A telemetry data analysis method, device and storage medium for a substation

Abstract

The present invention relates to the field of transformer technology, and discloses a telemetry data analysis method, device and storage medium for a substation, wherein a telemetry data analysis device for a substation includes a first data collection module, which is used to collect first data; an oil temperature prediction module, which is used to output an upper limit reference value and a lower limit reference value of the oil temperature of phase A of a main transformer; an oil temperature early warning module, which is used to output the oil temperature type; and a current limiting module, which is used to control whether the main transformer is operated at a reduced capacity. The present invention calculates the upper and lower limit reference values of the phase A oil temperature within one hour, fully considering various abnormal types such as high or low oil temperature, abnormal fluctuation and abnormal trend, etc., which is helpful to discover potential transformer faults. Compared with the method of judging whether the oil temperature is abnormal by comparing the reference value with the measured value, the early warning capability is stronger.

Description

Technical Field

[0001] The present invention relates to the technical field of transformers, and more particularly to a telemetry data analysis method, device and storage medium for a transformer substation. Background Art

[0002] When a fault occurs in the main transformer, abnormal oil temperature may occur. Therefore, abnormal oil temperature is also a monitoring indicator for detecting whether there is a fault in the transformer. However, the main transformer oil temperature itself fluctuates with the transformer power, phase voltage, and current. It is impossible to compare the monitored main transformer temperature with a fixed reference value to determine whether the oil temperature is abnormal. Model A can output the current reference value based on the main transformer operating parameters and the environmental data of the main transformer. Comparison with the measured value can determine whether the oil temperature is abnormal. Abnormal oil temperature indicates a possible fault.

[0003] However, a partial fault in the transformer will not cause the measured value to exceed the reference value, for example:

[0004] Failure of the convection oil circulation system. Failure of the convection oil circulation system inside the transformer will affect the circulation of the oil, causing the local oil temperature to rise, thereby causing oil temperature fluctuations.

[0005] Heat exchange system failure: If the transformer's oil cooler, oil pump or fan and other heat exchange systems fail, it will affect the transformer's heat dissipation efficiency and cause irregular changes in oil temperature.

[0006] Insulation failure: When insulation failures such as partial discharge or local overheating occur in the transformer coil or core, although it is not enough to directly cause the overall oil temperature to exceed the standard, it will cause the local oil temperature to rise and cause oil temperature fluctuations.

[0007] The conduction of oil temperature within a transformer takes time. When a local temperature rise occurs, this temperature change must be detected by the temperature sensor through oil convection and conduction. This temperature conduction process produces a certain amount of time delay and fluctuation. Therefore, the previous method can only detect oil temperature anomalies caused by a small number of faults. Summary of the Invention

[0008] The present invention provides a telemetry data analysis method, device and storage medium for a substation, which solves the technical problem in the related art that it is only possible to judge whether the oil temperature is abnormal based on the comparison between the reference value at the current moment and the measured value, and it is difficult to detect some faults that only cause abnormal delays and fluctuations in the oil temperature.

[0009] The present invention provides a telemetry data analysis device for a substation, comprising:

[0010] The first data collection module is configured to collect first data, where the first data includes:

[0011] The first data is collected once at each time sampling point. The first data includes telemetry data and environmental data. The telemetry data covers the monitoring information of the main transformer, and the environmental data includes the average temperature, maximum temperature, minimum temperature, rainfall, wind direction, wind speed, air pressure, and humidity of the substation at each time sampling point. The time sampling point refers to the moment when data is collected at a predetermined time interval, and each time interval is one minute.

[0012] Encode the first data collected at each time sampling point into a first data vector, wherein the j-th first data vector includes the data dimensions of all parameters collected at the j-th time sampling point;

[0013] an oil temperature prediction module, which inputs the first data vector set into a prediction neural network, and the prediction neural network outputs a prediction vector;

[0014] The first data vector set is composed of the first data vectors of the p-59th to pth time sampling points, where p represents the current time sampling point;

[0015] The prediction vector includes the upper limit reference value and the lower limit reference value of the main transformer A phase oil temperature at each time sampling point from the p-59th time sampling point to the pth time sampling point;

[0016] An oil temperature early warning module is configured to concatenate the first data vector with the upper and lower reference values of the A-phase oil temperature of the main transformer at the corresponding time sampling point output by the prediction neural network to generate a second data vector, sort all the second data vectors according to the corresponding time sampling points to generate a second data vector set, input the second data vector set into the early warning neural network, and output the abnormal value of the main transformer oil temperature;

[0017] Decode the abnormal value of the main transformer oil temperature and output the oil temperature type.

[0018] Current limiting module: When the oil temperature type is abnormal, if the real-time current of the main transformer exceeds the preset current threshold, the current limiting module will issue a control instruction to reduce the capacity of the main transformer.

[0019] Furthermore, the monitoring information of the main transformer includes the active power, reactive power, power factor, current value, phase A voltage amplitude, phase B voltage amplitude, phase C voltage amplitude, phase C current amplitude of the high voltage side, medium voltage side and low voltage side of the main transformer, as well as the winding temperature and oil temperature of the three phases A, B and C.

[0020] Furthermore, if the interval between the p-th time sampling point and the time sampling point at which acquisition starts is less than 59, an interpolation operation is performed on the missing first data vector, and the interpolated first data vector is a zero vector.

[0021] Furthermore, the prediction neural network uses a long short-term memory network, which includes a forget gate, an input gate, a cell state, and an output gate. The output of the forget gate for:

[0022] ;

[0023] in, is the weight matrix of the forget gate, is the bias term of the forget gate, is the hidden state at the previous time step and the input vector of the current time step The splicing, is the Sigmoid activation function;

[0024] For the The first data vector of time sampling points;

[0025] ;

[0026] Output of the input gate and candidate cell states The calculation method is:

[0027] ;

[0028] ;

[0029] in, is the weight matrix of the input gate, is the bias term of the input gate, is the hyperbolic tangent activation function, is the weight matrix for calculating candidate cell states, is the bias term of the candidate cell state;

[0030] Cell state The previous cell state And the weighted synthesis of candidate cell states is calculated as follows:

[0031] ;

[0032] in, Represents element-wise multiplication operation;

[0033] The activation value of the output gate and the hidden state at the current time step The calculation method is:

[0034] ;

[0035] ;

[0036] in, is the weight matrix of the output gate, is the bias term of the output gate.

[0037] Furthermore, the hidden state of each time step Processing is performed, and the oil temperature upper limit reference value is calculated by two independent fully connected layers and oil temperature lower limit reference value , the calculation method is:

[0038] ;

[0039] ;

[0040] in, and are the weight matrices of the two fully connected layers, and is the bias vector.

[0041] Furthermore, using the main loss term and penalty items As a loss function , the calculation method is:

[0042] ;

[0043] ;

[0044] ;

[0045] in 、 are the upper and lower limit reference values of the oil temperature predicted at time step k, 、 are the actual upper and lower limit reference values of the oil temperature at time step k, which are derived from the highest and lowest values of the oil temperature of the main transformer under historical normal operating conditions with the same operating parameters and environmental data. The operating parameters include active power, reactive power, voltage, current and winding temperature. is the weight coefficient, and its default value is 1.

[0046] Furthermore, the early warning neural network uses a gated recurrent unit algorithm. The calculation formula of the gated recurrent unit algorithm is as follows:

[0047] ;

[0048] ;

[0049] ;

[0050] ;

[0051] ;

[0052] ;

[0053] in, and denote the 1st and Lth second data vectors of the second data vector set, respectively, 、 、 、 represents the trainable weight parameters, 、 、 、 represents the trainable bias parameter, represents the dot product, represents the Lth first intermediate feature, represents the Lth second intermediate feature, represents the Lth third intermediate feature, and Represent the L-1th and Lth output features respectively, represents the abnormal value of the main transformer oil temperature, tanh represents the tanh function, Represents the Sigmoid function.

[0054] Furthermore, the oil temperature abnormality value is a probability value. If the probability is greater than 0.5, it indicates that the oil temperature type of the main transformer is abnormal; otherwise, the oil temperature type is normal.

[0055] The present invention provides a method for analyzing telemetry data of a substation. The method for analyzing telemetry data of a substation described above performs the following steps:

[0056] Step 301: collect telemetry data from the main transformer and environmental data from the substation once every minute;

[0057] Step 302: Encode the telemetry data and environmental data of each minute into an overall first data vector;

[0058] Step 303: Combine the first data vectors for a total of 60 minutes into a first data vector set, input the first data vector set into a prediction neural network, and output a prediction vector from the prediction neural network. The prediction vector represents the upper limit reference value and the lower limit reference value of the main transformer A phase oil temperature per minute.

[0059] Step 304: Combine the first data vector set and the prediction vector into a second data vector set, and input the second data vector set into a trained early warning neural network, which outputs an abnormal value of the main transformer oil temperature.

[0060] Step 305: Decode the abnormal value of the main transformer oil temperature and output the oil temperature type. If the oil temperature type is normal, return to step 301;

[0061] Step 306 , when the oil temperature type is abnormal, if the real-time current of the main transformer exceeds the preset current threshold, the current limiting module will issue a control instruction to reduce the capacity of the main transformer, and then return to step 301 .

[0062] The present invention provides a storage medium storing non-transitory computer-readable instructions. When the non-transitory computer-readable instructions are executed by a computer, the steps of the aforementioned method for analyzing telemetry data of a substation can be executed.

[0063] The beneficial effects of the present invention are as follows: by calculating the upper and lower limit reference values of the A-phase oil temperature within one hour, the present invention fully considers various abnormal types such as high oil temperature, low oil temperature, abnormal fluctuation and abnormal trend, which helps to discover potential transformer faults. Compared with the method of judging whether the oil temperature is abnormal by comparing the reference value with the measured value, the early warning capability is stronger. BRIEF DESCRIPTION OF THE DRAWINGS

[0064] Figure 1 This is a module diagram of a telemetry data analysis device for a substation according to the present invention;

[0065] Figure 2 The present invention is a flowchart of a telemetry data analysis method for a substation.

[0066] In the figure: a first data collection module 101 , an oil temperature prediction module 102 , an oil temperature early warning module 103 , and a current limiting module 104 . DETAILED DESCRIPTION

[0067] The subject matter described herein will now be discussed with reference to exemplary embodiments. It should be understood that these embodiments are discussed solely to enable those skilled in the art to better understand and implement the subject matter described herein, and that the functions and arrangements of the elements discussed may be varied without departing from the scope of this specification. Various examples may omit, substitute, or add various processes or components as needed. In addition, features described with respect to some examples may also be combined in other examples.

[0068] At least one embodiment of the present invention discloses a telemetry data analysis device for a substation, such as Figure 1 Shown, including:

[0069] The first data collection module 101 is used to collect first data, which includes:

[0070] The first data is collected once at each time sampling point, and includes telemetry data and environmental data;

[0071] Telemetry data covers the monitoring information of transformer, including main transformer high-voltage side active power, main transformer low-voltage side active power, main transformer medium-voltage side active power, main transformer high-voltage side reactive power, main transformer low-voltage side reactive power, main transformer medium-voltage side reactive power, main transformer high-voltage side power factor, main transformer low-voltage side power factor, main transformer medium-voltage side power factor, main transformer high-voltage side current value, main transformer low-voltage side current value, main transformer medium-voltage side current value (including the current amplitude of phase A, phase B and phase C), main transformer high-voltage side phase A voltage amplitude, main transformer low-voltage side phase A voltage amplitude, main transformer medium-voltage side phase A voltage amplitude, main transformer high-voltage side phase B voltage amplitude, main transformer low-voltage side phase A voltage amplitude, main transformer medium-voltage side phase B voltage amplitude, main transformer low-voltage side phase A voltage amplitude, main transformer medium-voltage side phase B voltage amplitude, main transformer high-voltage side phase B voltage amplitude, main transformer low-voltage side phase A voltage amplitude, main transformer medium-voltage side phase A ... Phase B voltage amplitude on the main transformer low-voltage side, phase B voltage amplitude on the main transformer medium-voltage side, phase C voltage amplitude on the main transformer high-voltage side, phase C voltage amplitude on the main transformer low-voltage side, phase C voltage amplitude on the main transformer medium-voltage side, phase A winding temperature of the main transformer, phase A oil temperature of the main transformer, phase B winding temperature of the main transformer, phase B oil temperature of the main transformer, phase C winding temperature of the main transformer, and phase C oil temperature of the main transformer; environmental data includes the average temperature, maximum temperature, minimum temperature, rainfall, wind direction, wind speed, air pressure, and humidity at each time sampling point of the substation; a time sampling point refers to the moment when data is collected at a predetermined time interval, each time interval is one minute;

[0072] The first data collected at each time sampling point is encoded into a first data vector, and the j-th first data vector includes the data dimensions of all parameters collected at the j-th time sampling point.

[0073] The oil temperature prediction module 102 inputs the first data vector set into a prediction neural network, which outputs a prediction vector;

[0074] The first data vector set is composed of the first data vectors of the p-59th to pth time sampling points, where p represents the current time sampling point; if the interval between the pth time sampling point and the time sampling point where acquisition starts is less than 59, then an interpolation operation is performed on the missing first data vector, and the interpolated first data vector is a zero vector;

[0075] The prediction neural network uses a long short-term memory network, which includes a forget gate, an input gate, a cell state, and an output gate. The output of the forget gate for:

[0076]

[0077] in, is the weight matrix of the forget gate, is the bias term of the forget gate, is the hidden state at the previous time step and the input vector of the current time step The splicing, is the Sigmoid activation function;

[0078] For the The first data vector of time sampling points;

[0079]

[0080] Output of the input gate and candidate cell states The calculation method is:

[0081]

[0082]

[0083] in, is the weight matrix of the input gate, is the bias term of the input gate, is the hyperbolic tangent activation function, is the weight matrix for calculating candidate cell states, is the bias term of the candidate cell state;

[0084] Cell state The previous cell state And the weighted synthesis of candidate cell states is calculated as follows:

[0085]

[0086] in, Represents element-wise multiplication operation;

[0087] The activation value of the output gate and the hidden state at the current time step The calculation method is:

[0088]

[0089]

[0090] in, is the weight matrix of the output gate, is the bias term of the output gate;

[0091] The hidden state for each time step Processing is performed, and the oil temperature upper limit reference value is calculated by two independent fully connected layers and oil temperature lower limit reference value , the calculation method is:

[0092]

[0093]

[0094] in, and are the weight matrices of the two fully connected layers, and is the bias vector;

[0095] Using the main loss term and penalty items As a loss function , the calculation method is:

[0096]

[0097]

[0098]

[0099] in 、 are the upper and lower limit reference values of the oil temperature predicted at time step k, 、 are the actual upper and lower limit reference values of the oil temperature at time step k, which are derived from the highest and lowest values of the oil temperature of the main transformer under historical normal operating conditions with the same operating parameters and environmental data. The operating parameters include active power, reactive power, voltage, current and winding temperature. is the weight coefficient, and its default value is 1.

[0100] The prediction vector includes the upper limit reference value and the lower limit reference value of the A-phase oil temperature of the main transformer at each time sampling point from the p-59th time sampling point to the pth time sampling point.

[0101] The oil temperature early warning module 103 is configured to concatenate the first data vector with the upper and lower reference values of the phase A oil temperature of the main transformer at the corresponding time sampling point output by the prediction neural network to generate a second data vector, sort all the second data vectors according to the corresponding time sampling points to generate a second data vector set, input the second data vector set into the early warning neural network, and output the abnormal value of the main transformer oil temperature;

[0102] In one embodiment of the present invention, the method of splicing the first data vector with the upper limit reference value and lower limit reference value of the main transformer A phase oil temperature at the corresponding time sampling point output by the prediction neural network is: splicing the upper limit reference value and lower limit reference value of the main transformer A phase oil temperature after the first data vector.

[0103] The early warning neural network uses the gated recurrent unit algorithm. The calculation formula of the gated recurrent unit algorithm is as follows:

[0104]

[0105]

[0106]

[0107]

[0108]

[0109]

[0110] in, and denote the 1st and Lth second data vectors of the second data vector set, respectively, 、 、 、 represents the trainable weight parameters, 、 、 、 represents the trainable bias parameter, represents the dot product, represents the Lth first intermediate feature, represents the Lth second intermediate feature, represents the Lth third intermediate feature, and Represent the L-1th and Lth output features respectively, represents the abnormal value of the main transformer oil temperature, tanh represents the tanh function, Represents the Sigmoid function;

[0111] In one embodiment of the present invention, is a probability value. If the probability is greater than 0.5, it means that the oil temperature type of the main transformer is abnormal; otherwise, the oil temperature type is normal.

[0112] The current limiting module 104, when the oil temperature type is abnormal, if the real-time current of the main transformer exceeds the preset current threshold, the current limiting module will issue a control instruction to reduce the capacity of the main transformer.

[0113] In one embodiment of the present invention, a method for analyzing telemetry data of a substation is provided. Figure 2 As shown, the following steps are included:

[0114] Step 301: collect telemetry data from the main transformer and environmental data from the substation once every minute;

[0115] Step 302: Encode the telemetry data and environmental data of each minute into an overall first data vector;

[0116] Step 303: Combine the first data vectors for a total of 60 minutes into a first data vector set, input the first data vector set into a prediction neural network, and output a prediction vector from the prediction neural network. The prediction vector represents the upper limit reference value and the lower limit reference value of the main transformer A phase oil temperature per minute.

[0117] Step 304: Combine the first data vector set and the prediction vector into a second data vector set, and input the second data vector set into a trained early warning neural network, which outputs an abnormal value of the main transformer oil temperature.

[0118] Step 305: Decode the abnormal value of the main transformer oil temperature and output the oil temperature type. If the oil temperature type is normal, return to step 301;

[0119] Step 306 , when the oil temperature type is abnormal, if the real-time current of the main transformer exceeds the preset current threshold, the current limiting module will issue a control instruction to reduce the capacity of the main transformer, and then return to step 301 .

[0120] At least one embodiment of the present invention provides a storage medium storing non-transitory computer-readable instructions. When the non-transitory computer-readable instructions are executed by a computer, the steps of the aforementioned method for analyzing telemetry data of a substation can be executed.

[0121] The above describes an embodiment of the present invention, but this embodiment is not limited to the above-mentioned specific implementation methods. The above-mentioned specific implementation methods are merely illustrative and not restrictive. Ordinary technicians in this field can also make more forms of equivalent embodiments based on the inspiration of this embodiment, all of which are protected by this embodiment.

Claims

1. A telemetry data analysis device for a substation, characterized in that: include: The first data collection module is configured to collect first data, where the first data includes: The first data is collected once at each time sampling point. The first data includes telemetry data and environmental data. The telemetry data covers the monitoring information of the main transformer, and the environmental data includes the average temperature, maximum temperature, minimum temperature, rainfall, wind direction, wind speed, air pressure, and humidity of the substation at each time sampling point. The time sampling point refers to the moment when data is collected at a predetermined time interval, and each time interval is one minute. The monitoring information of the main transformer includes the active power, reactive power, power factor, current value, phase A voltage amplitude, phase B voltage amplitude, phase C voltage amplitude, phase C current amplitude of the high-voltage side, medium-voltage side and low-voltage side of the main transformer, as well as the winding temperature and oil temperature of the three phases A, B and C; Encode the first data collected at each time sampling point into a first data vector, wherein the j-th first data vector includes the data dimensions of all parameters collected at the j-th time sampling point; an oil temperature prediction module, which inputs the first data vector set into a prediction neural network, and the prediction neural network outputs a prediction vector; The first data vector set is composed of the first data vectors of the p-59th to pth time sampling points, where p represents the current time sampling point; The prediction vector includes the upper limit reference value and the lower limit reference value of the main transformer A phase oil temperature at each time sampling point from the p-59th time sampling point to the pth time sampling point; An oil temperature early warning module is configured to concatenate the first data vector with the upper and lower reference values of the A-phase oil temperature of the main transformer at the corresponding time sampling point output by the prediction neural network to generate a second data vector, sort all the second data vectors according to the corresponding time sampling points to generate a second data vector set, input the second data vector set into the early warning neural network, and output the abnormal value of the main transformer oil temperature; Decode abnormal oil temperature value of main transformer and output oil temperature type; Current limiting module: When the oil temperature type is abnormal, if the real-time current of the main transformer exceeds the preset current threshold, the current limiting module will issue a control instruction to reduce the capacity of the main transformer; The substation's telemetry data, oil temperature fluctuations, and delay anomalies are optimized through a predictive neural network model using LSTM or GRU. The substation's telemetry data is collaboratively analyzed by designing a main transformer partitioning module, a data collection module, an oil temperature prediction module, and an oil temperature warning module. The warning module and gated recurrent unit algorithm are combined to further process the timing characteristics of oil temperature fluctuations and provide warnings. When data is missing, dynamic interpolation is used to fill in the gaps, enabling the model to accurately process telemetry data. Predictive neural network model using the main loss term and penalty items As a loss function , the calculation method is: ； ； ； in 、 are the upper and lower limit reference values of the oil temperature predicted at time step k, 、 are the actual upper and lower limit reference values of the oil temperature at time step k, which are derived from the highest and lowest values of the oil temperature of the main transformer under historical normal operating conditions with the same operating parameters and environmental data. The operating parameters include active power, reactive power, voltage, current and winding temperature. is the weight coefficient, and its default value is 1.

2. The telemetry data analysis device for a substation according to claim 1, characterized in that: If the interval between the p-th time sampling point and the time sampling point at which acquisition starts is less than 59, an interpolation operation is performed on the missing first data vector, and the interpolated first data vector is a zero vector.

3. The telemetry data analysis device for a substation according to claim 1, characterized in that: The prediction neural network uses a long short-term memory network, which includes a forget gate, an input gate, a cell state, and an output gate. The output of the forget gate for: ； in, is the weight matrix of the forget gate, is the bias term of the forget gate, is the hidden state at the previous time step and the input vector of the current time step The splicing, is the Sigmoid activation function; For the The first data vector of time sampling points; ； Output of the input gate and candidate cell states The calculation method is: ； ； in, is the weight matrix of the input gate, is the bias term of the input gate, is the hyperbolic tangent activation function, is the weight matrix for calculating candidate cell states, is the bias term of the candidate cell state; Cell state The previous cell state And the weighted synthesis of candidate cell states is calculated as follows: ； in, Represents element-wise multiplication operation; The activation value of the output gate and the hidden state at the current time step The calculation method is: ； ； in, is the weight matrix of the output gate, is the bias term of the output gate.

4. The telemetry data analysis device for a substation according to claim 3, characterized in that: The hidden state for each time step Processing is performed, and the oil temperature upper limit reference value is calculated by two independent fully connected layers and oil temperature lower limit reference value , the calculation method is: ； ； in, and are the weight matrices of the two fully connected layers, and is the bias vector.

5. The telemetry data analysis device for a substation according to claim 1, characterized in that: The early warning neural network uses the gated recurrent unit algorithm. The calculation formula of the gated recurrent unit algorithm is as follows: ； ； ； ； ； ； in, and denote the 1st and Lth second data vectors of the second data vector set, respectively, represents the trainable weight parameters, represents the trainable bias parameter, represents the dot product, represents the Lth first intermediate feature, represents the Lth second intermediate feature, represents the Lth third intermediate feature, and Represent the L-1th and Lth output features respectively, represents the abnormal value of the main transformer oil temperature, tanh represents the tanh function, Represents the Sigmoid function.

6. The telemetry data analysis device for a substation according to claim 1, characterized in that: The oil temperature abnormality value is a probability value. If the probability is greater than 0.5, it means that the main transformer oil temperature type is abnormal; otherwise, the oil temperature type is normal.

7. A telemetry data analysis method for a substation, characterized in that: The telemetry data analysis device for a substation according to any one of claims 1 to 6 performs the following steps: Step 301: collect telemetry data from the main transformer and environmental data from the substation once every minute; Step 302: Encode the telemetry data and environmental data of each minute into an overall first data vector; Step 303: Combine the first data vectors for a total of 60 minutes into a first data vector set, input the first data vector set into a prediction neural network, and output a prediction vector from the prediction neural network. The prediction vector represents the upper limit reference value and the lower limit reference value of the main transformer A phase oil temperature per minute. Step 304: Combine the first data vector set and the prediction vector into a second data vector set, and input the second data vector set into a trained early warning neural network, which outputs an abnormal value of the main transformer oil temperature. Step 305: Decode the abnormal value of the main transformer oil temperature and output the oil temperature type. If the oil temperature type is normal, return to step 301; Step 306 , when the oil temperature type is abnormal, if the real-time current of the main transformer exceeds the preset current threshold, the current limiting module will issue a control instruction to reduce the capacity of the main transformer, and then return to step 301 .

8. A storage medium, characterized in that: The non-transitory computer-readable instructions are stored therein. When the non-transitory computer-readable instructions are executed by a computer, the steps of the telemetry data analysis method for a substation as claimed in claim 7 can be executed.

Images