) are the best data value at the current moment and the best data value after the next iteration moment respectively.
[0032] Furthermore, in step S3, the obtained multiple sets of data sequence prediction results with different characteristics are linearly superimposed to obtain the prediction result of the traffic transmission rate at the next moment; the predicted data signal Y*(t) is expressed as:
[0033] Y*(t)=X * 1 (t)+X * 2 (t)+X * 3 (t)+…+X * N (t) (8).
[0034] Further, in the step S4, the evaluation index system is: using the error analysis index to analyze the prediction results, selecting the root mean square error RMSE and the fitting coefficient R2 to analyze the quality of the prediction results; here, the root mean square error RMSE cut-off value is set to 2, and the fitting coefficient R2 value is set to 1; information data that meets the evaluation index range requirements, that is, the root mean square error RMSE value does not exceed 2, and the fitting coefficient R2 value does not exceed 1, is saved as historical information data; information data that does not meet the requirements is discarded;
[0035]
[0036]
[0037] In the formula, n represents the number of prediction samples, p i Indicates the predicted fiber transmission rate value, q i Indicates the actual fiber transmission rate value. Indicates the average optical fiber transmission rate.
[0038] In order to achieve the above object, the present invention also provides a power station optical fiber flow transmission rate prediction system, comprising the following modules:
[0039] Data sending and receiving module: used to send marked original information data at the beginning of the optical fiber within a unit time, and receive the original information data sequence at the other end of the optical fiber;
[0040] Data processing module: used to decompose the original information data sequence to obtain multiple groups of data sequences with different characteristics, and obtain the weight coefficient of each group of data by calculation;
[0041] Prediction module: used to predict the multiple groups of data sequences with different characteristics; linearly superimpose the prediction results obtained from the multiple groups of data sequences with different characteristics according to the calculated weight coefficients, and use the superimposed results as the prediction results of the traffic transmission rate at the next moment;
[0042] Calibration module: used to establish a good evaluation index system, evaluate and analyze the prediction results obtained from predictions in different time periods, and save the information data that meets the evaluation index range requirements as historical information data, which is used as the historical value for predicting the traffic transmission rate at later moments; information data that does not meet the requirements is discarded, that is, the historical data of the previous moment is used to predict the traffic transmission rate value at this moment.
[0043] 10. Furthermore, in the data processing module, firstly, multiple groups of data sequences with different characteristics are decomposed into multiple data sequences with simple characteristics, and the method is as follows: firstly, all the maximum points and minimum points in the original data Y(t) or the signal are found, and all the maximum points and minimum points are connected respectively to obtain the upper and lower limits and range of the data Y(t), and calculate the mean n(t) of the upper and lower limits; let g 1 (t) = Y(t) - n(t), check g 1 (t) Whether the following two conditions are met:
[0044] (1) The number of extreme points (including maximum and minimum points) in the signal is equal to or differs by 1 from the number of zero-crossing points;
[0045] (2) At any point in time, the mean of the upper and lower limits determined by the local extreme value of the signal is zero;
[0046] If the above conditions are not met, repeat the above steps. If they are met, record f 1 (t) = g 1 (t), as the first decomposition sequence; repeat the above steps until the data sequence is a monotonic function or a constant value, and obtain a linear data sequence;
[0047] In the data processing module, the original data signal Y(t) is expressed as
[0048] Y(t)=f 1 (t)+f 2 (t)+f 3 (t)+…+f N (t)+i(t) (1)
[0049] Where t is the number of elements in the time series, N is the number of decomposed components, and i(t) is the component remaining after the original signal cannot be decomposed, i.e., the remainder;
[0050] Through the above n sets of original data input variables, that is, N sets of different sample training data x j , j = 1, ..., N, to obtain the setting function g(a i , b i , x j ), i = 1, 2, ..., T, thereby obtaining the optimized output matrix function Z(t) based on the sample data; the model expression is as follows:
[0051] Z=Hβ
[0052]
[0053] Where H is the hidden layer output matrix; Z is the output matrix; β is the weight matrix between the hidden layer and the output layer; a i is the matrix of corresponding weights between all nodes in the input layer and the i-th node in the hidden layer; b i is the threshold of the i-th hidden layer node;
[0054] In the matrix H, calculate the optimal weights for
[0055]
[0056] In the formula, H + is the generalized inverse matrix of the matrix H.
[0057] Furthermore, in the prediction module, the following two formulas are used to predict the positions of the multiple groups of data sequences with different characteristics obtained by decomposition within the value range, where A and C are random weight coefficients:
[0058] D=|CX rand (t)-X(t)| (4)
[0059] X(t+1)=X rand (t)-AD (5)
[0060] Where D is the distance from the data value at this moment to the target optimization value; X(t) and X(t+1) are the current position and the position after the next iteration respectively; t is the current iteration time; X rand It is a random value generated by a random function within the value range;
[0061] The following two formulas are used to search for the weight coefficient of each group of data and the optimal value of the data at this moment:
[0062] D=|CX * (t)-X(t)| (6)
[0063] X * (t+1)=X *(t)-AD (7)
[0064] Where, X * (t) and X * (t+1) are the best data value at the current moment and the best data value after the next iteration moment respectively;
[0065] In the prediction module, multiple sets of data sequence prediction results with different characteristics are linearly superimposed to obtain the prediction result of the traffic transmission rate at the next moment; the prediction data signal Y*(t) is expressed as:
[0066] Y*(t)=X * 1 (t)+X * 2 (t)+X * 3 (t)+…+X * N (t) (8);
[0067] In the calibration module, the evaluation index system is as follows: the prediction results are analyzed using the error analysis index, and the root mean square error RMSE and the fitting coefficient R2 are selected to analyze the quality of the prediction results; here, the root mean square error RMSE cut-off value is set to 2, and the fitting coefficient R2 value is set to 1; the information data that meets the evaluation index range requirements, that is, the root mean square error RMSE value does not exceed 2, and the fitting coefficient R2 value does not exceed 1, is saved as historical information data; the information data that does not meet the requirements is discarded;
[0068]
[0069]
[0070] In the formula, n represents the number of prediction samples, p i Indicates the predicted fiber transmission rate value, q i Indicates the actual fiber transmission rate value. Indicates the average optical fiber transmission rate.
[0071] Beneficial effects: The present invention reduces the influence of modal aliasing on the prediction results by decomposing the original data, and then predicts each sub-component after decomposition, so that the prediction effect is significantly improved. Finally, the prediction result is obtained by superposition, which has high prediction accuracy and better robust performance. The optical fiber flow transmission rate prediction result of the present invention can be used for optical fiber communication operation optimization and optical fiber installation planning. BRIEF DESCRIPTION OF THE DRAWINGS
[0072] The accompanying drawings constituting a part of the present invention are used to provide a further understanding of the present invention. The exemplary embodiments of the present invention and their descriptions are used to explain the present invention and do not constitute an improper limitation of the present invention. In the accompanying drawings:
[0073] Figure 1 It is a data decomposition flow chart of the prediction method of the present invention;
[0074] Figure 2 The value optimization structure diagram of the prediction method of the present invention;
[0075] Figure 3 The value optimization flow chart of the prediction method of the present invention;
[0076] Figure 4 It is a schematic diagram of the decomposition of a certain original optical fiber transmission rate data in the prediction method of the present invention;
[0077] Figure 5 The optical fiber transmission rate prediction and its comparison diagram obtained by applying the prediction method of the present invention;
[0078] Figure 6 The absolute error of the optical fiber transmission rate prediction obtained by applying the prediction method of the present invention and its comparison diagram;
[0079] Figure 7 It is a structural schematic diagram of the optical fiber flow transmission rate prediction system for power stations of the present invention. DETAILED DESCRIPTION
[0080] It should be noted that, in the absence of conflict, the embodiments of the present invention and the features in the embodiments may be combined with each other.
[0081] The present invention will be described in detail below with reference to the accompanying drawings and in conjunction with embodiments.
[0082] Example 1
[0083] See also Figure 1-3 :A method for predicting the optical fiber flow transmission rate for a power station, comprising the following steps:
[0084] S1, sending marked original information data at the beginning of the optical fiber within a unit time, and receiving the original information data sequence at the other end of the optical fiber;
[0085] S2, decomposing the original information data sequence to obtain multiple groups of data sequences with different characteristics, and obtaining weight coefficients of each group of data by calculation;
[0086] In step S2, firstly, multiple groups of data sequences with different characteristics are decomposed into multiple data sequences with simple characteristics, and the method is as follows: firstly, all the maximum points and minimum points in the original data Y(t) or the signal are found, and all the maximum points and minimum points are connected respectively to obtain the upper and lower limits and range of the data Y(t), and the mean value n(t) of the upper and lower limits is calculated; let g 1 (t) = Y(t) - n(t), check g 1 (t) Whether the following two conditions are met:
[0087] (1) The number of extreme points (including maximum and minimum points) in the signal is equal to or differs by 1 from the number of zero-crossing points;
[0088] (2) At any point in time, the mean of the upper and lower limits determined by the local extreme value of the signal is zero;
[0089] If the above conditions are not met, repeat the above steps. If they are met, record f 1 (t) = g 1 (t), as the first decomposition sequence; repeat the above steps until the data sequence is a monotonic function or a constant value, and obtain a linear data sequence.
[0090] In step S2, the original data signal Y(t) is expressed as
[0091] Y(t)=f 1 (t)+f 2 (t)+f 3 (t)+…+f N (t)+i(t) (1)
[0092] Where t is the number of elements in the time series, N is the number of decomposed components, and i(t) is the component remaining after the original signal cannot be decomposed, i.e., the remainder;
[0093] Through the above n sets of original data input variables, that is, N sets of different sample training data x j , j = 1, ..., N, to obtain the setting function g(a i , b i , x j ), i = 1, 2, ..., T, thereby obtaining the optimized output matrix function Z(t) based on the sample data; the model expression is as follows:
[0094] Z=Hβ
[0095]
[0096] Where H is the hidden layer output matrix; Z is the output matrix; β is the weight matrix between the hidden layer and the output layer; a iis the matrix of corresponding weights between all nodes in the input layer and the i-th node in the hidden layer; b i is the threshold of the i-th hidden layer node;
[0097] In the matrix H, calculate the optimal weights for
[0098]
[0099] In the formula, H + is the generalized inverse matrix of the matrix H.
[0100] S3, predicting each group of decomposed data components; linearly superimposing the prediction results obtained from each group of data components according to the calculated weight coefficients, and using the superimposed results as the prediction results of the traffic transmission rate at the next moment;
[0101] In step S3, the following two formulas are used to predict the positions of the multiple groups of data sequences with different characteristics obtained by decomposition within the value range, where A and C are random weight coefficients:
[0102] D=|CX rand (t)-X(t)| (4)
[0103] X(t+1)=X rand (t)-AD (5)
[0104] Where D is the distance from the data value at this moment to the target optimization value; X(t) and X(t+1) are the current position and the position after the next iteration respectively; t is the current iteration time; X rand It is a random value generated by a random function within the value range;
[0105] The following two formulas are used to search for the weight coefficient of each group of data and the optimal value of the data at this moment:
[0106] D=|CX * (t)-X(t)| (6)
[0107] X * (t+1)=X * (t)-AD (7)
[0108] Where, X * (t) and X * (t+1) are the best data value at the current moment and the best data value after the next iteration moment respectively;
[0109] In step S3, multiple sets of data sequence prediction results with different characteristics are linearly superimposed to obtain the prediction result of the traffic transmission rate at the next moment; the predicted data signal Y*(t) is expressed as:
[0110] Y*(t)=X * 1 (t)+X * 2 (t)+X * 3 (t)+…+X * N (t) (8).
[0111] S4. Establish an evaluation index system, evaluate and analyze the prediction results obtained in different time periods, and save the information data that meets the evaluation index range requirements as historical information data, which will be used as the historical value for predicting the traffic transmission rate at the next moment; the information data that does not meet the requirements will be discarded, that is, the historical data at the previous moment will be used to predict the traffic transmission rate value at the current moment;
[0112] In step 4, the evaluation index system is: using the error analysis index to analyze the prediction results, selecting the root mean square error RMSE and the fitting coefficient R2 to analyze the quality of the prediction results; here, the root mean square error RMSE cut-off value is set to 2, and the fitting coefficient R2 value is set to 1; information data that meets the evaluation index range requirements, that is, the root mean square error RMSE value does not exceed 2, and the fitting coefficient R2 value does not exceed 1, is saved as historical information data; information data that does not meet the requirements is discarded;
[0113]
[0114]
[0115] In the formula, n represents the number of prediction samples, p i Indicates the predicted fiber transmission rate value, q i Indicates the actual fiber transmission rate value. Indicates the average optical fiber transmission rate.
[0116] This embodiment reduces the influence of modal aliasing on the prediction results by decomposing the original data, and then predicts each sub-component after decomposition, so that the prediction effect is significantly improved. Finally, the prediction result is obtained by superposition, which has high prediction accuracy and better robust performance. The optical fiber flow transmission rate prediction result of this embodiment can be used for optical fiber communication operation optimization and optical fiber installation planning.
[0117] The advantages of this embodiment are further illustrated below through specific application comparison implementation:
[0118] Based on the historical data of optical fiber transmission of a power station, the EMD algorithm is used to decompose the historical data to obtain the decomposed linear component and nonlinear component. The specific decomposition diagram is as follows Figure 4 shown.
[0119] The original data of the optical fiber transmission rate of a power station in May were selected as training samples. The data interval was twenty minutes, and the data in each hour was averaged. There were 24 data in one day. The optical fiber flow transmission rate of the power station was predicted by MATLAB simulation software. The data interval of the prediction result was the same as the collection data interval. The prediction results were evaluated and analyzed by the root mean square error and the fitting degree coefficient, and then compared with other methods.
[0120] The data of June was selected as the training set, and one day was randomly selected in July as the test set. The optical fiber traffic transmission rate was predicted by MATLAB simulation software, and the prediction results of the traffic transmission rate were obtained. The error index selected in this paper was used as the standard to compare the method proposed in this paper with the traditional ELM prediction model and WOA-ELM prediction model. It can be seen that the EMD-WOA-ELM prediction model proposed in this paper has higher accuracy, better effect and better robust performance. Figure 6 The temperature prediction results of different prediction methods are shown in Figure 2. Figure 6 is the prediction error of different methods. The following table 1 shows the RMSE and R of different prediction methods. 2 Numeric value.
[0121] Table 1 Prediction error indicators of each model under different weather types
[0122]
[0123] According to Table 1, Figure 5 as well as Figure 6 The prediction results of each model are taken as an example to illustrate. Figure 6 , Figure 7 From the analysis of the evaluation index values of each model in Table 1, it can be seen that the prediction effect of EMD-WOA-ELM is the best among the three prediction methods, followed by the prediction effect of WOA-ELM prediction method, and the prediction effect of the traditional ELM prediction method is the worst. The prediction effect of the ELM prediction model improved by the whale optimization algorithm has been improved to a certain extent, and its RMSE and R2 values are 1.4527 and 0.9831 respectively. The prediction result obtained by the traditional ELM prediction model is the worst among the three models, and its RMSE and R2 in the evaluation index are 1.8946 and 0.9547. The prediction model of EMD-WOA-ELM proposed in this paper first processes the original data through the EMD algorithm to reduce the influence of modal aliasing on the prediction results. After decomposition, each sub-component is predicted separately, which significantly improves the prediction effect. Finally, the RMSE and R2 values of the superimposed prediction results are 0.9783 and 0.9927 respectively.
[0124] Through a detailed analysis of the evaluation indicators of the three prediction models, Figure 5 and Figure 6 It can be seen from the simulation diagram that the prediction effect of the present invention is the best compared with the other two methods. The prediction degree of the optical fiber flow transmission rate is the most accurate and the anti-interference ability is also the best.
[0125] Example 2
[0126] To achieve the above purpose, see Figure 7 :This embodiment also provides a power station optical fiber flow transmission rate prediction system, including the following modules:
[0127] Data sending and receiving module: used to send marked original information data at the beginning of the optical fiber within a unit time, and receive the original information data sequence at the other end of the optical fiber;
[0128] Data processing module: used to decompose the original information data sequence to obtain multiple groups of data sequences with different characteristics, and obtain the weight coefficient of each group of data by calculation;
[0129] In the data processing module, multiple groups of data sequences with different characteristics are first decomposed into multiple data sequences with simple characteristics. The method is as follows: first find all the maximum and minimum points in the original data Y(t) or the signal, connect all the maximum and minimum points respectively, obtain the upper and lower limits and range of the data Y(t), and calculate the mean n(t) of the upper and lower limits; let g 1 (t) = Y(t) - n(t), check g 1 (t) Whether the following two conditions are met:
[0130] (1) The number of extreme points (including maximum and minimum points) in the signal is equal to or differs by 1 from the number of zero-crossing points;
[0131] (2) At any point in time, the mean of the upper and lower limits determined by the local extreme value of the signal is zero;
[0132] If the above conditions are not met, repeat the above steps. If they are met, record f 1 (t) = g 1 (t), as the first decomposition sequence; repeat the above steps until the data sequence is a monotonic function or a constant value, and obtain a linear data sequence;
[0133] In the data processing module, the original data signal Y(t) is expressed as
[0134] Y(t)=f 1 (t)+f 2 (t)+f 3 (t)+…+f N (t)+i(t) (1)
[0135] Where t is the number of elements in the time series, N is the number of decomposed components, and i(t) is the component remaining after the original signal cannot be decomposed, i.e., the remainder;
[0136] Through the above n sets of original data input variables, that is, N sets of different sample training data x j , j = 1, ..., N, to obtain the setting function g(a i , b i , x j ), i = 1, 2, ..., T, thereby obtaining the optimized output matrix function Z(t) based on the sample data; the model expression is as follows:
[0137] Z=Hβ
[0138]
[0139] Where H is the hidden layer output matrix; Z is the output matrix; β is the weight matrix between the hidden layer and the output layer; a i is the matrix of corresponding weights between all nodes in the input layer and the i-th node in the hidden layer; b i is the threshold of the i-th hidden layer node;
[0140] In the matrix H, calculate the optimal weights for
[0141]
[0142] In the formula, H + is the generalized inverse matrix of the matrix H.
[0143] Prediction module: used to predict the multiple groups of data sequences with different characteristics; linearly superimpose the prediction results obtained from the multiple groups of data sequences with different characteristics according to the calculated weight coefficients, and use the superimposed results as the prediction results of the traffic transmission rate at the next moment;
[0144] In the prediction module, the following two formulas are used to predict the positions of multiple groups of data sequences with different characteristics obtained by decomposition within the value range, where A and C are random weight coefficients:
[0145] D=|CX rand (t)-X(t)| (4)
[0146] X(t+1)=X rand (t)-AD (5)
[0147] Where D is the distance from the data value at this moment to the target optimization value; X(t) and X(t+1) are the current position and the position after the next iteration respectively; t is the current iteration time; X randIt is a random value generated by a random function within the value range;
[0148] The following two formulas are used to search for the weight coefficient of each group of data and the optimal value of the data at this moment:
[0149] D=|CX * (t)-X(t)| (6)
[0150] X * (t+1)=X * (t)-AD (7)
[0151] Where, X * (t) and X * (t+1) are the best data value at the current moment and the best data value after the next iteration moment respectively;
[0152] In the prediction module, multiple sets of data sequence prediction results with different characteristics are linearly superimposed to obtain the prediction result of the traffic transmission rate at the next moment; the prediction data signal Y*(t) is expressed as:
[0153] Y*(t)=X * 1 (t)+X * 2 (t)+X * 3 (t)+…+X * N (t) (8);
[0154] Calibration module: used to establish a good evaluation index system, evaluate and analyze the prediction results obtained from predictions in different time periods, and save the information data that meets the evaluation index range requirements as historical information data, which is used as the historical value for predicting the traffic transmission rate at later moments; information data that does not meet the requirements is discarded, that is, the historical data of the previous moment is used to predict the traffic transmission rate value at this moment.
[0155] In the calibration module, the evaluation index system is: using the error analysis index to analyze the prediction results, selecting the root mean square error RMSE and the fitting coefficient R2 to analyze the quality of the prediction results; here, the root mean square error RMSE cut-off value is set to 2, and the fitting coefficient R2 value is set to 1; information data that meets the evaluation index range requirements, that is, the root mean square error RMSE value does not exceed 2, and the fitting coefficient R2 value does not exceed 1, is saved as historical information data; information data that does not meet the requirements is discarded;
[0156]
[0157]
[0158] In the formula, n represents the number of prediction samples, p i Indicates the predicted fiber transmission rate value, q i Indicates the actual fiber transmission rate value. Indicates the average optical fiber transmission rate.
[0159] The optical fiber flow transmission rate prediction system for power stations of this embodiment has the same advantages as the above-mentioned optical fiber flow transmission rate prediction method for power stations over the prior art, which will not be described in detail here.
[0160] The above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions, improvements, etc. made within the spirit and principle of the present invention should be included in the protection scope of the present invention.
Claims
1. A method for predicting optical fiber flow transmission rate for power stations, characterized in that: The following steps are involved: S1, sending marked original information data at the beginning of the optical fiber within a unit time, and receiving the original information data sequence at the other end of the optical fiber; S2, decomposing the original information data sequence to obtain multiple groups of data sequences with different characteristics, and obtaining weight coefficients of each group of data by calculation; S3, predicting the multiple groups of data sequences with different characteristics; linearly superimposing the prediction results obtained from the multiple groups of data sequences with different characteristics according to the calculated weight coefficients, and using the superimposed results as the prediction results of the flow transmission rate at the next moment, and also including step S4, establishing an evaluation index system, evaluating and analyzing the prediction result values obtained by prediction in different time periods, and saving the information data that meets the evaluation index range requirements as historical information data, as the historical value for predicting the flow transmission rate at the subsequent moment; the information data that does not meet the requirements is discarded, and the historical data of the previous moment is used to predict the flow transmission rate value at the current moment. In the step S2, firstly, the multiple groups of data sequences with different characteristics are decomposed into multiple data sequences with simple characteristics, and the method is: first find all the maximum and minimum points in the original data Y(t) or the signal, connect all the maximum points and the minimum N value points respectively, obtain the upper and lower limits and range of the data Y(t), and calculate the mean n(t) of the upper and lower limits; let g1(t) = Y(t)-n(t), check whether g1(t) meets the following two conditions: (1) The number of extreme points in the signal is equal to or differs by 1 from the number of zero crossing points; (2) At any point in time, the mean of the upper and lower limits determined by the local extreme value of the signal is zero; If the above conditions are not met, repeat the above steps. If they are met, record f1(t)=g1(t) as the first decomposition sequence. Repeat the above steps until the data sequence is a monotonic function or a constant value, and obtain a linear data sequence. In step S2, the signal Y(t) of the original data is expressed as Y(t)=f1(t)+f2(t)+f3(t)+...+f N (t)+i(t) (1) Where t is the number of elements in the time series, N is the number of decomposed components, and i(t) is the component left after the original signal cannot be decomposed, i.e., the remaining term. Through n sets of original data input variables, that is, N sets of different sample training data x j , j = 1, ..., N, to obtain the setting function g(a i , b i , x j ), i = 1, 2, ..., T, thereby obtaining the optimized output matrix function Z(t) based on the sample data; the model expression is as follows: Z=Hβ Where H is the hidden layer output matrix; z is the output matrix; β is the weight matrix between the hidden layer and the output layer; a i is the matrix of corresponding weights between all nodes in the input layer and the i-th node in the hidden layer; b i is the threshold of the i-th hidden layer node; In the matrix H, calculate the optimal weights for In the formula, H + is the generalized inverse matrix of the matrix H; In step S3, the following two formulas are used to predict the positions of the multiple groups of data sequences with different characteristics obtained by decomposition within the value range, where A and C are random weight coefficients: D=|CX rand (t)-X(t)| (4) X(t+1)=X rand (t)-AD (5) Where D is the distance from the data value at this moment to the target optimization value; X(t) and X(t+1) are the current position and the position after the next iteration respectively; t is the current iteration time; X rand It is a random value generated by a random function within the value range; The following two formulas are used to search for the weight coefficient of each group of data and the optimal value of the data at this moment: D=|CX * (t)-X(t)| (6) X * (t+1)=X * (t)-AD (7) Where, X * (t) and X * (t+1) are the best data value at the current moment and the best data value after the next iteration moment respectively; In step S3, multiple sets of data sequence prediction results with different characteristics are linearly superimposed to obtain the prediction result of the traffic transmission rate at the next moment; the predicted data signal Y*(t) is expressed as: Y*(t)=X*1(t)+X*2(t)+X*3(t)+…+X*N(t) (8); In step S4, the evaluation index system is: using the error analysis index to analyze the prediction results, selecting the root mean square error RM SE and the fitting degree coefficient R 2 To analyze the pros and cons of the prediction results; here, the root mean square error RM SE cut-off value is set to 2, and the fitting degree coefficient R is set 2 The value is 1; the information data meets the requirements of the evaluation index range, that is, the root mean square error RM SE value does not exceed 2, and the fitting degree coefficient R 2 If the value does not exceed 1, it will be saved as historical information data; information data that does not meet the requirements will be discarded; In the formula, n represents the number of prediction samples, p i Represents the predicted fiber transmission rate value, q i Indicates the actual fiber transmission rate value. Indicates the average optical fiber transmission rate.
2. A power station optical fiber flow transmission rate prediction system, characterized in that: Includes the following modules: Data sending and receiving module: used to send marked original information data at the beginning of the optical fiber within a unit time, and receive the original information data sequence at the other end of the optical fiber; Data processing module: used to decompose the original information data sequence to obtain multiple groups of data sequences with different characteristics, and obtain the weight coefficient of each group of data by calculation; Prediction module: used to predict the multiple groups of data sequences with different characteristics; linearly superimpose the prediction results obtained from the multiple groups of data sequences with different characteristics according to the calculated weight coefficients, and use the superimposed results as the prediction results of the traffic transmission rate at the next moment; Calibration module: used to establish a good evaluation index system, evaluate and analyze the prediction results obtained from predictions in different time periods, and save the information data that meets the evaluation index range requirements as historical information data, which is used as the historical value for predicting the traffic transmission rate at the following moment; the information data that does not meet the requirements is discarded, and the historical data at the previous moment is used to predict the traffic transmission rate value at the current moment; In the data processing module, multiple groups of data sequences with different characteristics are first decomposed into multiple data sequences with simple characteristics. The method is as follows: first find all the maximum and minimum points in the original data Y(t) or the signal, connect all the maximum and minimum points respectively, obtain the upper and lower limits and range of the data Y(t), and calculate the mean n(t) of the upper and lower limits; let g1(t) = Y(t)-n(t), and check whether g1(t) meets the following two conditions: (1) The number of extreme points in the signal is equal to or differs by 1 from the number of zero crossing points; (2) At any point in time, the mean of the upper and lower limits determined by the local extreme value of the signal is zero; If the above conditions are not met, repeat the above steps. If they are met, record f1(t) = g1(t) as the first decomposition sequence. Repeat the above steps until the data sequence is a monotonic function or a constant value, and obtain a linear data sequence. In the data processing module, the signal Y(t) of the original data is expressed as Y(t)=f1(t)+f2(t)+f3(t)+...+f N (t)+i(t) (1) Where t is the number of elements in the time series, N is the number of decomposed components, and i(t) is the component left after the original signal cannot be decomposed, i.e., the remaining term. Through n sets of original data input variables, that is, N sets of different sample training data x j , j = 1, ..., N, to obtain the setting function g(a i , b i , x i ), i = 1, 2, ..., T, thereby obtaining the optimized output matrix function Z(t) based on the sample data; the model expression is as follows: Z=Hβ Where H is the hidden layer output matrix; z is the output matrix; β is the weight matrix between the hidden layer and the output layer; a i is the matrix of corresponding weights between all nodes in the input layer and the i-th node in the hidden layer; b i is the threshold of the i-th hidden layer node; In the matrix H, calculate the optimal weights for In the formula, H + is the generalized inverse matrix of the matrix H; In the prediction module, the following two formulas are used to predict the positions of multiple groups of data sequences with different characteristics obtained by decomposition within the value range, where A and C are random weight coefficients: D=|CX rand (t)-X(t)| (4) X(t+1)=X rand (t)-AD (5) Where D is the distance from the data value at this moment to the target optimization value; X(t) and X(t+1) are the current position and the position after the next iteration respectively; t is the current iteration time; X rand It is a random value generated by a random function within the value range; The following two formulas are used to search for the weight coefficient of each group of data and the optimal value of the data at this moment: D=|CX * (t)-X(t)| (6) X * (t+1)=X * (t)-AD (7) Where, X * (t) and X * (t+1) are the best data value at the current moment and the best data value after the next iteration moment respectively; In the prediction module, multiple sets of data sequence prediction results with different characteristics are linearly superimposed to obtain the prediction result of the traffic transmission rate at the next moment; the prediction data signal Y*(t) is expressed as: Y*(t)=X*1(t)+X*2(t)+X*3(t)+…+X*N(t) (8); In the calibration module, the evaluation index system is: using the error analysis index to analyze the prediction results, selecting the root mean square error RM SE and the fitting degree coefficient R 2 To analyze the pros and cons of the prediction results; here, the root mean square error RM SE cut-off value is set to 2, and the fitting degree coefficient R is set 2 The value is 1; the information data meets the requirements of the evaluation index range, that is, the root mean square error RMS E value does not exceed 2, and the fitting degree coefficient R 2 If the value does not exceed 1, it will be saved as historical information data; information data that does not meet the requirements will be discarded; In the formula, n represents the number of prediction samples, p i Represents the predicted fiber transmission rate value, q i Indicates the actual fiber transmission rate value. Indicates the average optical fiber transmission rate.
