Transformer oil temperature prediction method, device, computer equipment and storage medium

Abstract

The present application relates to a transformer oil temperature prediction method, device, computer equipment, storage medium and computer program product. By decomposing a known oil temperature sequence based on an empirical mode decomposition algorithm, a plurality of intrinsic modal components and residual components corresponding to the known oil temperature sequence are obtained, and a target long-short-term memory artificial neural network outputs a predicted intrinsic modal component containing a plurality of first prediction components and a predicted residual component containing a plurality of second prediction components based on the plurality of intrinsic modal components and residual components, and obtains the oil temperature prediction result of the transformer to be predicted according to the plurality of predicted intrinsic modal components and the predicted residual components. Compared with the traditional method of prediction through a fuzzy model, this scheme decomposes the known oil temperature sequence into a plurality of intrinsic modal components and residual components through an empirical mode decomposition algorithm, and predicts each intrinsic modal component and residual component based on a neural network to obtain the oil temperature prediction result of the transformer to be predicted, thereby improving the efficiency of oil temperature prediction.

Description

Technical Field

[0001] The present application relates to the field of electric power technology, and in particular to a transformer oil temperature prediction method, device, computer equipment, storage medium, and computer program product. Background Art

[0002] Power transformers are critical components of power systems, and their stable operation significantly impacts system reliability and economic efficiency. To a certain extent, a transformer's load capacity and insulation aging rate are primarily determined by its internal thermal state. Transformer top oil temperature (TOT) and winding hotspot temperature are key parameters for assessing this thermal state. Therefore, while ensuring safe operation, it is crucial to fully utilize the transformer's dynamic load capacity and accurately predict its winding hotspot temperature. Currently, transformer winding hotspot temperature is typically determined by monitoring the top oil temperature. Specifically, this is achieved by using the TS model to predict oil temperature. However, using the TS model increases the complexity of oil temperature prediction due to its complex application.

[0003] Therefore, the current transformer oil temperature prediction method has the defect of high prediction complexity. Summary of the Invention

[0004] Based on this, it is necessary to provide a transformer oil temperature prediction method, device, computer equipment, computer-readable storage medium and computer program product that can reduce the prediction complexity in order to address the above technical problems.

[0005] In a first aspect, the present application provides a transformer oil temperature prediction method, the method comprising:

[0006] Obtaining a known oil temperature sequence of the transformer to be predicted within a preset time period;

[0007] Decomposing the known oil temperature sequence according to an empirical mode decomposition algorithm to obtain a plurality of intrinsic mode components and residual components corresponding to the known oil temperature sequence;

[0008] The target long short-term memory artificial neural network outputs a first prediction component corresponding to each intrinsic modal component and a second prediction component corresponding to the residual component based on the multiple intrinsic modal components and the residual component, thereby obtaining a predicted intrinsic modal component corresponding to each intrinsic modal component and a predicted residual component corresponding to the residual component; the predicted intrinsic modal component includes the first prediction components at multiple time points, and the predicted residual component includes the second prediction components at multiple time points;

[0009] The oil temperature prediction result of the transformer to be predicted is obtained according to the multiple prediction intrinsic modal components and the prediction residual component.

[0010] In one embodiment, decomposing the known oil temperature sequence according to an empirical mode decomposition algorithm to obtain a plurality of intrinsic mode components and residual components corresponding to the known oil temperature sequence includes:

[0011] Acquire multiple maximum value points and multiple minimum value points in the known oil temperature sequence;

[0012] Determining a first envelope curve of the known oil temperature sequence according to the multiple maximum value points, and determining a second envelope curve of the known oil temperature sequence according to the multiple minimum value points;

[0013] Obtaining mean envelope curves corresponding to the first envelope curve and the second envelope curve;

[0014] Obtaining the difference between the known oil temperature sequence and the mean envelope curve to obtain a difference oil temperature sequence;

[0015] If the differential oil temperature sequence meets a preset component condition, determining that the differential oil temperature sequence is an intrinsic modal component, separating the intrinsic modal component from the known oil temperature sequence to obtain a new known oil temperature sequence;

[0016] If the number of the intrinsic modal components is less than a preset numerical threshold, obtaining a plurality of maximum points and a plurality of minimum points in the new known oil temperature sequence, and returning to the step of determining a first envelope curve of the known oil temperature sequence based on the plurality of maximum points, and determining a second envelope curve of the known oil temperature sequence based on the plurality of minimum points;

[0017] If the number of the intrinsic modal components is greater than or equal to a preset numerical threshold, multiple intrinsic modal components corresponding to the known oil temperature sequence are obtained based on the current multiple intrinsic modal components, and the new known oil temperature sequence is used as the residual component.

[0018] In one embodiment, after obtaining the difference between the known oil temperature sequence and the mean envelope curve to obtain the difference oil temperature sequence, the method further includes:

[0019] If the difference oil temperature sequence does not meet the preset component condition, obtaining multiple maximum value points and multiple minimum value points in the difference oil temperature sequence;

[0020] Determining a difference envelope curve according to a plurality of maximum value points and a plurality of minimum value points in the difference oil temperature sequence, and obtaining a mean envelope curve of the difference envelope curve;

[0021] obtaining a new differential oil temperature sequence based on the difference between the differential oil temperature sequence and the mean envelope curve; if the new differential oil temperature sequence does not meet the preset component condition, obtaining multiple maximum points and multiple minimum points in the new differential oil temperature sequence, and returning to the step of determining the differential envelope curve based on the multiple maximum points and multiple minimum points in the differential oil temperature sequence;

[0022] If the new differential oil temperature sequence meets a preset component condition, it is determined that the new differential oil temperature sequence is an intrinsic modal component.

[0023] In one embodiment, the method further comprises:

[0024] Acquire multiple sample connotation modal components and sample residual components corresponding to the sample oil temperature sequence, and obtain a long short-term memory artificial neural network to be trained; the long short-term memory artificial neural network to be trained includes an input gate, a forget gate, and an output gate;

[0025] For each sample connotation modal component, the sample connotation modal component is divided into an input sample connotation modal component and an output sample connotation modal component; the time corresponding to the input sample connotation modal component is shorter than the time corresponding to the output sample connotation modal component;

[0026] Inputting the input sample connotation modal component into the long short-term memory artificial neural network to be trained, the long short-term memory artificial neural network to be trained sequentially predicts the input sample connotation modal component through an input gate, a forget gate, and an output gate, and outputs a predicted sample connotation modal component corresponding to the input sample connotation modal component; the predicted sample connotation modal component represents the connotation modal component after the time corresponding to the input sample connotation modal component;

[0027] According to the comparison result of the predicted sample connotation modal component and the output sample connotation modal component, the model parameters of the long short-term memory artificial neural network to be trained are adjusted until the training conditions are met, thereby obtaining the trained target long short-term memory artificial neural network.

[0028] In one embodiment, the target long short-term memory artificial neural network outputs a first prediction component corresponding to each connotation modal component and a second prediction component corresponding to the residual component based on the multiple connotation modal components and the residual component, to obtain a predicted connotation modal component corresponding to each connotation modal component and a predicted residual component corresponding to the residual component, including:

[0029] For each connotation modal component, the connotation modal component is input into a target long short-term memory artificial neural network, and the target long short-term memory artificial neural network sequentially passes the connotation modal component through an input gate, a forget gate, and an output gate, and outputs a plurality of first predicted components after the time of the connotation modal component, to obtain a predicted connotation modal component corresponding to the connotation modal component;

[0030] The residual component is input into the target long short-term memory artificial neural network, and the target long short-term memory artificial neural network passes the residual component through the input gate, the forgetting gate and the output gate in sequence and outputs multiple second prediction components after the time of the residual component to obtain the predicted residual component corresponding to the residual component.

[0031] In one embodiment, obtaining the prediction result of the oil temperature of the transformer to be predicted based on the multiple predicted intrinsic modal components and the residual component includes:

[0032] The oil temperature prediction result of the transformer to be predicted is obtained according to the sum of the multiple predicted intrinsic modal components and the residual component.

[0033] In a second aspect, the present application provides a transformer oil temperature prediction device, the device comprising:

[0034] An acquisition module is used to obtain a known oil temperature sequence of the transformer to be predicted within a preset time period;

[0035] a decomposition module, configured to decompose the known oil temperature sequence according to an empirical mode decomposition algorithm to obtain a plurality of intrinsic mode components and residual components corresponding to the known oil temperature sequence;

[0036] a prediction module configured to output, by a target long short-term memory artificial neural network, a first prediction component corresponding to each connotational modal component and a second prediction component corresponding to the residual component based on the multiple connotational modal components and the residual component, to obtain a predicted connotational modal component corresponding to each connotational modal component and a predicted residual component corresponding to the residual component; the predicted connotational modal component includes the first prediction components at multiple time points, and the predicted residual component includes the second prediction components at multiple time points;

[0037] The reconstruction module is used to obtain the oil temperature prediction result of the transformer to be predicted based on the multiple predicted intrinsic modal components and the residual component.

[0038] In a third aspect, the present application provides a computer device comprising a memory and a processor, wherein the memory stores a computer program, and the processor implements the steps of the above method when executing the computer program.

[0039] In a fourth aspect, the present application provides a computer-readable storage medium having a computer program stored thereon, which implements the steps of the above method when executed by a processor.

[0040] In a fifth aspect, the present application provides a computer program product, comprising a computer program, which implements the steps of the above method when executed by a processor.

[0041] The above-mentioned transformer oil temperature prediction method, apparatus, computer equipment, storage medium, and computer program product decompose the known oil temperature sequence based on the empirical mode decomposition algorithm to obtain multiple intrinsic modal components and residual components corresponding to the known oil temperature sequence. The target long-short-term memory artificial neural network outputs a predicted intrinsic modal component containing multiple first prediction components and a predicted residual component containing multiple second prediction components based on the multiple intrinsic modal components and residual components, and obtains the oil temperature prediction result of the transformer to be predicted based on the multiple predicted intrinsic modal components and predicted residual components. Compared with the traditional method of prediction through fuzzy models, this solution decomposes the known oil temperature sequence into multiple intrinsic modal components and residual components through the empirical mode decomposition algorithm, and predicts each intrinsic modal component and residual component based on the neural network to obtain the oil temperature prediction result of the transformer to be predicted, thereby improving the efficiency of oil temperature prediction. BRIEF DESCRIPTION OF THE DRAWINGS

[0042] Figure 1 1 is a flow chart of a transformer oil temperature prediction method according to an embodiment;

[0043] Figure 2 A schematic diagram of the structure of a long short-term memory artificial neural network in one embodiment;

[0044] Figure 3 A schematic flow chart of a transformer oil temperature prediction method in another embodiment;

[0045] Figure 4 A schematic diagram of a known oil temperature sequence in one embodiment;

[0046] Figure 5 A schematic diagram of the structure of the decomposition steps in one embodiment;

[0047] Figure 6 A schematic structural diagram of the frequencies of the decomposed components in one embodiment;

[0048] Figure 7 A schematic diagram of a prediction result in one embodiment;

[0049] Figure 8 1 is a structural block diagram of a transformer oil temperature prediction device in one embodiment;

[0050] Figure 9FIG. 1 is a diagram showing the internal structure of a computer device in one embodiment. DETAILED DESCRIPTION

[0051] In order to make the purpose, technical solutions and advantages of this application more clear, the following further describes this application in detail with reference to the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are only used to explain this application and are not intended to limit this application.

[0052] In one embodiment, Figure 1 As shown, a transformer oil temperature prediction method is provided. This embodiment uses the method applied to a terminal as an example for illustration. It is understandable that the method can also be applied to a server, and can also be applied to a system including a terminal and a server, and is implemented through interaction between the terminal and the server, including the following steps:

[0053] Step S202: obtaining a known oil temperature sequence of the transformer to be predicted within a preset time period.

[0054] The transformer to be predicted can be an electrical device requiring oil temperature prediction. The terminal can predict the transformer's oil temperature and thereby obtain the hotspot temperature of the transformer's windings. The terminal can record the oil temperature of the transformer to be predicted at all times, obtaining chronologically ordered oil temperature data. These data can form a known oil temperature sequence. That is, the terminal can sort the historically recorded oil temperature data chronologically to obtain a known oil temperature sequence. The terminal can obtain the known oil temperature sequence of the transformer to be predicted within a preset time period, and thus the terminal can predict the transformer's oil temperature based on this known oil temperature sequence.

[0055] In step S204 , the known oil temperature sequence is decomposed according to an empirical mode decomposition algorithm to obtain a plurality of intrinsic mode components and residual components corresponding to the known oil temperature sequence.

[0056] Among them, the Empirical Mode Decomposition (EMD) method is a signal analysis method that decomposes signals based on the time scale characteristics of the data itself, without the need to predefine any basis functions. It is suitable for analyzing nonlinear and non-stationary signal sequences and has a high signal-to-noise ratio. The terminal can decompose the known oil temperature sequence using the EMD algorithm to obtain multiple IMF (Intrinsic Mode Function) components and a residual component R corresponding to the known oil temperature sequence. The IMF components are the signal components of each layer obtained after the original signal is decomposed by EMD. Any signal can be split into the sum of several intrinsic mode components. After the terminal decomposes the known oil temperature sequence into multiple IMF components and residual components, it can obtain deeper time series characteristics through a prediction model. The IMF components are the components formed by dividing the known oil temperature sequence into components of different time lengths after appropriate processing, and the residual components are the sequences remaining after the known oil temperature sequence is segmented.

[0057] Specifically, since the oil temperature data of power transformers has the characteristics of large scale and low density, before oil temperature prediction, it is necessary to use the EMD algorithm to decompose the oil temperature data to obtain several IMF components and a residual component R, so that the prediction model can obtain deeper time series characteristics.

[0058] In step S206, the target long short-term memory artificial neural network outputs a first prediction component corresponding to each intrinsic modal component and a second prediction component corresponding to the residual component based on multiple intrinsic modal components and residual components, thereby obtaining a predicted intrinsic modal component corresponding to each intrinsic modal component and a predicted residual component corresponding to the residual component; the predicted intrinsic modal component includes the first prediction components of multiple time points, and the predicted residual component includes the second prediction components of multiple time points.

[0059] The LSTM (Long Short-Term Memory) artificial neural network is a time-recurrent neural network specifically designed to address the long-term dependency issues inherent in typical RNNs (recurrent neural networks). All RNNs have a chain-like structure of repetitive neural network modules. There are multiple intrinsic modal components. The terminal can use the target LSTM artificial neural network to output a first prediction component corresponding to each intrinsic modal component and a second prediction component corresponding to the residual component based on the multiple intrinsic modal components and residual components corresponding to the known oil temperature sequence. This allows the terminal to obtain a predicted intrinsic modal component corresponding to each intrinsic modal component and a predicted residual component corresponding to the residual component. The predicted intrinsic modal components include first prediction components at multiple time points, and the predicted residual components include second prediction components at multiple time points. In other words, the terminal can use the target LSTM artificial neural network to predict each intrinsic modal component, predicting the prediction component after the last moment of that intrinsic modal component. One intrinsic modal component can correspond to multiple prediction components, resulting in a predicted intrinsic modal component that includes the first prediction components at multiple time points. The terminal can also predict the residual component through the target long short-term memory artificial neural network, predicting the predicted component after the last moment of the residual component. The residual component can correspond to multiple predicted components, and the terminal can obtain the predicted residual component of the second predicted component including multiple time points.

[0060] Step S208: Obtain the prediction result of the oil temperature of the transformer to be predicted based on the multiple predicted intrinsic modal components and the predicted residual components.

[0061] Since there are multiple intrinsic modal components, the terminal can predict multiple intrinsic modal components using LSTM. The terminal can obtain a prediction result for the oil temperature of the transformer to be predicted based on the multiple predicted intrinsic modal components and the prediction residual components.

[0062] Specifically, the terminal can determine the oil temperature prediction result of the transformer to be predicted by summing. For example, in one embodiment, the oil temperature prediction result of the transformer to be predicted is obtained according to a plurality of predicted intrinsic modal components and residual components, including: obtaining the oil temperature prediction result of the transformer to be predicted according to the sum of a plurality of predicted intrinsic modal components and residual components. In this embodiment, the terminal can obtain the sum of the above-mentioned multiple predicted intrinsic modal components, and obtain the sum of the sum of the multiple predicted intrinsic modal components and the above-mentioned residual components, and then obtain the final predicted oil temperature sequence as the oil temperature prediction result of the transformer to be predicted. Specifically, each of the above-mentioned predicted intrinsic modal components represents the components of the original sequence at different frequencies. After predicting the data of each component point by point, the terminal merges the data of different frequencies to obtain the final predicted oil temperature sequence. The predicted intrinsic modal component is also called IMFS. The terminal can let the predicted intrinsic modal component be P mfs(i) , the predicted residual component is P r The terminal can reconstruct multiple predicted intrinsic modal components and residual components to obtain the oil temperature prediction result. The specific formula is as follows:

[0063] Among them, P mfs(i) (t) represents the predicted data of the i-th IMFS at time t; P r (t) represents the predicted data of the residual trend term at time t; P represents the final predicted sequence.

[0064] In the above-mentioned transformer oil temperature prediction method, the known oil temperature sequence is decomposed based on the empirical mode decomposition algorithm to obtain multiple intrinsic modal components and residual components corresponding to the known oil temperature sequence. The target long-short-term memory artificial neural network outputs a predicted intrinsic modal component containing multiple first prediction components and a predicted residual component containing multiple second prediction components based on the multiple intrinsic modal components and residual components. The predicted oil temperature of the transformer to be predicted is obtained based on the multiple predicted intrinsic modal components and the predicted residual components. Compared with the traditional method of prediction through fuzzy models, this scheme decomposes the known oil temperature sequence into multiple intrinsic modal components and residual components through the empirical mode decomposition algorithm, and predicts each intrinsic modal component and residual component based on the neural network to obtain the oil temperature prediction result of the transformer to be predicted, thereby improving the efficiency of oil temperature prediction.

[0065] In one embodiment, a known oil temperature sequence is decomposed according to an empirical mode decomposition algorithm to obtain multiple intrinsic modal components and residual components corresponding to the known oil temperature sequence, including: obtaining multiple maximum points and multiple minimum points in the known oil temperature sequence; determining a first envelope curve of the known oil temperature sequence according to the multiple maximum points, and determining a second envelope curve of the known oil temperature sequence according to the multiple minimum points; obtaining a mean envelope curve corresponding to the first envelope curve and the second envelope curve; obtaining a difference between the known oil temperature sequence and the mean envelope curve to obtain a difference oil temperature sequence; if the difference oil temperature sequence meets a preset component condition, determining the difference oil temperature sequence is Intrinsic modal components, separating the intrinsic modal components in the known oil temperature sequence to obtain a new known oil temperature sequence; if the number of intrinsic modal components is less than a preset numerical threshold, obtaining multiple maximum points and multiple minimum points in the new known oil temperature sequence, and returning to the step of determining a first envelope curve of the known oil temperature sequence according to the multiple maximum points, and determining a second envelope curve of the known oil temperature sequence according to the multiple minimum points; if the number of intrinsic modal components is greater than or equal to the preset numerical threshold, obtaining multiple intrinsic modal components corresponding to the known oil temperature sequence according to the current multiple intrinsic modal components, and using the new known oil temperature sequence as a residual component.

[0066] In this embodiment, the terminal can decompose the known oil temperature sequence using an EMD algorithm. The EMD algorithm includes the following steps: determining local maximum and minimum points in the time series; connecting all maximum and minimum points using cubic spline interpolation to determine the upper and lower envelopes of the signal; averaging the maximum and minimum envelopes to obtain a mean envelope; and calculating IMF components and residual components. The terminal can obtain multiple maximum and minimum points in the known oil temperature sequence and determine a first envelope curve for the known oil temperature sequence based on the multiple maximum points and a second envelope curve for the known oil temperature sequence based on the multiple minimum points. The terminal can obtain the mean envelope curve corresponding to the first and second envelope curves and obtain the difference between the known oil temperature sequence and the mean envelope curve to obtain a difference oil temperature sequence. The terminal can then determine whether the difference oil temperature sequence meets a preset component condition, which can be determined based on the number of extreme points and zero crossings in the sequence. If the terminal determines that the above-mentioned differential oil temperature sequence meets the preset component conditions, the terminal determines that the differential oil temperature sequence is an intrinsic modal component, separates the intrinsic modal component in the known oil temperature sequence, and obtains a new known oil temperature sequence. The terminal can also obtain the number of currently existing intrinsic modal components and the length of the above-mentioned known oil temperature sequence. If the terminal detects that the number of currently existing intrinsic modal components is less than the preset numerical threshold and the length of the known oil temperature sequence is greater than the preset length threshold, the terminal determines that the above-mentioned new known oil temperature sequence can continue to be decomposed, so that the terminal can obtain multiple maximum points and multiple minimum points in the new known oil temperature sequence, and returns to the step of determining the first envelope curve of the known oil temperature sequence based on the multiple maximum points, and determining the second envelope curve of the known oil temperature sequence based on the multiple minimum points.

[0067] If the terminal detects that the number of intrinsic modal components is greater than or equal to a preset numerical threshold or the length of the known oil temperature sequence is less than or equal to a preset length threshold, the terminal determines that the above-mentioned known oil temperature sequence cannot be further decomposed. The terminal can then obtain multiple intrinsic modal components corresponding to the known oil temperature sequence based on the current multiple intrinsic modal components, and use the new known oil temperature sequence as the residual component.

[0068] Specifically, the EMD decomposition process is based on the following assumptions: the signal being analyzed has at least one maximum and one minimum, the time interval between extremes is a characteristic time dimension, and if the signal data has no extreme points but only inflection points, the extreme points can be obtained by differentiating the data once or several times, and then the decomposition result is obtained by integration. Due to the inherent complexity of oil temperature data, many different oscillation modes coexist at the same time. The terminal can first determine the mean envelope from the known oil temperature series mentioned above. The formula for the mean envelope can be as follows:

[0069] Among them, emax1 (t) is the upper envelope formed by connecting all the maximum points of the known oil temperature sequence through cubic spline interpolation, e min1 (t) is the lower envelope formed by connecting all minimum points of the known oil temperature sequence through cubic spline interpolation, and m1(t) is the average envelope obtained by the maximum envelope and the minimum envelope, that is, the above-mentioned mean envelope.

[0070] After the terminal obtains the mean envelope, it can obtain the difference oil temperature sequence, and its formula can be as follows: h1(t) = x(t) - m1(t). Among them, x(t) represents the original sequence, that is, the known oil temperature sequence mentioned above; m1(t) represents the average envelope obtained by the maximum envelope and the minimum envelope. The original sequence can be subtracted from the envelope value point by point, and a new sequence is formed after subtraction. The value of each point represents the deviation of the point from the mean; h1(t) represents the difference oil temperature sequence. After the terminal obtains the difference oil temperature sequence, it can determine whether the difference oil temperature sequence meets the preset component conditions. If it does, the terminal can use the above h1(t) as the first IMF component c1(t), regarded as IMF1, and the terminal can separate the above c1(t) from the above x(t), and use the remainder r1(t) = x(t) - c1(t) as the new known oil temperature sequence, and repeat the above steps to obtain IMF2, IMF3... and other intrinsic modal components in turn until c n (t) or r n (t) When the given termination condition is met, the decomposition process ends. At this point, the terminal can obtain multiple intrinsic modal components and residual components, and each intrinsic modal component can be recorded as c i (t), the residual component can be recorded as r(t). Then the above known oil temperature series can be expressed as the sum of IMF and remainder, that is:

[0071] Among them, c i (t) is the i-th IMF component decomposed from x(t), and r(t) represents the residual residual decomposed from x(t). Each IMF component represents a component at a different frequency, and the terminals can be sorted in order from high frequency to low frequency.

[0072] Through this embodiment, the terminal can decompose the known oil temperature sequence multiple times based on the envelope curve and EMD algorithm to obtain multi-layer IMF components and residual components, so that the terminal can predict the oil temperature based on the multi-layer IMF components and residual components, thereby improving the efficiency of oil temperature prediction.

[0073] In one embodiment, after obtaining the difference between a known oil temperature sequence and a mean envelope curve and obtaining the difference oil temperature sequence, the method further includes: if the difference oil temperature sequence does not meet the preset component condition, obtaining multiple maximum points and multiple minimum points in the difference oil temperature sequence; determining the difference envelope curve based on the multiple maximum points and multiple minimum points in the difference oil temperature sequence, and obtaining the mean envelope curve of the difference envelope curve; obtaining a new difference oil temperature sequence based on the difference between the difference oil temperature sequence and the mean envelope curve, and if the new difference oil temperature sequence does not meet the preset component condition, obtaining multiple maximum points and multiple minimum points in the new difference oil temperature sequence, and returning to the step of determining the difference envelope curve based on the multiple maximum points and multiple minimum points in the difference oil temperature sequence; if the new difference oil temperature sequence meets the preset component condition, determining the new difference oil temperature sequence as an intrinsic modal component.

[0074] In this embodiment, after obtaining the above-mentioned differential oil temperature sequence, the terminal also needs to determine whether the differential oil temperature sequence can be used as an intrinsic modal component. If the terminal detects that the above-mentioned differential oil temperature sequence does not meet the preset component condition, the terminal can treat the differential oil temperature sequence as the original sequence and obtain multiple maximum points and multiple minimum points in the differential oil temperature sequence; determine a differential envelope curve based on the multiple maximum points and multiple minimum points in the differential oil temperature sequence, and obtain a mean envelope curve of the differential envelope curve; the terminal can obtain a new differential oil temperature sequence based on the difference between the differential oil temperature sequence and the mean envelope curve. If the new differential oil temperature sequence does not meet the preset component condition, the terminal can obtain multiple maximum points and multiple minimum points in the new differential oil temperature sequence and return to the step of determining the differential envelope curve based on the multiple maximum points and multiple minimum points in the differential oil temperature sequence; if the terminal detects that the new differential oil temperature sequence meets the preset component condition, it determines that the new differential oil temperature sequence is an intrinsic modal component.

[0075] Specifically, EMD is based on the local characteristic scale of the data itself and can parse out these intrinsic patterns, namely, the intrinsic pattern components IMF, from the original data. These IMFs need to meet two conditions: in the entire data sequence, the number of extreme points and the number of zero-crossing points must be equal or differ by at most 1; and at any point, the upper envelope determined by the local maximum point of the data sequence and the lower envelope determined by the local minimum point have a mean of 0, that is, the signal is symmetrical about the time axis. The above two conditions are the preset component conditions. When the terminal determines that the above-mentioned differential oil temperature sequence h1(t) does not meet the preset component conditions, it can treat h1(t) as the original sequence to obtain the envelope of the maximum and minimum values ​​of h1(t), that is, e max11 (t) and e min11 (t), and then calculate the mean envelope m of the upper and lower envelopes 11 (t), then judge h 11 (t)=h1(t)-m 11(t) Whether the IMF conditions are met. If not, repeat the cycle k times to obtain h 1k (t) = h 1(k-1) (t)-m 1k (t), so that h 1k (t) satisfies the IMF conditions. Let c1(t) = h 1k (t), then it is the first IMF of x(t), which is regarded as IMF1; separate c1(t) from x(t), and regard the remainder r1(t) = x(t) - c1(t) as the original sequence. Repeat the above steps to obtain IMF2, IMF3, ..., until c n (t) or r n (t) When the given termination condition is met, the screening process ends.

[0076] Through this embodiment, the terminal can determine whether the differential oil temperature sequence meets the component conditions and then determine the division structure of the intrinsic modal components, so that the terminal can predict the oil temperature based on the divided multi-layer IMF components and residual components, thereby improving the efficiency of oil temperature prediction.

[0077] In one embodiment, it also includes: obtaining multiple sample connotation modal components and sample residual components corresponding to the sample oil temperature sequence, and obtaining a long short-term memory artificial neural network to be trained; the long short-term memory artificial neural network to be trained includes an input gate, a forget gate and an output gate; for each sample connotation modal component, the sample connotation modal component is divided into an input sample connotation modal component and an output sample connotation modal component; the time corresponding to the input sample connotation modal component is less than the time corresponding to the output sample connotation modal component; the input sample connotation modal component is input into the long short-term memory artificial neural network to be trained, and the long short-term memory artificial neural network to be trained predicts the input sample connotation modal component through the input gate, the forget gate and the output gate in sequence, and outputs the predicted sample connotation modal component corresponding to the input sample connotation modal component; the predicted sample connotation modal component represents the connotation modal component after the time corresponding to the input sample connotation modal component; according to the comparison result of the predicted sample connotation modal component and the output sample connotation modal component, the model parameters of the long short-term memory artificial neural network to be trained are adjusted until the training conditions are met, and the trained target long short-term memory artificial neural network is obtained.

[0078] In this embodiment, the terminal can pre-train a long short-term memory artificial neural network. The terminal can obtain multiple sample intrinsic modal components and sample residual components corresponding to the sample oil temperature sequence, and obtain the long short-term memory artificial neural network to be trained. Among them, the long short-term memory artificial neural network to be trained includes an input gate, a forget gate and an output gate. Specifically, the above-mentioned long short-term memory artificial neural network can be an LSTM neural network. The terminal uses the LSTM algorithm to predict the data of each component. Compared with RNN, the LSTM algorithm introduces cell state memory information in structure, and the input gate, forget gate, and output gate introduce sigmoid function and combine it with tanh function, add summation operation, and reduce the possibility of gradient vanishing and gradient explosion; in terms of performance, the LSTM algorithm can handle both short-term dependency problems and long-term dependency problems, making the prediction results more accurate and reliable.

[0079] Since there can be multiple sample connotation modal components, for each sample connotation modal component, the terminal can divide the sample connotation modal component into an input sample connotation modal component and an output sample connotation modal component. The time corresponding to the input sample connotation modal component is less than the time corresponding to the output sample connotation modal component. Specifically, the sample connotation modal component can be a sequence. The terminal can set a time point in the sample connotation modal component, and use the sample connotation modal component before the time point as the input sample connotation modal component, and use the sample connotation modal component after the time point as the input sample connotation modal component, that is, the input sample connotation modal component is the training set, and the output sample connotation modal component is the validation set. The terminal can input the input sample connotation modal component into the long short-term memory artificial neural network to be trained, and the long short-term memory artificial neural network to be trained predicts the input sample connotation modal component through the input gate, forget gate and output gate in sequence, and outputs the predicted sample connotation modal component corresponding to the input sample connotation modal component. The predicted sample connotation modal component represents the connotation modal component after the time corresponding to the input sample connotation modal component. The terminal may obtain a comparison result, such as a similarity comparison result, between the intensional modal component of the predicted sample and the intensional modal component of the output sample, and adjust the model parameters of the long short-term memory artificial neural network to be trained based on the comparison result until a training condition is met, thereby obtaining a trained target long short-term memory artificial neural network. The training condition may be satisfied when the similarity between the intensional modal component of the predicted sample and the intensional modal component of the output sample is greater than or equal to a preset similarity threshold, at which point the terminal may determine that training is complete.

[0080] Specifically, the structure of the above-mentioned long short-term memory artificial neural network can be as follows Figure 2 As shown, Figure 2FIG. 1 is a schematic diagram of the structure of a long short-term memory artificial neural network in an embodiment. In the memory unit of the LSTM neural network, the “forget gate” f t Can control the state c of the previous memory unit t-1 The degree of being forgotten, the "forget gate" t The output h of the upper unit t-1 With the input x of this unit t Obtained by the activation function, its value range is [0,1], "0" means completely forgotten, and "1" means completely retained. t Expressed as: f t =σ(W xf x t +W hf h t-1 +b f ); where W xf 、W hf is the weight coefficient of the “forget gate”, b f is the bias coefficient of the “forget gate”, σ is the “sigmoid” activation function, h t-1 represents the sequence of outputs of the previous unit, x t Represents the input sequence of this unit.

[0081] The terminal needs to determine the new information stored in the unit state. First, use the output h of the previous unit t-1 With the input x of this unit t A new candidate unit information is obtained through the tanh function Then through the "input gate" i t Control how much new information is retained and update the memory cell state C accordingly t . The “input gate” i t , candidate unit information Updated cell state C t The expressions are: i t =σ(W xi x t +W hi h t-1 +b i ), Where W is the weight coefficient, b is the bias coefficient, σ is the "sigmoid" activation function, and the symbol "·" represents the dot product. The memory unit of the LSTM neural network also includes an output gate. The "output gate" can control the output of the memory unit. The final output of the memory unit is the updated unit state C after the tanh function is activated. t And through the "output gate", the mathematical expression is as follows: t =σ(W xo x t +W ho ht-1 +b o ), h t =O t tanh(C t ). Among them, W xo 、W ho is the weight coefficient of the output gate, b o is the bias coefficient of the output gate, σ is the “sigmoid” activation function, and the symbol “·” represents the dot product.

[0082] Through this embodiment, the terminal can train the LSTM neural network to obtain a target LSTM neural network that can be used to predict oil temperature data, thereby improving the efficiency of oil temperature prediction.

[0083] In one embodiment, a target long short-term memory artificial neural network outputs a first prediction component corresponding to each intrinsic modal component and a second prediction component corresponding to the residual component based on multiple intrinsic modal components and residual components, thereby obtaining a predicted intrinsic modal component corresponding to each intrinsic modal component and a predicted residual component corresponding to the residual component, including: for each intrinsic modal component, inputting the intrinsic modal component into the target long short-term memory artificial neural network, passing the intrinsic modal component through an input gate, a forgetting gate, and an output gate in the target long short-term memory artificial neural network in sequence, and outputting multiple first prediction components after the time when the intrinsic modal component is located, thereby obtaining a predicted intrinsic modal component corresponding to the intrinsic modal component; inputting the residual component into the target long short-term memory artificial neural network, passing the residual component through an input gate, a forgetting gate, and an output gate in the target long short-term memory artificial neural network in sequence, and outputting multiple second prediction components after the time when the residual component is located, thereby obtaining a predicted residual component corresponding to the residual component.

[0084] In this embodiment, a terminal can predict the oil temperature data of a transformer to be predicted using a trained target long-short-term memory artificial neural network. The terminal decomposes the known oil temperature data using an EMD algorithm to obtain multiple intrinsic modal components. For each intrinsic modal component, the terminal can input the intrinsic modal component into the target long-short-term memory artificial neural network. The target long-short-term memory artificial neural network includes an input gate, a forget gate, and an output gate. The terminal can sequentially pass the intrinsic modal component through the input gate, the forget gate, and the output gate in the target long-short-term memory artificial neural network, and output multiple first prediction components after the time at which the intrinsic modal component is located, thereby obtaining a predicted intrinsic modal component corresponding to the intrinsic modal component. The terminal can also input the residual component into the target long-short-term memory artificial neural network, and the target long-short-term memory artificial neural network can sequentially pass the residual component through the input gate, the forget gate, and the output gate in the target long-short-term memory artificial neural network, and output multiple second prediction components after the time at which the residual component is located, thereby obtaining a predicted residual component corresponding to the residual component. The term "after the time" refers to the time starting from the last time point of the intrinsic modal component or residual component. This means that the terminal can predict oil temperature data at various future time points based on the known oil temperature. Specifically, the terminal performs EMD decomposition on the original sequence to obtain multiple intrinsic mode functions (IMFS) and a residual trend term R. That is, the terminal can decompose the non-stationary runoff time series into multiple stationary runoff components and a trend runoff component. Then, these runoff components and the residual trend term R are predicted using LSTM to obtain the predicted item of each IMFS, that is, the above-mentioned predicted intrinsic mode component P mfs(i) (t), and the prediction result of the residual trend term, that is, the above prediction residual component P r (t). The terminal can reconstruct the above-mentioned prediction connotation Motai component and prediction residual component, that is, add them together to obtain the sequence corresponding to the prediction result, that is, the above-mentioned Among them, P mfs(i) (t) represents the predicted data of the i-th IMFS at time t; P r (t) represents the predicted data of the residual trend term at time t; P represents the final predicted sequence.

[0085] Through this embodiment, the terminal can perform oil temperature prediction based on the LSTM neural network, thereby improving the efficiency of oil temperature prediction.

[0086] In one embodiment, Figure 3 As shown, Figure 3This is a flow chart of a transformer oil temperature prediction method according to another embodiment. In this embodiment, the terminal can obtain short-term load data, such as oil temperature data, of the transformer to be predicted, forming a known oil temperature sequence. The terminal can decompose the oil temperature data using the EMD algorithm to obtain multiple IMF components and a residual component R, enabling the prediction model to capture deeper time series features. The terminal can then use LSTM to predict each IMF component and residual component R, and reconstruct the predicted components to obtain the prediction results.

[0087] Specifically, an application embodiment is provided below. The terminal can pre-acquire the ETT-small dataset in the power transformer dataset (ETDataset), pre-process the data therein, and store it in the '.csv' format. This dataset is very important for the study of time series. The data is real and valid, and the data volume is also very large. The ETT-small dataset contains two power transformers and two years of data from two sites. Each data point is recorded once a minute (marked with m), and comes from two regions in the same province in China, named ETT-small-m1 and ETT-small-m2. Each dataset contains 2×365×24×4=70080 data points. In addition, we use h to mark a dataset variant with hourly granularity, namely ETT-small-h1 and ETT-small-h2. Each data point has 8-dimensional features, including the recording date of the data point, the predicted value "oil temperature" and 6 different types of external load values. This embodiment uses the EMD-LSTM combination algorithm to further optimize the prediction effect of the ETT-small-h1 transformer dataset. The structure of the oil temperature data can be as follows. Figure 4 As shown, Figure 4 Schematic diagram of the structure of a known oil temperature sequence in one embodiment. Figure 4 This is a visualization of the ETT-small-h1.CSV data. It can be seen that the transformer oil temperature variable data is non-stationary. The terminal uses the EMD algorithm to decompose the sequence data and convert it into stationary data, which can optimize the LSTM model fitting effect. The LSTM model itself is suitable for stationary, non-trending data.

[0088] Specifically, the terminal first extracts the OT variables in the ETT-samll-h1.CSV data set, and uses EMD sequence decomposition on the 17421 OT variables. Finally, it decomposes them into 11 IMF components and 1 residual component, forming a (17421, 12)-dimensional vector data. The data fluctuations of the 11 IMF components and the residual component are visualized as follows: Figure 5 As shown, Figure 5Schematic diagram of the structure of the decomposition step in one embodiment. Each component represents the component of the original sequence at different frequencies, and the terminals can be arranged in order from high frequency to low frequency. The frequency of the component can be as follows: Figure 6 As shown, Figure 6 FIG. 4 is a structural diagram of the frequencies of the decomposed components in one embodiment. Figure 6 The middle is the frequency of data generation. It can be seen that the fluctuation range of each decomposed component data is much smaller.

[0089] The terminal can then input each component data into the LSTM model for training, and the resulting prediction fitting graph is as follows: Figure 7 As shown, Figure 7 FIG. 7 is a schematic diagram of a prediction result structure in an embodiment. It can be seen that the training prediction result 700 and the test prediction result 702 almost cover the original data, and the fitting effect is very good.

[0090] The terminal can also obtain the RMSE (Root Mean Square Error) of the above training process and test process, which can be shown in the following table:

[0091] IMF components Train-RMSE Test-RMSE 1 0.44 0.26 2 0.97 0.57 3 0.5 0.32 4 0.31 0.2 5 0.16 0.13 6 0.15 0.14 7 0.03 0.02 8 0.01 0.01 9 0.01 0.01 10 0.01 0.0 11 0.0 0.0 res 1.18 0.46

[0092] The table above shows the RMSE values ​​for each component and residual. This clearly demonstrates the significant improvement in prediction accuracy. Combining the components and residuals yields an 87.5% improvement in training data prediction accuracy and a 94.8% improvement in test data prediction accuracy, demonstrating the excellent performance of EMD sequence decomposition and the LSTM model.

[0093] Through the above embodiment, the terminal decomposes the known oil temperature series into multiple intrinsic mode components and residual components using the empirical mode decomposition algorithm, and predicts each intrinsic mode component and residual component based on a neural network to obtain the oil temperature prediction result of the transformer to be predicted, thereby improving the efficiency of oil temperature prediction. In addition, the terminal uses the EMD algorithm to decompose the oil temperature data to obtain several IMF components and a residual component R, allowing the prediction model to obtain deeper time series features. This method can not only perform signal decomposition processing based on its own time scale characteristics, reducing the training pressure of the model, but also has a high signal-to-noise ratio when processing non-stationary and nonlinear data, reducing information loss, and more effectively ensuring the efficiency and effectiveness of data processing. The LSTM algorithm used by the terminal predicts the data of each component and obtains the prediction result. The LSTM algorithm structurally introduces cell state memory information, and the input gate, forget gate, and output gate introduce the sigmoid function and combine it with the tanh function, adding a summation operation to reduce the possibility of gradient vanishing and gradient explosion. In terms of performance, the LSTM algorithm can handle both short-term and long-term dependency problems, making the prediction results more accurate and reliable.

[0094] It should be understood that, although the various steps in the flowcharts involved in the various embodiments described above are displayed in sequence according to the instructions of the arrows, these steps are not necessarily executed in sequence in the order indicated by the arrows. Unless otherwise specified herein, there is no strict order restriction on the execution of these steps, and these steps can be executed in other orders. Moreover, at least a portion of the steps in the flowcharts involved in the various embodiments described above can include multiple steps or multiple stages, and these steps or stages are not necessarily executed and completed at the same time, but can be executed at different times, and the execution order of these steps or stages is not necessarily to be carried out in sequence, but can be executed in turn or alternately with other steps or at least a portion of steps or stages in other steps.

[0095] Based on the same inventive concept, embodiments of the present application also provide a transformer oil temperature prediction device for implementing the aforementioned transformer oil temperature prediction method. The solution provided by this device is similar to the solution described in the aforementioned method. Therefore, the specific limitations of one or more embodiments of the transformer oil temperature prediction device provided below can be found in the above-described limitations of the transformer oil temperature prediction method and will not be further elaborated here.

[0096] In one embodiment, Figure 8 As shown, a transformer oil temperature prediction device is provided, comprising: an acquisition module 500, a decomposition module 502, a prediction module 504 and a reconstruction module 506, wherein:

[0097] The acquisition module 500 is used to obtain a known oil temperature sequence of the transformer to be predicted within a preset time period.

[0098] The decomposition module 502 is configured to decompose the known oil temperature sequence according to an empirical mode decomposition algorithm to obtain a plurality of intrinsic mode components and residual components corresponding to the known oil temperature sequence.

[0099] The prediction module 504 is used to output a first prediction component corresponding to each intrinsic modal component and a second prediction component corresponding to the residual component based on multiple intrinsic modal components and residual components by the target long short-term memory artificial neural network, so as to obtain a predicted intrinsic modal component corresponding to each intrinsic modal component and a predicted residual component corresponding to the residual component; the predicted intrinsic modal component includes the first prediction components of multiple time points, and the predicted residual component includes the second prediction components of multiple time points.

[0100] The reconstruction module 506 is used to obtain the prediction result of the oil temperature of the transformer to be predicted based on the multiple predicted intrinsic modal components and residual components.

[0101] In one embodiment, the decomposition module 502 is specifically used to obtain multiple maximum points and multiple minimum points in a known oil temperature sequence; determine a first envelope curve of the known oil temperature sequence based on the multiple maximum points, and determine a second envelope curve of the known oil temperature sequence based on the multiple minimum points; obtain a mean envelope curve corresponding to the first envelope curve and the second envelope curve; obtain the difference between the known oil temperature sequence and the mean envelope curve to obtain a difference oil temperature sequence; if the difference oil temperature sequence meets a preset component condition, determine that the difference oil temperature sequence is an intrinsic modal component, separate the intrinsic modal component from the known oil temperature sequence, and obtain a new known oil temperature sequence; if the intrinsic modal component meets the preset component condition, determine that the difference oil temperature sequence is an intrinsic modal component, and separate the intrinsic modal component from the known oil temperature sequence to .... If the number of intrinsic modal components is less than a preset numerical threshold and the length of the known oil temperature sequence is greater than a preset length threshold, multiple maximum points and multiple minimum points in the new known oil temperature sequence are obtained, and the steps of determining a first envelope curve of the known oil temperature sequence based on the multiple maximum points and determining a second envelope curve of the known oil temperature sequence based on the multiple minimum points are returned; if the number of intrinsic modal components is greater than or equal to the preset numerical threshold or the length of the known oil temperature sequence is less than or equal to the preset length threshold, multiple intrinsic modal components corresponding to the known oil temperature sequence are obtained based on the current multiple intrinsic modal components, and the new known oil temperature sequence is used as the residual component.

[0102] In one embodiment, the above-mentioned decomposition module 502 is specifically used to obtain multiple maximum points and multiple minimum points in the difference oil temperature sequence if the difference oil temperature sequence does not meet the preset component conditions; determine the difference envelope curve based on the multiple maximum points and multiple minimum points in the difference oil temperature sequence, and obtain the mean envelope curve of the difference envelope curve; obtain a new difference oil temperature sequence based on the difference between the difference oil temperature sequence and the mean envelope curve, if the new difference oil temperature sequence does not meet the preset component conditions, obtain multiple maximum points and multiple minimum points in the new difference oil temperature sequence, and return to the step of determining the difference envelope curve based on the multiple maximum points and multiple minimum points in the difference oil temperature sequence; if the new difference oil temperature sequence meets the preset component conditions, determine that the new difference oil temperature sequence is an intrinsic modal component.

[0103] In one embodiment, the above-mentioned device also includes: a training module, which is used to obtain multiple sample intrinsic modal components and sample residual components corresponding to the sample oil temperature sequence, and obtain a long short-term memory artificial neural network to be trained; the long short-term memory artificial neural network to be trained includes an input gate, a forget gate and an output gate; for each sample intrinsic modal component, the sample intrinsic modal component is divided into an input sample intrinsic modal component and an output sample intrinsic modal component; the time corresponding to the input sample intrinsic modal component is less than the time corresponding to the output sample intrinsic modal component; the input sample intrinsic modal component is input into the long short-term memory artificial neural network to be trained, and the long short-term memory artificial neural network to be trained predicts the input sample intrinsic modal component through the input gate, the forget gate and the output gate in sequence, and outputs the predicted sample intrinsic modal component corresponding to the input sample intrinsic modal component; the predicted sample intrinsic modal component represents the intrinsic modal component after the time corresponding to the input sample intrinsic modal component; according to the comparison result of the predicted sample intrinsic modal component and the output sample intrinsic modal component, the model parameters of the long short-term memory artificial neural network to be trained are adjusted until the training conditions are met, and the target long short-term memory artificial neural network with complete training is obtained.

[0104] In one embodiment, the prediction module 504 is specifically configured to input each intrinsic modal component into a target long short-term memory artificial neural network, and have the target long short-term memory artificial neural network pass the intrinsic modal component through an input gate, a forget gate, and an output gate in sequence, and output a plurality of first prediction components after the time at which the intrinsic modal component is located, thereby obtaining a predicted intrinsic modal component corresponding to the intrinsic modal component; and input the residual component into the target long short-term memory artificial neural network, and have the target long short-term memory artificial neural network pass the residual component through an input gate, a forget gate, and an output gate in sequence, and output a plurality of second prediction components after the time at which the residual component is located, thereby obtaining a predicted residual component corresponding to the residual component.

[0105] In one embodiment, the reconstruction module 506 is specifically configured to obtain a prediction result of the oil temperature of the transformer to be predicted based on the sum of multiple predicted intrinsic modal components and residual components.

[0106] Each module in the aforementioned transformer oil temperature prediction device can be implemented in whole or in part through software, hardware, or a combination thereof. Each module can be embedded in or independent of a processor in a computer device in hardware form, or can be stored in a computer device memory in software form, so that the processor can call and execute the corresponding operations of each module.

[0107] In one embodiment, a computer device is provided. The computer device may be a terminal, and its internal structure diagram may be as follows: Figure 9 As shown. The computer device includes a processor, a memory, a communication interface, a display screen and an input device connected via a system bus. The processor of the computer device is used to provide computing and control capabilities. The memory of the computer device includes a non-volatile storage medium and an internal memory. The non-volatile storage medium stores an operating system and a computer program. The internal memory provides an environment for the operation of the operating system and the computer program in the non-volatile storage medium. The communication interface of the computer device is used to communicate with an external terminal in a wired or wireless manner, and the wireless manner can be achieved through WIFI, a mobile cellular network, NFC (near field communication) or other technologies. When the computer program is executed by the processor, a transformer oil temperature prediction method is implemented. The display screen of the computer device can be a liquid crystal display screen or an electronic ink display screen, and the input device of the computer device can be a touch layer covering the display screen, or a key, trackball or touchpad provided on the computer device housing, or an external keyboard, touchpad or mouse.

[0108] Those skilled in the art will understand that Figure 9 The structure shown in the figure is only a block diagram of a part of the structure related to the solution of the present application, and does not constitute a limitation on the computer device to which the solution of the present application is applied. The specific computer device may include more or fewer components than shown in the figure, or combine certain components, or have a different component arrangement.

[0109] In one embodiment, a computer device is provided, including a memory and a processor. The memory stores a computer program, and the processor implements the above-mentioned transformer oil temperature prediction method when executing the computer program.

[0110] In one embodiment, a computer-readable storage medium is provided, on which a computer program is stored. When the computer program is executed by a processor, the above-mentioned transformer oil temperature prediction method is implemented.

[0111] In one embodiment, a computer program product is provided, comprising a computer program, which implements the above transformer oil temperature prediction method when executed by a processor.

[0112] It should be noted that the user information (including but not limited to user device information, user personal information, etc.) and data (including but not limited to data used for analysis, stored data, displayed data, etc.) involved in this application are all information and data authorized by the user or fully authorized by all parties.

[0113] Those skilled in the art will appreciate that all or part of the processes in the above-mentioned embodiment methods can be implemented by instructing the relevant hardware through a computer program, and the computer program can be stored in a non-volatile computer-readable storage medium. When the computer program is executed, it can include the processes of the embodiments of the above-mentioned methods. Among them, any reference to memory, database or other media used in the embodiments provided in this application may include at least one of non-volatile and volatile memory. Non-volatile memory may include read-only memory (ROM), magnetic tape, floppy disk, flash memory, optical memory, high-density embedded non-volatile memory, resistive random access memory (ReRAM), magnetic random access memory (MRAM), ferroelectric random access memory (FRAM), phase change memory (PCM), graphene memory, etc. Volatile memory may include random access memory (RAM) or external cache memory, etc. By way of illustration and not limitation, RAM can be in various forms, such as static random access memory (SRAM) or dynamic random access memory (DRAM). The database involved in the various embodiments provided herein may include at least one of a relational database and a non-relational database. Non-relational databases may include, but are not limited to, distributed databases based on blockchains. The processor involved in the various embodiments provided herein may be, but are not limited to, a general-purpose processor, a central processing unit, a graphics processing unit, a digital signal processor, a programmable logic unit, a data processing logic unit based on quantum computing, and the like.

[0114] The technical features of the above embodiments can be combined arbitrarily. To make the description concise, not all possible combinations of the technical features in the above embodiments are described. However, as long as there is no contradiction in the combination of these technical features, they should be considered to be within the scope of this specification.

[0115] The above-described embodiments merely represent several implementation methods of the present application. While the descriptions are relatively specific and detailed, they should not be construed as limiting the scope of the present application. It should be noted that a person of ordinary skill in the art may make various modifications and improvements without departing from the spirit of the present application, and these modifications and improvements fall within the scope of protection of the present application. Therefore, the scope of protection of the present application shall be determined by the appended claims.

Claims

1. A transformer oil temperature prediction method, characterized in that: The method comprises: Obtaining a known oil temperature sequence of the transformer to be predicted within a preset time period; Determine a first envelope curve of the known oil temperature sequence according to a plurality of maximum points in the known oil temperature sequence, and determine a second envelope curve of the known oil temperature sequence according to a plurality of minimum points in the known oil temperature sequence; obtain the difference between the known oil temperature sequence and the mean envelope curves corresponding to the first envelope curve and the second envelope curve to obtain a differential oil temperature sequence; if the differential oil temperature sequence meets a preset component condition, determine that the differential oil temperature sequence is an intrinsic modal component, separate the intrinsic modal component from the known oil temperature sequence, and obtain a new known oil temperature sequence; if the intrinsic modal component .... If the number of the intrinsic modal components is less than a preset numerical threshold, a plurality of maximum points and a plurality of minimum points in the new known oil temperature sequence are obtained, and the steps of determining a first envelope curve of the known oil temperature sequence according to the plurality of maximum points in the known oil temperature sequence and determining a second envelope curve of the known oil temperature sequence according to the plurality of minimum points in the known oil temperature sequence are returned; if the number of the intrinsic modal components is greater than or equal to the preset numerical threshold, a plurality of intrinsic modal components corresponding to the known oil temperature sequence are obtained according to the currently existing plurality of intrinsic modal components, and the new known oil temperature sequence is used as a residual component; The target long short-term memory artificial neural network outputs a first prediction component corresponding to each intrinsic modal component and a second prediction component corresponding to the residual component based on the multiple intrinsic modal components and the residual component, thereby obtaining a predicted intrinsic modal component corresponding to each intrinsic modal component and a predicted residual component corresponding to the residual component; the predicted intrinsic modal component includes the first prediction components at multiple time points, and the predicted residual component includes the second prediction components at multiple time points; The oil temperature prediction result of the transformer to be predicted is obtained according to the prediction residual component and multiple prediction intrinsic modal components.

2. The method according to claim 1, characterized in that After obtaining the difference between the known oil temperature sequence and the mean envelope curve corresponding to the first envelope curve and the second envelope curve to obtain the difference oil temperature sequence, the method further includes: If the difference oil temperature sequence does not meet the preset component condition, obtaining multiple maximum value points and multiple minimum value points in the difference oil temperature sequence; Determining a difference envelope curve according to a plurality of maximum value points and a plurality of minimum value points in the difference oil temperature sequence, and obtaining a mean envelope curve of the difference envelope curve; obtaining a new differential oil temperature sequence based on the difference between the differential oil temperature sequence and the mean envelope curve; if the new differential oil temperature sequence does not meet the preset component condition, obtaining multiple maximum points and multiple minimum points in the new differential oil temperature sequence, and returning to the step of determining the differential envelope curve based on the multiple maximum points and multiple minimum points in the differential oil temperature sequence; If the new differential oil temperature sequence meets a preset component condition, it is determined that the new differential oil temperature sequence is an intrinsic modal component.

3. The method according to claim 1, characterized in that The method further comprises: Acquire multiple sample connotation modal components and sample residual components corresponding to the sample oil temperature sequence, and obtain a long short-term memory artificial neural network to be trained; the long short-term memory artificial neural network to be trained includes an input gate, a forget gate, and an output gate; For each sample connotation modal component, the sample connotation modal component is divided into an input sample connotation modal component and an output sample connotation modal component; the time corresponding to the input sample connotation modal component is shorter than the time corresponding to the output sample connotation modal component; Inputting the input sample connotation modal component into the long short-term memory artificial neural network to be trained, the long short-term memory artificial neural network to be trained sequentially predicts the input sample connotation modal component through an input gate, a forget gate, and an output gate, and outputs a predicted sample connotation modal component corresponding to the input sample connotation modal component; the predicted sample connotation modal component represents the connotation modal component after the time corresponding to the input sample connotation modal component; According to the comparison result of the predicted sample connotation modal component and the output sample connotation modal component, the model parameters of the long short-term memory artificial neural network to be trained are adjusted until the training conditions are met, thereby obtaining the trained target long short-term memory artificial neural network.

4. The method according to claim 3, characterized in that The target long short-term memory artificial neural network outputs a first prediction component corresponding to each connotation modal component and a second prediction component corresponding to the residual component based on the multiple connotation modal components and the residual component, to obtain a predicted connotation modal component corresponding to each connotation modal component and a predicted residual component corresponding to the residual component, including: For each connotation modal component, the connotation modal component is input into a target long short-term memory artificial neural network, and the target long short-term memory artificial neural network sequentially passes the connotation modal component through an input gate, a forget gate, and an output gate, and outputs a plurality of first predicted components after the time of the connotation modal component, to obtain a predicted connotation modal component corresponding to the connotation modal component; The residual component is input into the target long short-term memory artificial neural network, and the target long short-term memory artificial neural network passes the residual component through the input gate, the forgetting gate and the output gate in sequence and outputs multiple second prediction components after the time of the residual component to obtain the predicted residual component corresponding to the residual component.

5. The method according to claim 1, wherein Obtaining a prediction result of the oil temperature of the transformer to be predicted based on the multiple predicted intrinsic modal components and the residual component includes: The oil temperature prediction result of the transformer to be predicted is obtained according to the sum of the multiple predicted intrinsic modal components and the residual component.

6. A transformer oil temperature prediction device, characterized in that: The device comprises: An acquisition module is used to obtain a known oil temperature sequence of the transformer to be predicted within a preset time period; A decomposition module is used to determine a first envelope curve of the known oil temperature sequence according to a plurality of maximum points in the known oil temperature sequence, and to determine a second envelope curve of the known oil temperature sequence according to a plurality of minimum points in the known oil temperature sequence; obtain the difference between the known oil temperature sequence and the mean envelope curve corresponding to the first envelope curve and the second envelope curve to obtain a difference oil temperature sequence; if the difference oil temperature sequence meets a preset component condition, determine that the difference oil temperature sequence is an intrinsic modal component, separate the intrinsic modal component from the known oil temperature sequence, and obtain a new known oil temperature sequence; if the intrinsic modal component meets a preset component condition, determine that the difference oil temperature sequence is an intrinsic modal component, and separate the intrinsic modal component from the known oil temperature sequence to ... If the number of intrinsic modal components is less than a preset numerical threshold, a plurality of maximum points and a plurality of minimum points in the new known oil temperature sequence are obtained, and the steps of determining a first envelope curve of the known oil temperature sequence according to the plurality of maximum points in the known oil temperature sequence and a second envelope curve of the known oil temperature sequence according to the plurality of minimum points in the known oil temperature sequence are returned; if the number of intrinsic modal components is greater than or equal to the preset numerical threshold, a plurality of intrinsic modal components corresponding to the known oil temperature sequence are obtained according to the currently existing plurality of intrinsic modal components, and the new known oil temperature sequence is used as a residual component; a prediction module configured to output, by a target long short-term memory artificial neural network, a first prediction component corresponding to each connotational modal component and a second prediction component corresponding to the residual component based on the multiple connotational modal components and the residual component, to obtain a predicted connotational modal component corresponding to each connotational modal component and a predicted residual component corresponding to the residual component; the predicted connotational modal component includes the first prediction components at multiple time points, and the predicted residual component includes the second prediction components at multiple time points; The reconstruction module is used to obtain the oil temperature prediction result of the transformer to be predicted based on the prediction residual component and the multiple prediction intrinsic modal components.

7. The device according to claim 6, characterized in that The decomposition module is specifically used to: If the difference oil temperature sequence does not meet the preset component condition, obtaining multiple maximum value points and multiple minimum value points in the difference oil temperature sequence; Determining a difference envelope curve according to a plurality of maximum value points and a plurality of minimum value points in the difference oil temperature sequence, and obtaining a mean envelope curve of the difference envelope curve; obtaining a new differential oil temperature sequence based on the difference between the differential oil temperature sequence and the mean envelope curve; if the new differential oil temperature sequence does not meet the preset component condition, obtaining multiple maximum points and multiple minimum points in the new differential oil temperature sequence, and returning to the step of determining the differential envelope curve based on the multiple maximum points and multiple minimum points in the differential oil temperature sequence; If the new differential oil temperature sequence meets a preset component condition, it is determined that the new differential oil temperature sequence is an intrinsic modal component.

8. A computer device comprising a memory and a processor, wherein the memory stores a computer program, wherein: When the processor executes the computer program, the steps of the method according to any one of claims 1 to 5 are implemented.

9. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by a processor, the steps of the method according to any one of claims 1 to 5 are implemented.

10. A computer program product comprising a computer program, characterized in that When the computer program is executed by a processor, the steps of the method according to any one of claims 1 to 5 are implemented.

Images