Empirical mode decomposition (EMD)-combined method for predicting circulating oil temperature of hot oil during installation of LSTM main equipment

Abstract

The invention relates to the technical field of extra-high voltage power transformation project main equipment installation, and provides an EMD-combined LSTM main equipment installation hot oil circulation oil temperature prediction method, which comprises the steps of collecting time sequence data of instantaneous flow, water content, gas content, granularity, oil filter outlet oil temperature and transformer outlet oil temperature for oil temperature prediction in a main equipment installation hot oil circulation link; using an EMD decomposition method to decompose the transformer outlet oil temperature characteristic time sequence data into a plurality of IMF components and an RES component; calculating a Pearson's correlation coefficient, and screening and filtering components of which the correlation is lower than a preset threshold value; and inputting data obtained by combining the screened component after EMD decomposition of the transformer outlet oil temperature within a certain time range with the instantaneous flow rate, the water content, the gas content, the granularity and the oil filter outlet oil temperature into an LSTM model for multivariable oil temperature prediction. The method can effectively improve the prediction precision of the standard reaching time of the outlet oil temperature of the transformer.

Description

Technical Field

[0001] The present invention relates to the technical field of main equipment installation of ultra-high voltage power transformation projects, and in particular to a method for predicting the oil temperature of hot oil circulation during the installation of LSTM main equipment combined with EMD. Background Art

[0002] At present, the field of installation of main equipment of UHV projects has gradually applied the on-site monitoring device (PLC) to implement installation process monitoring. By integrating oil filter, on-site monitoring device, transformer, professional tooling, gateway and other tools with main equipment, a hot oil circulation path is formed. The transformer oil is filtered and heated by the oil filter and then transmitted to the transformer oil tank. It is connected to the on-site monitoring device through a pipeline and then returns to the oil filter. The monitoring data is transmitted to the cloud backend through the gateway, realizing the real-time monitoring of the outlet oil temperature of the vacuum oil filter, the outlet oil temperature of the transformer, the instantaneous flow rate, the water content, the gas content, the particle size and other indicators of the hot oil circulation process of the oil-immersed equipment installation. The outlet oil temperature of the transformer, as an important indicator of the process quality control of the hot oil circulation process, needs to be gradually heated to 55°C during the hot oil circulation process, and it starts to be maintained for 96 hours after 55°C. Knowing in advance when the transformer outlet oil temperature reaches the standard timing is also of great guiding significance to the site. However, the current monitoring results only realize the real-time monitoring of the hot oil circulation indicators, and fail to predict the time when the transformer outlet oil temperature reaches the standard timing.

[0003] It can be seen that the traditional solution only provides real-time monitoring of the hot oil circulation indicators of the main equipment installed in the UHV substation project, lacks intelligent analysis links, and cannot provide a time prediction for the transformer outlet oil temperature to reach the standard. The transformer outlet oil temperature is affected by many factors, such as the instantaneous flow rate in the pipeline (affecting the flow rate), the water content, gas content, and particle size of the transformer oil (affecting the physical, chemical and electrical properties of the oil), the oil temperature at the outlet of the oil filter (affecting the initial oil temperature), etc. It is difficult to judge the time when the transformer outlet oil temperature reaches the standard in different transformer hot oil circulation processes based on experience. Summary of the invention

[0004] The purpose of the present invention is to solve at least one technical problem in the background technology and to provide a method for predicting the oil temperature of a hot oil circulation installed in an LSTM main device combined with EMD.

[0005] To achieve the above object, the present invention provides a method for predicting the oil temperature of a hot oil circulation installed in an LSTM main equipment combined with EMD, comprising:

[0006] Collect the time series data of instantaneous flow rate, water content, gas content, particle size, oil filter outlet oil temperature and transformer outlet oil temperature for oil temperature prediction in the hot oil circulation link of the main equipment installation;

[0007] The transformer outlet oil temperature characteristic time series data is decomposed into multiple IMF components and one RES component using the EMD decomposition method;

[0008] Calculate the Pearson correlation coefficient between each decomposed component and the original time series of the transformer outlet oil temperature, and filter out the components whose correlation is lower than the preset threshold T;

[0009] The data of the transformer outlet oil temperature within a certain time range, which is decomposed by EMD and filtered, combined with the instantaneous flow rate, water content, gas content, particle size, and oil filter outlet oil temperature, are input into the LSTM model for multivariate prediction.

[0010] According to one aspect of the present invention, the use of the EMD decomposition method to decompose the transformer outlet oil temperature time series data into multiple IMF components and one RES component includes:

[0011] S1. For a non-stationary original transformer outlet oil temperature time series x(t), use the cubic spline function to fit all its maximum points and all its minimum points, and obtain the maximum envelope c1(t) and minimum envelope f1(t) respectively;

[0012] S2. Calculate the mean m1(t) of the maximum envelope c1(t) and the minimum envelope f1(t):

[0013]

[0014] S3. Subtract m1(t) from the original transformer outlet oil temperature time series to obtain a new time series x ′ 1(t):

[0015] x ′ 1(t)=x(t)―m1(t);

[0016] S4. Judge x ′ 1(t) whether it meets the IMF conditions. If it does, then x ′ 1(t) is the first IMF component. If it is not satisfied, repeat steps S1-S3 until the IMF condition is satisfied, and the first IMF component IMF1 is obtained.

[0017] The IMF conditions include: (1) the number of extreme points and zero-crossing points of the IMF component should be equal or differ by at most one; (2) for each time point, the average value of the maximum envelope formed by the local maximum points and the average value of the minimum envelope formed by the local minimum points is zero; the above two conditions must be met at the same time to be judged to meet the IMF conditions;

[0018] Step S5. The first IMF component IMF1 represents the highest frequency component of the original sequence. Subtract the first IMF component IMF1 from the original sequence x(t) to obtain the first difference sequence r1(t) without high-frequency components. Repeat the stationary process of steps S1-S4 for r1(t) to obtain the second IMF component IMF2 until it cannot be decomposed. Then r n (t) is the residual component after decomposition, recorded as Res component:

[0019] r1(t)=x(t)―IMF1;

[0020] r2(t)=r1(t)―IMF2;

[0021] …

[0022] Res=r n (t) = r n―1 (t) ―IMF n .

[0023] According to one aspect of the present invention, the calculation formula of the Pearson correlation coefficient R is:

[0024]

[0025] Among them, X i (i=1,2…n),Y i (i=1,2…n) represent the time series of the two sets of transformer outlet oil temperatures, and represent the means of X and Y respectively, σX and σY represent the standard deviations of X and Y respectively.

[0026] According to one aspect of the present invention, the training process of the LSTM model includes:

[0027] S11. Build a memory unit and calculate the value of the input gate;

[0028] S12. Calculate the forgetting state information;

[0029] S13. Update the memory unit state information and calculate the candidate value of the memory unit state;

[0030] S14. Output memory unit state information and calculate output gate;

[0031] S15. Repeat steps S11-S14 until the preset termination condition is met, and the training process ends.

[0032] To achieve the above object, the present invention also provides a hot oil circulation oil temperature prediction system for LSTM main equipment installation combined with EMD, comprising:

[0033] The data acquisition module collects the time series data of instantaneous flow rate, water content, gas content, particle size, oil filter outlet oil temperature and transformer outlet oil temperature for oil temperature prediction in the hot oil circulation link of the main equipment installation;

[0034] The data decomposition module uses the EMD decomposition method to decompose the transformer outlet oil temperature characteristic time series data into multiple IMF components and one RES component;

[0035] The data component screening module calculates the Pearson correlation coefficient between each decomposed component and the original time series of the transformer outlet oil temperature, and filters out the components whose correlation is lower than the preset threshold T;

[0036] The oil temperature prediction module inputs the data of the transformer outlet oil temperature within a certain time range, which is decomposed by EMD and filtered, and combined with the instantaneous flow rate, water content, gas content, particle size, and oil filter outlet oil temperature, into the LSTM model for multivariate prediction.

[0037] To achieve the above objectives, the present invention also provides an electronic device, comprising a processor, a memory, and a computer program stored in the memory and executable on the processor, wherein when the computer program is executed by the processor, the method for predicting the oil temperature of a hot oil circulation installed in an LSTM main device combined with EMD as described above is implemented.

[0038] To achieve the above objectives, the present invention also provides a computer-readable storage medium, on which a computer program is stored. When the computer program is executed by a processor, the method for predicting the oil temperature of a hot oil circulation installed in an LSTM main device combined with EMD as described above is implemented.

[0039] According to the solution of the present invention, the present invention decomposes the transformer outlet oil temperature data into multiple components of different frequencies by using EMD decomposition, realizes the smooth processing of sequence data, and removes components with low correlation by filtering through the Pearson correlation coefficient, which can reduce the complexity of the model to a certain extent, and uses the retained components and other original indicator sequence data as the input of the LSTM model to perform multivariate prediction. Through decomposition and reconstruction, the hidden information in the transformer outlet oil temperature data can be better extracted, and the accuracy of the prediction of the transformer outlet oil temperature reaching the standard time can be effectively improved.

[0040] According to the solution of the present invention, the present invention constructs an LSTM model combined with EMD, uses EMD to carry out transformer outlet oil temperature data processing, constructs input gates, forget gates, and output gates, and dynamically analyzes the importance of time series monitoring indicators during the installation of hot oil circulation in the main equipment, thereby realizing dynamic prediction of the time when the outlet oil temperature of the main equipment reaches the standard, and providing guidance for on-site understanding of when the outlet oil temperature of the transformer hot oil circulation reaches the standard.

[0041] According to the solution of the present invention, the LSTM UHV main equipment installation hot oil circulation oil temperature prediction method combined with EMD designed by the present invention, on the one hand, the method fully utilizes the monitoring index data in the process of main equipment installation hot oil circulation for model training and verification, without the need to collect other data additionally; on the other hand, the method can realize the prediction of real-time hot oil circulation qualified time, and predict the latest hot oil circulation qualified time through all index data before the current moment, which is not available in the current hot oil main equipment installation hot oil circulation circulation monitoring scheme. BRIEF DESCRIPTION OF THE DRAWINGS

[0042] Figure 1 Schematically showing a flow chart of a method for predicting oil temperature of hot oil circulation installed in a LSTM main device combined with EMD according to an embodiment of the present invention;

[0043] Figure 2 Schematically showing a flow chart of inputting data into an LSTM model for prediction according to an embodiment of the present invention;

[0044] Figure 3 The figure schematically shows a logical operation diagram of an LSTM model according to an embodiment of the present invention. DETAILED DESCRIPTION

[0045] The present invention will now be discussed with reference to exemplary embodiments. It should be understood that the embodiments discussed are only to enable those skilled in the art to better understand and thus implement the present invention, rather than implying any limitation on the scope of the present invention.

[0046] As used herein, the term “including” and variations thereof are to be interpreted as open-ended terms meaning “including, but not limited to.” The term “based on” is to be interpreted as “based, at least in part, on.” The terms “one embodiment” and “an embodiment” are to be interpreted as “at least one embodiment.”

[0047] Figure 1 The flowchart of the method for predicting the oil temperature of the hot oil circulation installed in the LSTM main equipment combined with the EMD according to one embodiment of the present invention is schematically shown. Figure 1 As shown, in this embodiment, the hot oil circulation oil temperature prediction method is installed in combination with the LSTM main device of the EMD, including:

[0048] Collect the time series data of instantaneous flow rate, water content, gas content, particle size, oil filter outlet oil temperature and transformer outlet oil temperature for oil temperature prediction in the hot oil circulation link of the main equipment installation;

[0049] The transformer outlet oil temperature characteristic time series data is decomposed into multiple IMF components and one RES component using the EMD decomposition method;

[0050] Calculate the Pearson correlation coefficient between each decomposed component and the original time series of the transformer outlet oil temperature, and filter out the components whose correlation is lower than the preset threshold T;

[0051] The data of the transformer outlet oil temperature combined with the instantaneous flow rate, water content, gas content, particle size, and oil filter outlet oil temperature after EMD decomposition within a certain time range are input into the LSTM model for multivariate prediction. With this setting, considering that the transformer outlet oil temperature has a greater correlation with the transformer compliance time, the EMD empirical mode decomposition is used to process the transformer outlet oil temperature data to reduce the non-stationarity of the original data, and the transformer outlet oil temperature is decomposed into multiple IMF components and participating components. Different components reflect the degree of correlation with the original transformer outlet oil temperature data; the IMF components with greater correlation are selected and recomposed into a new input data set with the remaining instantaneous flow rate, water content, gas content, particle size, and oil filter outlet oil temperature indicators; based on the new input data set, the LSTM transformer outlet oil temperature compliance time prediction is carried out.

[0052] In this embodiment, the various index data (time series data) used for predicting the oil temperature of the hot oil circulation link in the installation of the main equipment are associated with the transformer outlet oil temperature, including the instantaneous flow rate, water content, gas content, particle size, oil filter outlet oil temperature, and transformer outlet oil temperature. The six index data are all associated with the time series and belong to the same time series. They can be divided into installation process indicators and transformer oil quality indicators, among which the installation process indicators are divided into instantaneous flow rate (affecting the pipeline flow rate of the oil, thereby affecting the oil temperature transfer), oil filter outlet oil temperature (the oil needs to be heated by the oil filter, affecting the oil outlet oil temperature, less than or equal to 75°C), and transformer outlet oil temperature (an indicator to be predicted, which itself is also correlated with the prediction result), and the transformer oil quality indicators are divided into water content, gas content, and particle size (all of which will affect the electrical properties, chemical properties and physical properties of the transformer oil, thereby affecting the oil temperature transfer).

[0053] Further, according to an embodiment of the present invention, the transformer outlet oil temperature time series data is decomposed into multiple IMF components and one RES component using the EMD decomposition method, including:

[0054] S1. For a non-stationary original transformer outlet oil temperature time series x(t), use the cubic spline function to fit all its maximum points and all its minimum points, and obtain the maximum envelope c1(t) and minimum envelope f1(t) respectively;

[0055] S2. Calculate the mean m1(t) of the maximum envelope c1(t) and the minimum envelope f1(t):

[0056]

[0057] S3. Subtract m1(t) from the original transformer outlet oil temperature time series to obtain a new time series x ′ 1(t):

[0058] x ′ 1(t)=x(t)―m1(t);

[0059] S4. Judge x ′ 1(t) whether it meets the IMF conditions. If it does, then x ′ 1(t) is the first IMF component. If it is not satisfied, repeat steps S1-S3 until the IMF condition is satisfied, and the first IMF component IMF1 is obtained.

[0060] The IMF conditions include: (1) the number of extreme points and zero-crossing points of the IMF component should be equal or differ by at most one; (2) for each time point, the average value of the maximum envelope formed by the local maximum points and the average value of the minimum envelope formed by the local minimum points is zero; the above two conditions must be met at the same time to be judged to meet the IMF conditions;

[0061] Step S5. The first IMF component IMF1 represents the highest frequency component of the original sequence. Subtract the first IMF component IMF1 from the original sequence x(t) to obtain the first difference sequence r1(t) without high-frequency components. Repeat the stationary process of steps S1-S4 for r1(t) to obtain the second IMF component IMF2 until it cannot be decomposed. Then r n (t) is the residual component after decomposition, recorded as Res component:

[0062] r1(t)=x(t)―IMF1;

[0063] r2(t)=r1(t)―IMF2;

[0064] …

[0065] Res=r n (t) = r n―1 (t) ―IMF n .

[0066] Further, according to an embodiment of the present invention, the calculation formula of the Pearson correlation coefficient R is:

[0067]

[0068] Among them, X i (i=1,2…n),Y i (i=1,2…n) represent the time series of the two sets of transformer outlet oil temperatures, and represent the means of X and Y respectively, σX and σY represent the standard deviations of X and Y respectively.

[0069] Further, according to an embodiment of the present invention, when the data of the transformer outlet oil temperature combined with the instantaneous flow rate, water content, gas content, particle size, and oil filter outlet oil temperature within a certain time range is input into the LSTM model for multivariate oil temperature prediction, the data within the time period from the start of hot oil circulation to the start of timing when the transformer inlet and outlet temperatures reach the process requirements is selected as input information. The flowchart of data input into the LSTM model for prediction is as follows: Figure 2 shown.

[0070] Further, Figure 3 The following is a schematic diagram showing a logical operation diagram of an LSTM model according to an embodiment of the present invention. Figure 2 and Figure 3 As shown, in this embodiment, the training process of the LSTM model includes:

[0071] S11. Build a memory unit and calculate the value of the input gate;

[0072] In this implementation, the input gate is used to control the impact of the current data input on the memory cell value. In addition to being affected by the current input data and the LSTM cell output value at the previous moment, the calculation of all gates is also affected by the memory cell value at the previous moment. This layer can be divided into two parts. First, the sigmoid layer determines which values ​​need to be updated, which are calculated by the input gate. Then, a vector is created in the tanh layer, which contains the new information to be added and can be added to the new state cell.

[0073] Assume that i t Represents the input gate, with f t Represents the forget gate, with o t represents the output gate, the subscript t represents the time, c t represents the cell state at time t, c′ t represents the candidate value of the cell state at time t, x t represents the input at time t, h t represents the hidden state at time t, σ represents the activation function, W and b are the weight and bias respectively, then:

[0074] i t =σ(W i ·[h t―1 ,x t ]+b i );

[0075] c′ t =tanh(WG ·[h t―1 ,x t ]+b G );

[0076] Among them, W i , W G , b i , b G are the relevant weights and biases, respectively, [h t―1 ,x t ] is the hidden state of the previous time step and the input of the current time step.

[0077] S12. Calculate the forgetting state information (i.e., forgetting gate);

[0078] In this implementation, the forget gate determines which information from the previous unit needs to be discarded from the unit state, outputs a value of [0,1] for each number in the unit state, and takes the input of the current layer and the output of the previous layer as input. The forget state information at time t is:

[0079] f t =σ(W f ·[h t―1 ,x t ]+b f );

[0080] Among them, W f and b f They are the weight matrix and bias term of the forget gate respectively.

[0081] S13. Update the memory unit state information and calculate the candidate value of the memory unit state;

[0082] In this embodiment, this step will determine which information is stored in the unit state information, and then update the unit state by combining the information of the input gate and the forget gate. First, the state of the old unit is updated, and then the new information is input for the new unit state, that is, the information is updated in the unit state:

[0083] c t =f t ·c t―1 +i t ·c′ t .

[0084] S14. Output memory unit state information and calculate output gate;

[0085] In this embodiment, a sigmoid layer is used to determine the relevant information of the output unit state, and a tanh layer is used to process the unit state to obtain the output gate information;

[0086] o t=σ(W o ·[h t―1 ,x t ]+b o );

[0087] h t =o t ·tanh (c t );

[0088] Among them, t is the output of the sigmoid of the output gate, c t is the updated memory cell state, h t is the final hidden state output.

[0089] S15. Repeat steps S11-S14 until the preset termination condition is met, and the training process ends.

[0090] According to the above scheme of the present invention, the present invention decomposes the transformer outlet oil temperature data into multiple components of different frequencies by using EMD decomposition, realizes the smooth processing of sequence data, and removes components with low correlation by filtering through the Pearson correlation coefficient, which can reduce the complexity of the model to a certain extent, and uses the retained components and other original indicator sequence data as the input of the LSTM model for multivariate prediction. Through decomposition and reconstruction, the hidden information in the transformer outlet oil temperature data can be better extracted, and the accuracy of the prediction of the transformer outlet oil temperature reaching the standard time can be effectively improved.

[0091] According to the above scheme of the present invention, the present invention constructs an LSTM model combined with EMD, uses EMD to carry out transformer outlet oil temperature data processing, constructs input gates, forget gates, and output gates, and dynamically analyzes the importance of time series monitoring indicators during the installation of hot oil circulation in the main equipment, thereby realizing dynamic prediction of the time when the outlet oil temperature of the main equipment reaches the standard, and providing guidance for on-site understanding of when the outlet oil temperature of the transformer hot oil circulation reaches the standard.

[0092] According to the above scheme of the present invention, the LSTM UHV main equipment installation hot oil circulation oil temperature prediction method combined with EMD designed by the present invention, on the one hand, the method fully utilizes the monitoring index data in the process of main equipment installation hot oil circulation for model training and verification, without the need to collect other data additionally; on the other hand, the method can realize the real-time prediction of the hot oil circulation qualified time, and predict the latest hot oil circulation qualified time through all the index data before the current moment, which is not available in the current hot oil main equipment installation hot oil circulation circulation monitoring scheme.

[0093] Furthermore, to achieve the above-mentioned purpose, the present invention also provides a hot oil circulation oil temperature prediction system for LSTM main equipment installation combined with EMD, comprising:

[0094] The data acquisition module collects the time series data of instantaneous flow rate, water content, gas content, particle size, oil filter outlet oil temperature and transformer outlet oil temperature for oil temperature prediction in the hot oil circulation link of the main equipment installation;

[0095] The data decomposition module uses the EMD decomposition method to decompose the transformer outlet oil temperature characteristic time series data into multiple IMF components and one RES component;

[0096] The data component screening module calculates the Pearson correlation coefficient between each decomposed component and the original time series of the transformer outlet oil temperature, and filters out the components whose correlation is lower than the preset threshold T;

[0097] The oil temperature prediction module inputs the data of the transformer outlet oil temperature combined with the instantaneous flow rate, water content, gas content, particle size, and oil filter outlet oil temperature within a certain time range after the EMD decomposition and screening components into the LSTM model for multivariate prediction. With this setting, considering that the transformer outlet oil temperature has a greater correlation with the transformer compliance time, the EMD empirical mode decomposition is used to process the transformer outlet oil temperature data to reduce the non-stationarity of the original data, and the transformer outlet oil temperature is decomposed into multiple IMF components and participating components. Different components reflect the degree of correlation with the original transformer outlet oil temperature data; select the IMF component with greater correlation and re-compose the new input data set with the remaining instantaneous flow rate, water content, gas content, particle size, and oil filter outlet oil temperature indicators; based on the new input data set, the LSTM transformer outlet oil temperature compliance time prediction is carried out.

[0098] In this embodiment, the various index data (time series data) used for predicting the oil temperature of the hot oil circulation link in the installation of the main equipment are associated with the transformer outlet oil temperature, including the instantaneous flow rate, water content, gas content, particle size, oil filter outlet oil temperature, and transformer outlet oil temperature. The six index data are all associated with the time series and belong to the same time series. They can be divided into installation process indicators and transformer oil quality indicators, among which the installation process indicators are divided into instantaneous flow rate (affecting the pipeline flow rate of the oil, thereby affecting the oil temperature transfer), oil filter outlet oil temperature (the oil needs to be heated by the oil filter, affecting the oil outlet oil temperature, less than or equal to 75°C), and transformer outlet oil temperature (an indicator to be predicted, which itself is also correlated with the prediction result), and the transformer oil quality indicators are divided into water content, gas content, and particle size (all of which will affect the electrical properties, chemical properties and physical properties of the transformer oil, thereby affecting the oil temperature transfer).

[0099] Further, according to an embodiment of the present invention, the transformer outlet oil temperature time series data is decomposed into multiple IMF components and one RES component using the EMD decomposition method, including:

[0100] S1. For a non-stationary original transformer outlet oil temperature time series x(t), use the cubic spline function to fit all its maximum points and all its minimum points, and obtain the maximum envelope c1(t) and minimum envelope f1(t) respectively;

[0101] S2. Calculate the mean m1(t) of the maximum envelope c1(t) and the minimum envelope f1(t):

[0102]

[0103] S3. Subtract m1(t) from the original transformer outlet oil temperature time series to obtain a new time series x ′ 1(t):

[0104] x ′ 1(t)=x(t)―m1(t);

[0105] S4. Judge x ′ 1(t) whether it meets the IMF conditions. If it does, then x ′ 1(t) is the first IMF component. If it is not satisfied, repeat steps S1-S3 until the IMF condition is satisfied, and the first IMF component IMF1 is obtained.

[0106] The IMF conditions include: (1) the number of extreme points and zero-crossing points of the IMF component should be equal or differ by at most one; (2) for each time point, the average value of the maximum envelope formed by the local maximum points and the average value of the minimum envelope formed by the local minimum points is zero; the above two conditions must be met at the same time to be judged to meet the IMF conditions;

[0107] Step S5. The first IMF component IMF1 represents the highest frequency component of the original sequence. Subtract the first IMF component IMF1 from the original sequence x(t) to obtain the first difference sequence r1(t) without high-frequency components. Repeat the stationary process of steps S1-S4 for r1(t) to obtain the second IMF component IMF2 until it cannot be decomposed. Then r n (t) is the residual component after decomposition, recorded as Res component:

[0108] r1(t)=x(t)―IMF1;

[0109] r2(t)=r1(t)―IMF2;

[0110] …

[0111] Res=r n (t) = r n―1 (t) ―IMF n .

[0112] Further, according to an embodiment of the present invention, the calculation formula of the Pearson correlation coefficient R is:

[0113]

[0114] Among them, X i (i=1,2…n),Y i (i=1,2…n) represent the time series of the two sets of transformer outlet oil temperatures, and represent the means of X and Y respectively, σX and σY represent the standard deviations of X and Y respectively.

[0115] Further, according to an embodiment of the present invention, when the data of the transformer outlet oil temperature combined with the instantaneous flow rate, water content, gas content, particle size, and oil filter outlet oil temperature within a certain time range is input into the LSTM model for multivariate oil temperature prediction, the data within the time period from the start of hot oil circulation to the start of timing when the transformer inlet and outlet temperatures reach the process requirements is selected as input information. The flowchart of data input into the LSTM model for prediction is as follows: Figure 2 shown.

[0116] Furthermore, combined with Figure 2 and Figure 3 As shown, in this embodiment, the training process of the LSTM model includes:

[0117] S11. Build a memory unit and calculate the value of the input gate;

[0118] In this implementation, the input gate is used to control the impact of the current data input on the memory cell value. In addition to being affected by the current input data and the LSTM cell output value at the previous moment, the calculation of all gates is also affected by the memory cell value at the previous moment. This layer can be divided into two parts. First, the sigmoid layer determines which values ​​need to be updated, which are calculated by the input gate. Then, a vector is created in the tanh layer, which contains the new information to be added and can be added to the new state cell.

[0119] Assume that i t Represents the input gate, with f t Represents the forget gate, with o t represents the output gate, the subscript t represents the time, c t represents the cell state at time t, c′ t represents the candidate value of the cell state at time t, x t represents the input at time t, h t represents the hidden state at time t, σ represents the activation function, W and b are the weight and bias respectively, then:

[0120] i t =σ(W i ·[h t―1 ,x t ]+b i );

[0121] c′ t =tanh(W G ·[h t―1 ,x t ]+b G );

[0122] Among them, W i , W G , b i , b G are the relevant weights and biases, respectively, [h t―1 ,x t ] is the hidden state of the previous time step and the input of the current time step.

[0123] S12. Calculate the forgetting state information (i.e., forgetting gate);

[0124] In this implementation, the forget gate determines which information from the previous unit needs to be discarded from the unit state, outputs a value of [0,1] for each number in the unit state, and takes the input of the current layer and the output of the previous layer as input. The forget state information at time t is:

[0125] f t =σ(W f ·[h t―1 ,x t ]+b f );

[0126] Among them, W f and b f They are the weight matrix and bias term of the forget gate respectively.

[0127] S13. Update the memory unit state information and calculate the candidate value of the memory unit state;

[0128] In this embodiment, this step will determine which information is stored in the unit state information, and then update the unit state by combining the information of the input gate and the forget gate. First, the state of the old unit is updated, and then the new information is input for the new unit state, that is, the information is updated in the unit state:

[0129] c t =f t ·c t―1 +i t ·c′ t .

[0130] S14. Output memory unit state information and calculate output gate;

[0131] In this embodiment, a sigmoid layer is used to determine the relevant information of the output unit state, and a tanh layer is used to process the unit state to obtain the output gate information;

[0132] o t =σ(W o ·[h t―1 ,x t ]+b o );

[0133] h t =o t ·tanh (c t );

[0134] Among them, t is the output of the sigmoid of the output gate, c t is the updated memory cell state, h t is the final hidden state output.

[0135] S15. Repeat steps S11-S14 until the preset termination condition is met, and the training process ends.

[0136] According to the above scheme of the present invention, the present invention decomposes the transformer outlet oil temperature data into multiple components of different frequencies by using EMD decomposition, realizes the smooth processing of sequence data, and removes components with low correlation by filtering through the Pearson correlation coefficient, which can reduce the complexity of the model to a certain extent, and uses the retained components and other original indicator sequence data as the input of the LSTM model for multivariate prediction. Through decomposition and reconstruction, the hidden information in the transformer outlet oil temperature data can be better extracted, and the accuracy of the prediction of the transformer outlet oil temperature reaching the standard time can be effectively improved.

[0137] According to the above scheme of the present invention, the present invention constructs an LSTM model combined with EMD, uses EMD to carry out transformer outlet oil temperature data processing, constructs input gates, forget gates, and output gates, and dynamically analyzes the importance of time series monitoring indicators during the installation of hot oil circulation in the main equipment, thereby realizing dynamic prediction of the time when the outlet oil temperature of the main equipment reaches the standard, and providing guidance for on-site understanding of when the outlet oil temperature of the transformer hot oil circulation reaches the standard.

[0138] According to the above scheme of the present invention, the LSTM UHV main equipment installation hot oil circulation oil temperature prediction method combined with EMD designed by the present invention, on the one hand, the method fully utilizes the monitoring index data in the process of main equipment installation hot oil circulation for model training and verification, without the need to collect other data additionally; on the other hand, the method can realize the real-time prediction of the hot oil circulation qualified time, and predict the latest hot oil circulation qualified time through all the index data before the current moment, which is not available in the current hot oil main equipment installation hot oil circulation circulation monitoring scheme.

[0139] Furthermore, to achieve the above-mentioned purpose, the present invention also provides an electronic device, including a processor, a memory, and a computer program stored in the memory and executable on the processor. When the computer program is executed by the processor, the method for predicting the oil temperature of a hot oil circulation installed in the LSTM main device combined with the EMD as described above is implemented.

[0140] Furthermore, to achieve the above-mentioned purpose, the present invention also provides a computer-readable storage medium, on which a computer program is stored. When the computer program is executed by a processor, the method for predicting the oil temperature of a hot oil circulation installed in an LSTM main device combined with EMD as described above is implemented.

[0141] Those skilled in the art will appreciate that the modules and algorithm steps described in conjunction with the embodiments disclosed herein can be implemented in electronic hardware, or a combination of computer software and electronic hardware. Whether these functions are performed in hardware or software depends on the specific application and design constraints of the technical solution. Professional and technical personnel can use different methods to implement the described functions for each specific application, but such implementation should not be considered to be beyond the scope of the present invention.

[0142] Those skilled in the art can clearly understand that, for the convenience and brevity of description, the specific working processes of the above-described devices and equipment can refer to the corresponding processes in the aforementioned method implementation methods, and will not be repeated here.

[0143] In the embodiments provided in the present application, it should be understood that the disclosed devices and methods can be implemented in other ways. For example, the device embodiments described above are only schematic. For example, the division of the modules is only a logical function division. There may be other division methods in actual implementation, such as multiple modules or components can be combined or integrated into another system, or some features can be ignored or not executed. Another point is that the mutual coupling or direct coupling or communication connection shown or discussed can be through some interfaces, indirect coupling or communication connection of devices or modules, which can be electrical, mechanical or other forms.

[0144] The modules described as separate components may or may not be physically separated, and the components shown as modules may or may not be physical modules, that is, they may be located in one place or distributed on multiple network modules. Some or all of the modules may be selected according to actual needs to achieve the purpose of the embodiments of the present invention.

[0145] In addition, each functional module in the embodiment of the present invention may be integrated into one processing module, or each module may exist physically separately, or two or more modules may be integrated into one module.

[0146] If the functions are implemented in the form of software function modules and sold or used as independent products, they can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of the present invention can be essentially or partly embodied in the form of a software product that contributes to the prior art. The computer software product is stored in a storage medium and includes several instructions for enabling a computer device (which can be a personal computer, server, or network device, etc.) to perform all or part of the steps of the energy-saving signal sending / receiving method of each embodiment of the present invention. The aforementioned storage medium includes: various media that can store program codes, such as USB flash drives, mobile hard disks, ROM, RAM, magnetic disks, or optical disks.

[0147] The above description is only a preferred embodiment of the present application and an explanation of the technical principles used. Those skilled in the art should understand that the scope of the invention involved in the present application is not limited to the technical solution formed by a specific combination of the above technical features, but should also cover other technical solutions formed by any combination of the above technical features or their equivalent features without departing from the inventive concept. For example, the above features are replaced with the technical features with similar functions disclosed in this application (but not limited to) by each other.

[0148] It should be understood that the size of the serial numbers of each step in the content of the invention and the implementation methods of the present invention does not absolutely mean the order of execution. The execution order of each process should be determined by its function and internal logic, and should not constitute any limitation on the implementation process of the implementation methods of the present invention.

Claims

1. Combined with EMD LSTM main equipment installation hot oil circulation oil temperature prediction method, characterized in that, include: Collect the time series data of instantaneous flow rate, water content, gas content, particle size, oil filter outlet oil temperature and transformer outlet oil temperature for oil temperature prediction in the hot oil circulation link of the main equipment installation; The transformer outlet oil temperature characteristic time series data is decomposed into multiple IMF components and one RES component using the EMD decomposition method; Calculate the Pearson correlation coefficient between each decomposed component and the original time series of the transformer outlet oil temperature, and filter out the components whose correlation is lower than the preset threshold T; The data of the transformer outlet oil temperature within a certain time range, which is decomposed by EMD and filtered, combined with the instantaneous flow rate, water content, gas content, particle size, and oil filter outlet oil temperature, are input into the LSTM model for multivariate prediction.

2. The method for predicting the oil temperature of hot oil circulation installed by LSTM main equipment combined with EMD according to claim 1 is characterized in that: The EMD decomposition method is used to decompose the transformer outlet oil temperature time series data into multiple IMF components and one RES component, including: S1. For a non-stationary original transformer outlet oil temperature time series x(t), use the cubic spline function to fit all its maximum points and all its minimum points, and obtain the maximum envelope c1(t) and minimum envelope f1(t) respectively; S2. Calculate the mean m1(t) of the maximum envelope c1(t) and the minimum envelope f1(t): S3. Subtract m1(t) from the original transformer outlet oil temperature time series to obtain a new time series x′1(t): x′1(t)=x(t)-m1(t); S4. Determine whether x′1(t) satisfies the IMF condition. If so, x′1(t) is the first IMF component. If not, repeat steps S1-S3 until the IMF condition is satisfied, thus obtaining the first IMF component IMF1. The IMF conditions include: (1) the number of extreme points and zero crossing points of the IMF component should be equal or differ by at most one; (2) for each time point, the average value of the maximum envelope formed by the local maximum points and the average value of the minimum envelope formed by the local minimum points is zero: the above two conditions must be met at the same time to be judged to meet the IMF conditions; Step S5. The first IMF component IMF1 represents the highest frequency component of the original sequence. Subtract the first IMF component IMF1 from the original sequence x(t) to obtain the first difference sequence r1(t) without high-frequency components. Repeat the stationary process of steps S1-S4 for r1(t) to obtain the second IMF component IMF2 until it cannot be decomposed. Then r n (t) is the residual component after decomposition, recorded as Res component: r1(t)=x(t)-IMF1; r2(t)=r1(t)-IMF2; Res=r n (t)=r n-1 (t)-IMF n 。 3. The method for predicting the oil temperature of hot oil circulation installed by LSTM main equipment combined with EMD according to claim 1 is characterized in that: The calculation formula of the Pearson correlation coefficient R is: Among them, Xi (i = 1, 2...n) and Yi (i = 1, 2...n) represent the time series of the oil temperature at the outlet of two transformers, and represent the means of X and Y respectively, σX and σY represent the standard deviations of X and Y respectively.

4. The method for predicting the oil temperature of hot oil circulation installed by the LSTM main equipment combined with EMD according to any one of claims 1-3, characterized in that: The training process of the LSTM model includes: S11. Build a memory unit and calculate the value of the input gate; S12. Calculate the forgetting state information; S13. Update the memory unit state information and calculate the candidate value of the memory unit state; S14. Output memory unit state information and calculate output gate; S15. Repeat steps S11-S14 until the preset termination condition is met, and the training process ends.

5. Combined with EMD's LSTM main equipment, a hot oil circulation oil temperature prediction system is installed, which is characterized by: include: The data acquisition module collects the time series data of instantaneous flow rate, water content, gas content, particle size, oil filter outlet oil temperature and transformer outlet oil temperature for oil temperature prediction in the hot oil circulation link of the main equipment installation; The data decomposition module uses the EMD decomposition method to decompose the transformer outlet oil temperature characteristic time series data into multiple IMF components and one RES component; The data component screening module calculates the Pearson correlation coefficient between each decomposed component and the original time series of the transformer outlet oil temperature, and filters out the components whose correlation is lower than the preset threshold T; The oil temperature prediction module inputs the data of the transformer outlet oil temperature within a certain time range, which is decomposed by EMD and filtered, and combined with the instantaneous flow rate, water content, gas content, particle size, and oil filter outlet oil temperature, into the LSTM model for multivariate prediction.

6. An electronic device, characterized in that It includes a processor, a memory, and a computer program stored in the memory and executable on the processor. When the computer program is executed by the processor, the method for predicting the oil temperature of a hot oil circulation installed in an LSTM main device combined with an EMD as described in any one of claims 1 to 4 is implemented.

7. A computer-readable storage medium, characterized in that The computer-readable storage medium stores a computer program, and when the computer program is executed by the processor, the method for predicting the oil temperature of a hot oil circulation installed in an LSTM main device combined with EMD as described in any one of claims 1-4 is implemented.

Images