Power transmission line icing thickness prediction method based on interpretable physical deep learning model

Abstract

The invention discloses a power transmission line icing thickness prediction method based on an interpretable physical deep learning model. The method comprises the steps that meteorological data and power system data are acquired and preprocessed; performing correlation analysis on the preprocessed data, selecting input features by using a Pearson's correlation coefficient, and performing data standardization to construct a power transmission line icing database; constructing a power transmission line icing thickness prediction model based on the deep learning model; constructing a physical model based on the physical rule, wherein the physical model calculates the icing thickness of the power transmission line according to the ice load; establishing a physical error term in combination with the physical model, establishing a data error term by taking error minimization as a target, and establishing a total loss function through the data error term and the physical error term; and in combination with the total loss function, optimizing the power transmission line icing thickness prediction model by using a white whale optimization algorithm, and predicting the icing thickness of the power transmission line by using the optimized model. The method provided by the invention has relatively high prediction precision and good interpretability.

Description

Technical Field

[0001] The present invention belongs to the technical field of natural disaster risk assessment of power systems, and specifically relates to a method for predicting the icing thickness of transmission lines based on an interpretable physical deep learning model. Background Art

[0002] To strengthen the forecasting and early warning of power meteorological disasters and provide refined meteorological services for the safe operation of the power grid and power dispatching, it is urgent to carry out research on predicting the icing thickness of transmission lines, improve the prediction accuracy of the icing thickness of transmission lines, and provide support for the safe operation of new power system equipment and reliable power supply.

[0003] However, in existing research, less comprehensive consideration is given to the establishment of a prediction model for the icing thickness of transmission lines, which usually requires considering physical laws, deep learning models, and optimization of error minimization, while taking into account both accuracy and interpretability. Therefore, it is necessary to provide a method for predicting the icing thickness of transmission lines based on an interpretable physical deep learning model to effectively balance the accuracy and interpretability of the icing prediction of transmission lines under the background of natural disasters. Summary of the Invention

[0004] The purpose of the present invention is to provide a method for predicting the icing thickness of transmission lines based on an interpretable physical deep learning model, which comprehensively considers physical laws, deep learning models, and optimization of error minimization when establishing a prediction model for the icing thickness of transmission lines, so that the model takes into account both accuracy and interpretability.

[0005] Next, in the first aspect of the present invention, a method for predicting the icing thickness of transmission lines based on an interpretable physical deep learning model is provided, and the method includes:

[0006] Obtain meteorological data and power system data and perform preprocessing;

[0007] Perform a correlation analysis on the preprocessed data, select input features using the Pearson correlation coefficient, and perform data standardization to construct a transmission line icing database;

[0008] Based on a deep learning model, construct a prediction model for the icing thickness of transmission lines, where the prediction model for the icing thickness of transmission lines takes the selected input features as input and the icing thickness as output;

[0009] Based on physical laws, construct a physical model, and the physical model calculates the icing thickness of the transmission line according to the ice load. The formula is as follows:

[0010]

[0011] In the formula, y n is the icing thickness of the nth transmission line, g ice is the ice load, π is the pi, ρ0 ρ is the ice density, and d is the diameter of the transmission line;

[0012] A physical error term is established by combining a physical model. A data error term is established with the goal of minimizing the error. And a total loss function is established through the data error term and the physical error term, as follows:

[0013]

[0014] L = λ data + αλ physical

[0015] In the formula, λ data is the data error term, λ physical is the physical error term, L is the total loss function, α is the weight coefficient, is the model prediction value, is the true value, is the ice coating thickness y n of the first derivative, T true is the temperature, T melt is the melting temperature of ice, β is the coefficient of the wind deflection angle, V WINDDEFANG is the actual angle of the wind deflection angle, d base is the reference diameter;

[0016] Combined with the total loss function, the beluga optimization algorithm is used to optimize the prediction model of the ice coating thickness of the transmission line, and the optimized prediction model of the ice coating thickness of the transmission line is used to predict the ice coating thickness of the transmission line.

[0017] In some embodiments, the meteorological data and power system data include ice coating thickness, temperature, humidity, wind deflection angle, wind direction, tensile force, and line inclination.

[0018] In some embodiments, the preprocessing includes median filling for missing values in discrete data, interpolation filling for missing values in continuous data, mode filling for missing values in categorical data, and removing data with large noise or outliers.

[0019] In some embodiments, the Pearson correlation coefficient is used to select input features, including:

[0020] Calculate the Pearson correlation coefficient between other data except the ice coating thickness, select the data pairs with the Pearson correlation coefficient exceeding the preset threshold, and select one of them for removal. Finally, the remaining data is selected as the input features.

[0021] In some embodiments, the ice load calculation method is:

[0022] Calculate the wind load through the diameter of the overhead transmission line and the wind deflection angle:

[0023]

[0024] Where: g wind is the wind load, C d is the drag coefficient, ρ 1 is the air density, L is the length of the transmission line, v is the wind speed, is the wind deflection angle.

[0025] The ice load is obtained by subtracting the wind load from the total load.

[0026] In some embodiments, the method further includes:

[0027] Evaluating the optimized prediction model of the ice thickness on the transmission line using an error index and a fitting index; wherein, the error index includes the root mean square error RMSE and the mean absolute error MAE, and the fitting index includes the explained variance EV and the coefficient of determination R 2 , and the formulas are as follows:

[0028]

[0029] Where, M RMSE is the root mean square error, M MAE is the mean absolute error, M EV is the explained variance, M R2 is the coefficient of determination, N is the total number of samples, is the model prediction value, is the true value, y n is the average value of the true values.

[0030] In some embodiments, the method further includes:

[0031] If the coefficient of determination is less than 0.70, continue to optimize the prediction model of the ice thickness on the transmission line.

[0032] In some embodiments, the method further includes:

[0033] Performing an interpretability analysis on the optimized prediction model of the ice thickness on the transmission line using SHAP values, and calculating the contribution of each feature to the prediction result:

[0034]

[0035] Where, is the model prediction value, f(x ij ) is the contribution of the jth variable in the ith input feature to , is the mean value of all variables in the input feature;

[0036] Performing an interpretability analysis on the feature dependence through SHAP values:

[0037]

[0038] In the formula, is the interaction between features i and j, is the pure effect value of each feature after excluding other features. S is the set of all features except features i and j, |S| is the total number of variables in S, N is the total number of samples, and f x (S ∪ {i, j}) is the predicted value including features i and j, and f x (S ∪ {i}) is the predicted value including feature i, and f x (S ∪ {j}) is the predicted value including feature j, and f x (S) is the predicted value excluding features i and j, and α i n is the SHAP value of the model features.

[0039] According to a second aspect of the present invention, there is provided an electronic device, including: a processor and a memory. The memory stores a program or instruction that can run on the processor. When the program or instruction is executed by the processor, the steps of the transmission line icing thickness prediction method of the interpretable physical deep learning model described in any one of the first aspects are implemented.

[0040] According to a third aspect of the present invention, there is provided a readable storage medium, on which a program or instruction is stored. When the program or instruction is executed by the processor, the steps of the transmission line icing thickness prediction method of the interpretable physical deep learning model described in any one of the first aspects are implemented.

[0041] Generally speaking, compared with the prior art by the above technical solution conceived by the present invention, the following beneficial effects can be achieved:

[0042] The present invention provides a transmission line icing thickness prediction model that minimizes prediction error and has interpretability. First, a physical deep learning model is constructed by combining a deep learning model, a physical model, and a beluga optimization algorithm. Then, the predicted values are evaluated using error metrics and fitting metrics. Secondly, the influence of input features on output features and model prediction results is analyzed through SHAP values. This method has high prediction accuracy and good interpretability, and can provide support for the safe operation of new power system equipment and reliable power supply, as well as improve prediction accuracy. BRIEF DESCRIPTION OF THE DRAWINGS

[0043] Figure 1 is a framework diagram of a method for predicting the icing thickness of a transmission line by an interpretable physical deep learning model provided by an embodiment of the present application;

[0044] Figure 2 is an analysis diagram of the Pearson correlation coefficient of features provided by an embodiment of the present application;

[0045] Figure 3 It is a comparison graph of the predicted value and the true value of a prediction model provided by an embodiment of the present application;

[0046] Figure 4 It is an interpretability analysis graph of the influence of model prediction provided by an embodiment of the present application;

[0047] Figure 5 It is a schematic diagram of the hardware structure of a computer device provided by an embodiment of the present application. Detailed implementation manners

[0048] In order to make the objectives, technical solutions and advantages of the present invention clearer, the present invention will be further described in detail below with reference to the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are only used to explain the present invention and are not used to limit the present invention. Based on the embodiments provided in the present application, all other embodiments obtained by those of ordinary skill in the art without making creative efforts fall within the scope of protection of the present invention.

[0049] Obviously, the accompanying drawings in the following description are only some examples or embodiments of the present application. For those of ordinary skill in the art, the present application can be applied to other similar scenarios based on these drawings without making creative efforts. In addition, it can also be understood that although the efforts made in this development process may be complex and lengthy, for those of ordinary skill in the art related to the content disclosed in the present application, some design, manufacturing or production changes based on the technical content disclosed in the present application are only conventional technical means and should not be understood as the content disclosed in the present application being insufficient.

[0050] Referring to "embodiment" in the present application means that the specific features, structures or characteristics described in connection with the embodiment may be included in at least one embodiment of the present application. The phrase appears in various positions in the specification does not necessarily refer to the same embodiment, nor is it an independent or alternative embodiment mutually exclusive with other embodiments. Those of ordinary skill in the art explicitly and implicitly understand that the embodiments described in the present application can be combined with other embodiments without conflict.

[0051] Unless otherwise defined, the technical terms or scientific terms involved in this application shall have the ordinary meanings understood by those with ordinary skills in the technical field to which this application belongs. The words such as "a", "an", "one kind", "the" and the like involved in this application do not indicate a quantity limitation and may represent a singular or plural number. The terms "include", "comprise", "have" and any variations thereof involved in this application are intended to cover non-exclusive inclusion; for example, a process, method, system, product or device that includes a series of steps or modules (units) is not limited to the listed steps or units, but may further include unlisted steps or units, or may further include other steps or units inherent to these processes, methods, products or devices. The similar words such as "connect", "be connected", "couple" and the like involved in this application are not limited to physical or mechanical connections, but may include electrical connections, whether direct or indirect. The "multiple" involved in this application means two or more. "And / or" describes the association relationship of associated objects and indicates that three relationships may exist. For example, "A and / or B" may represent: A exists alone, A and B exist simultaneously, and B exists alone. The character " / " generally represents an "or" relationship between the front and back associated objects. The terms "first", "second", "third" and the like involved in this application are only used to distinguish similar objects and do not represent a specific order for the objects.

[0052] In the context of frequent ice disasters and increasing complexity of prediction models, this application proposes a prediction model for the ice coating thickness of transmission lines aimed at minimizing prediction errors and making predictions interpretable. First, a physical deep learning model is constructed by combining a deep learning model, a physical model, and a beluga optimization algorithm. Then, error metrics and fitting metrics are used to evaluate the predicted values. Secondly, the influence of input features on output features and model prediction results is analyzed through SHAP values. To verify the accuracy and interpretability of the proposed method, this application takes the actual ice coating data of transmission lines as an example for verification. The results show that the method has high prediction accuracy and good interpretability, so the method is advanced.

[0053] Figure 1 It is a framework diagram of a method for predicting the ice coating thickness of a transmission line by an interpretable physical deep learning model provided in an embodiment of this application. As Figure 1 shown, the method for predicting the ice coating thickness of a transmission line by an interpretable physical deep learning model in an embodiment of this application specifically includes the following steps:

[0054] Step 1: Perform data preprocessing on meteorological data and power system data, convert historical data into input features, and construct a transmission line ice coating database.

[0055] Step 2: With the goal of minimizing errors, establish an optimization model with physical significance and use the physical model to modify the parameters of the deep learning prediction model.

[0056] Step 3: According to the prediction model obtained in Step 2, predict the ice thickness of the transmission line, and evaluate the predicted values using error indicators and fitting indicators.

[0057] Step 4: According to the prediction model obtained in Step 2, use SHAP values to evaluate and analyze the interpretability of the prediction model and its feature dependencies.

[0058] Step 5: Taking a certain transmission line as an example, illustrate the specific prediction process and results.

[0059] In some of these embodiments, the data preprocessing of the meteorological data and power system data related to the ice coating of the transmission line in Step 1, converting historical data into input features, and constructing a transmission line ice coating database are specifically as follows:

[0060] Step 11: Collect historical real data of meteorological data and power systems, including the ice thickness of the transmission line, ambient temperature, ambient humidity, wind deflection angle, wind direction, tensile force, and line inclination angle.

[0061] Step 12: Preprocess the historical meteorological data and power system data.

[0062] (1) For missing values in discrete data, use median filling; for missing values in continuous data, use interpolation filling; for missing values in categorical data, use mode filling method.

[0063] (2) For data with large noise or outliers, eliminate them to reduce prediction errors.

[0064] Step 13: Conduct a correlation analysis on the features in the historical data of meteorology and power systems. Use the Pearson correlation coefficient to evaluate the relationship between each influencing feature, and select key input features. The Pearson correlation coefficient satisfies:

[0065]

[0066] In the formula: ρ X,Y is the Pearson correlation coefficient value of each feature, X and Y are different features respectively, and σ X and σ Y are the standard deviations of features X and Y respectively.

[0067] Step 14: Standardize the features in the historical database, satisfying:

[0068]

[0069] In the formula: xscale is the data after standardization, x max and x min are the maximum and minimum values of the feature data respectively.

[0070] Step 15: Integrate the pre - processed meteorological data and the historical real - time data of the power system to construct a transmission line icing database.

[0071] In some of these embodiments, in step 2, with the goal of minimizing the error, an optimization model with physical significance is established, and the parameters of the deep - learning prediction model are modified using the physical model as follows:

[0072] Step 21: Use the beluga whale optimization algorithm to establish an optimization model with the goal of minimizing the error as the objective function.

[0073] (1) The exploration stage is established by the swimming of beluga whales. The position of the search agent is determined by the swimming of a pair of beluga whales, so the search model is as follows:

[0074]

[0075] In the formula: is the new position of the i - th beluga whale in the j - th dimension, and the current positions of the i - th beluga whale and the r - th beluga whale, r 1 and r 2 are random numbers between the interval (0, 1) respectively, which can enhance the random operator in the exploration stage. The selection of even and odd numbers in the formula can update the position to reflect the swimming or diving of beluga whales.

[0076] (2) After the beluga whales prey, they enter the exploitation stage. Assuming they adopt the Lévy flight strategy to prey, the exploitation model is as follows:

[0077]

[0078] In the formula: is the best position of the beluga whale, r 3 and r 4 are random numbers between the interval (0, 1) respectively, C 1 is the random jump intensity of the Lévy flight, L flight is the Lévy flight function.

[0079] (3) The whale fall stage means that beluga whales face orcas, polar bears, and humans. The probability of whale fall is selected as a minor change within the group, so the whale fall model is as follows:

[0080]

[0081] In the formula: r 5 、r6 and r 7 are random numbers between the intervals (0, 1), and X step is the step length of the whale's fall.

[0082] Step 22: Construct a physical model based on physical laws to ensure the maximum and minimum values of the prediction curve, and use the coefficient of determination to evaluate whether the parameters of the prediction model need to be modified. When its value is less than 0.70, modify the parameters of the prediction model.

[0083] (1) The balance point of the overhead transmission line is based on a theoretical model, and its combined load includes ice load and wind load, where the wind load can be calculated through the line diameter and wind deflection angle of the overhead transmission line:

[0084]

[0085] In the formula: g wind is the wind load, C d is the drag coefficient, ρ 1 is the air density, L is the length of the transmission line, v is the wind speed, is the wind deflection angle.

[0086] Thus, the ice load g can be calculated through the combined load and wind load ice .

[0087] (2) Combining the known ice density and the line diameter of the overhead transmission line and other conditions, calculate the specific icing thickness. The standard wire icing thickness is as follows:

[0088]

[0089] In the formula: y n is the icing thickness of the transmission line calculated by the physical model, n is the nth transmission line, ρ 0 = 0.9×10 -3 kg / (m×mm 2 ) is the ice density, and d is the diameter of the transmission line.

[0090] (3) The data error term and the physical error term can constrain the predicted value of the ice thickness of the overhead transmission line within the range allowed by physical laws, and they respectively satisfy:

[0091]

[0092] L = λ data +αλ physical

[0093] In the formula: λ data and λ physical are the data error term and the physical error term respectively, is the true value, is the predicted value, is the ice thickness y n changing with time, T true is the actual temperature, T melt is the melting temperature of ice, β is the coefficient of wind deflection angle, representing the influence degree of wind deflection angle on ice layer thickening, V WINDDEFANG is the actual angle of wind deflection, d base is the reference diameter, L is the total loss function, α is the weight coefficient that can control the influence of physical constraints.

[0094] In some of these embodiments, for the prediction of the ice thickness of the transmission line according to the prediction model obtained in step 2 in step 3, the predicted value is evaluated using the error index and the fitting index as follows:

[0095] Step 31: Evaluate the performance of the prediction model using the error index and the fitting index. The error index is the root mean square error (RMSE) and the mean absolute error (MAE), and the fitting index is the explained variance (EV) and the coefficient of determination (R-squared, R2). The specific expressions are as follows:

[0096]

[0097] In the formula: is the true value, is the predicted value, y n is the average value of the true values, and N is the number of samples.

[0098] In some of these embodiments, for the evaluation and analysis of the interpretability of the prediction model and its feature dependencies according to the prediction model obtained in step 2 in step 4, the following is specific:

[0099] Step 41: Conduct an interpretability analysis of the deep learning model using SHAP values, satisfying:

[0100]

[0101] In the formula: is the mean value of all variables in the feature, f(x ij ) is the contribution of the jth variable in the ith feature to

[0102] Step 42: Conduct an interpretability analysis of the feature dependencies using SHAP values, satisfying:

[0103]

[0104] In the formula: is the interaction between features i and j, is the pure effect value of each feature after excluding other features. S is the set of all feature subsets except features i and j, |S| is the total number of variables in S, N is the number of samples, and f x (S ∪ {i, j}) is the predicted value including features i and j, and f x (S ∪ {i}) is the predicted value including feature i, and f x (S ∪ {j}) is the predicted value including feature j, and f x (S) is the predicted value without features i and j, is the SHAP value of the model features.

[0105] In some of these embodiments, taking a certain transmission line as an example in step 5, the specific prediction process and results are as follows:

[0106] Step 51: Collect the meteorological data and historical real data of the power system of a certain transmission line, including the ice thickness of the transmission line, ambient temperature, ambient humidity, wind deflection angle, wind direction, tensile force, and line inclination angle, and construct a database after preprocessing.

[0107] Step 52: Predict the ice thickness of the transmission line according to the proposed ice thickness prediction method for transmission lines based on the interpretable physical deep learning model.

[0108] Step 53: Analyze the accuracy and interpretability of the prediction results.

[0109] The relevant evaluation index values of the predicted ice thickness of the transmission line obtained according to the embodiments of the present application are shown in Table 1.

[0110] Table 1 Ice thickness prediction evaluation index table for transmission lines

[0111]

[0112] Figure 2 It shows that the Pearson correlation values between the features in the database are all less than 0.5. Therefore, there is no strong correlation between each feature, so the influencing features can be used as key input features and input into the model. If there is a strong correlation between two features, then one of these two features is selected as the key feature.

[0113] Figure 3The results calculated using the model are shown. The model's prediction accuracy is highlighted because it takes into account both the physical model and the optimization model. The predicted curve is very close to the actual curve in terms of overall shape and behavior. The prediction results show that the thickness of ice on overhead transmission lines increases with time under favorable weather conditions. Due to the constraints of physical laws, the physical model ensures that the maximum and minimum values ​​of the predicted curve are consistent with the actual curve. The optimization model ensures that the error between the predicted value and the actual value is minimized.

[0114] Figure 4 The interpretability of the input features’ impact on the model’s predictions is shown. The input features are sorted by the average of their SHAP values, which are the most important features for the model. Wide areas indicate large clusters of samples. The redder the color, the larger the value of the feature itself, and the bluer the color, the smaller the value of the feature itself. The feature line inclination has the highest overall importance, followed by the feature wind direction. In the horizontal direction, the feature line inclination has a relatively scattered distribution of samples, which has a greater impact. Most of the data for the feature humidity is spread around 0, so it has little impact on the model’s predictions.

[0115] In summary, this application comprehensively considers the great challenges posed by frequent ice disasters to the stability of the power system and the increasing complexity of the prediction model in existing research. The proposed method uses the Beluga optimization method with the smallest prediction error to establish an optimization model, combines the deep learning prediction model with the physical model, and uses the SHAP value to explain the impact of input features on output features and model predictions. It is verified by taking the actual ice coverage data of the transmission line as an example, which shows the advanced nature of the proposed method, which can provide support for the safe operation of new power system equipment and reliable power supply, as well as improve the prediction accuracy. The transmission line ice thickness prediction method of the interpretable physical deep learning model proposed in this application takes into account the accuracy and interpretability of the prediction model, so this method is advanced.

[0116] In addition, combined Figure 1 The method for predicting ice thickness on transmission lines using an interpretable physical deep learning model described in the embodiments of the present application can be implemented by a computer device. Figure 5 Schematic diagram of the hardware structure of the computer device of the embodiment of the present application. Figure 5 As shown, the device may include a processor 201 and a memory 202 storing computer program instructions.

[0117] Specifically, the processor 201 may include a central processing unit (CPU), or an application specific integrated circuit (ASIC), or may be configured to implement one or more integrated circuits of the embodiments of the present application.

[0118] Among them, the memory 202 may include a mass storage for data or instructions. By way of example and not limitation, the memory 202 may include a hard disk drive (HDD), a floppy disk drive, a solid state drive (SSD), a flash memory, an optical disc, a magneto-optical disc, a magnetic tape, or a universal serial bus (USB) drive, or a combination of two or more of these. In appropriate cases, the memory 202 may include removable or non-removable (or fixed) media. In appropriate cases, the memory 202 may be internal or external to the data processing device. In a particular embodiment, the memory 202 is a non-volatile memory. In a particular embodiment, the memory 202 includes a read-only memory (ROM) and a random access memory (RAM). In appropriate cases, the ROM may be a mask-programmed ROM, a programmable ROM (PROM), an erasable PROM (EPROM), an electrically erasable PROM (EEPROM), an electrically alterable ROM (EAROM), or a flash memory, or a combination of two or more of these. In appropriate cases, the RAM may be a static random access memory (SRAM) or a dynamic random access memory (DRAM), where the DRAM may be a fast page mode dynamic random access memory (FPMDRAM), an extended data out dynamic random access memory (EDODRAM), a synchronous dynamic random access memory (SDRAM), etc.

[0119] The memory 202 can be used to store or cache various data files required for processing and / or communication, as well as possible computer program instructions executed by the processor 201.

[0120] The processor 201 reads and executes the computer program instructions stored in the memory 202 to implement the transmission line icing thickness prediction method of any interpretable physical deep learning model in the above embodiments.

[0121] In some embodiments, the point cloud generation device may further include a communication interface 203 and a bus 200. Among them, as Figure 5 shown, the processor 201, the memory 202, and the communication interface 203 are connected through the bus 200 to complete communication with each other.

[0122] The communication interface 203 is used to implement communication between the modules, devices, units, and / or devices in the embodiments of the present application. The communication interface 203 can also implement data communication with other components such as external devices, image / data acquisition devices, databases, external storage, and image / data processing workstations.

[0123] The bus 200 includes hardware, software, or both, and couples components of the point cloud generation device to each other. The bus 200 includes, but is not limited to, at least one of the following: Data Bus, Address Bus, Control Bus, Expansion Bus, Local Bus. By way of example and not limitation, the bus 200 may include an Accelerated Graphics Port (AGP) or other graphics bus, an Extended Industry Standard Architecture (EISA) bus, a Front Side Bus (FSB), a Hyper Transport (HT) interconnect, an Industry Standard Architecture (ISA) bus, an InfiniBand interconnect, a Low Pin Count (LPC) bus, a memory bus, a MicroChannel Architecture (MCA) bus, a Peripheral Component Interconnect (PCI) bus, a PCI-Express (PCI-X) bus, a Serial Advanced Technology Attachment (SATA) bus, a Video Electronics Standards Association Local Bus (VLB) bus, or other suitable bus or a combination of two or more of these. Where appropriate, the bus 200 may include one or more buses. Although embodiments of the present application describe and illustrate specific buses, the present application contemplates any suitable bus or interconnect.

[0124] The computer device can execute the method for predicting the ice coating thickness of a transmission line of the interpretable physical deep learning model in the embodiments of the present application, thereby realizing the combination Figure 1 with the method for predicting the ice coating thickness of a transmission line of the interpretable physical deep learning model described.

[0125] In addition, in combination with the method for predicting the icing thickness of transmission lines by the interpretable physical deep learning model in the above embodiments, the embodiments of the present application can provide a computer-readable storage medium to implement. Computer program instructions are stored on the computer-readable storage medium; when the computer program instructions are executed by a processor, the method for predicting the icing thickness of transmission lines by any one of the interpretable physical deep learning models in the above embodiments is implemented.

[0126] It should be noted that the technical features of the above-described embodiments can be combined arbitrarily. For the sake of brevity of description, 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, it should be considered as the scope described in this specification. In addition, according to the needs of implementation, each step / component described in the present application can be split into more steps / components, or two or more steps / components or partial operations of steps / components can be combined into new steps / components to achieve the purpose of the present invention.

[0127] Those skilled in the art can easily understand that the above-described embodiments only represent several implementation manners of the present application, and the description is relatively specific and detailed, but it cannot be understood as a limitation on the scope of the invention patent. It should be pointed out that for those of ordinary skill in the art, without departing from the concept of the present application, several deformations and improvements can still be made, and these all belong to the protection scope of the present application. Therefore, the protection scope of the patent of the present application should be subject to the appended claims.

Claims

1. A method for predicting ice thickness of power transmission lines based on an interpretable physical deep learning model, characterized in that: The method includes: Acquire meteorological data and power system data and perform preprocessing; The preprocessed data were subjected to correlation analysis, and the input features were selected using the Pearson correlation coefficient and the data were standardized to construct a transmission line icing database. A transmission line ice thickness prediction model is constructed based on the deep learning model. The transmission line ice thickness prediction model takes the selected input features as input and takes the ice thickness as output. A physical model is constructed based on physical laws. The physical model calculates the ice thickness of the transmission line according to the ice load. The formula is as follows: In the formula, y n is the ice thickness of the nth transmission line, g ice is the ice load, π is the circumference of a circle, ρ0 is the ice density, and d is the diameter of the transmission line; The physical error term is established in combination with the physical model, the data error term is established with the goal of minimizing the error, and the total loss function is established through the data error term and the physical error term, as follows: L=λ data +al physical In the formula, λ data is the data error term, λ physical is the physical error term, L is the total loss function, α is the weight coefficient, is the model prediction value, is the true value, is the ice thickness y n The first derivative of T true is the temperature, T melt is the melting temperature of ice, β is the coefficient of wind deflection angle, V WINDDEFANG is the actual angle of wind deflection, d base is the reference diameter; Combined with the total loss function, the Beluga optimization algorithm is used to optimize the transmission line ice thickness prediction model, and the optimized transmission line ice thickness prediction model is used to predict the transmission line ice thickness.

2. The method for predicting ice thickness of power transmission lines based on an interpretable physical deep learning model according to claim 1 is characterized in that: Meteorological data and power system data include ice thickness, temperature, humidity, wind deflection angle, wind direction, tension and line inclination.

3. The method for predicting ice thickness of power transmission lines based on an interpretable physical deep learning model according to claim 1, characterized in that: Preprocessing includes filling missing values ​​in discrete data with median values, filling missing values ​​in continuous data with interpolation, filling missing values ​​in categorical data with pattern filling, and removing data with large noise or outliers.

4. The method for predicting ice thickness of power transmission lines based on an interpretable physical deep learning model according to claim 1, characterized in that: Use the Pearson correlation coefficient to select input features, including: The Pearson correlation coefficients between the data except ice thickness are calculated, and the data pairs whose Pearson correlation coefficients exceed the preset threshold are selected, and one of them is selected to be eliminated, and finally the remaining data is selected as the input features.

5. The method for predicting ice thickness of power transmission lines based on an interpretable physical deep learning model according to claim 1, characterized in that: The ice load calculation method is: Calculate wind loads based on overhead transmission line path and wind deflection angle: Where: g wind is the wind load, C d is the drag coefficient, ρ1 is the air density, L is the length of the transmission line, v is the wind speed, is the wind deflection angle. The ice load is obtained by subtracting the wind load from the total load.

6. The method for predicting ice thickness of power transmission lines based on an interpretable physical deep learning model according to claim 1, characterized in that: The method further includes: The optimized transmission line ice thickness prediction model is evaluated using error indicators and fitting indicators; the error indicators include root mean square error RMSE and mean absolute error MAE, and the fitting indicators include explained variance EV and determination coefficient R 2 , the formula is as follows: Where M RMSE is the root mean square error, M MAE is the mean absolute error, M EV To explain the variance, M R2 is the coefficient of determination, N is the total number of samples, is the model prediction value, is the true value, is the average of the true values.

7. The method for predicting ice thickness of power transmission lines based on an interpretable physical deep learning model according to claim 6 is characterized in that: The method further includes: If the determination coefficient is less than 0.70, continue to optimize the transmission line ice thickness prediction model.

8. The method for predicting ice thickness of power transmission lines based on an interpretable physical deep learning model according to claim 1, characterized in that: The method further includes: The SHAP value is used to perform interpretability analysis on the optimized transmission line ice thickness prediction model, and the contribution of each feature to the prediction result is calculated: In the formula, is the model prediction value, f(x ij ) is the jth variable pair in the i-th input feature Contribution is the mean of all variables in the input features; Interpretability analysis of feature dependencies through SHAP values: In the formula, is the interaction between features i and j, is the pure effect value of each feature after excluding other features, S is the subset of all features except features i and j, |S| is the total number of variables in S, N is the total number of samples, and f x (S∪{i,j}) is the predicted value containing features i and j, f x (S∪{i}) is the predicted value containing feature i, f x (S∪{j}) contains the predicted value of feature j, and fx(S) is the predicted value without including features i and j. is the SHAP value of the model feature.

9. An electronic device, characterized in that: include: A processor and a memory, the memory storing programs or instructions that can be run on the processor, and the programs or instructions, when executed by the processor, implement the steps of the method for predicting ice thickness of transmission lines based on an interpretable physical deep learning model as described in any one of claims 1 to 8.

10. A readable storage medium, characterized in that: Programs or instructions are stored thereon, and when the programs or instructions are executed by a processor, the steps of the method for predicting ice thickness of transmission lines based on an interpretable physical deep learning model as described in any one of claims 1 to 8 are implemented.

Images