Zone area load prediction method and device based on micro-service

Abstract

The invention relates to the technical field of electric power systems, and discloses a micro-service-based transformer area load prediction method and a micro-service-based transformer area load prediction device. The method comprises the following steps: acquiring power station deployment data for each power station in a target station area, and extracting power station operation data and power station weather data under typical weather; performing time sequence stability analysis, spatial coupling degree analysis and environmental sensitivity analysis on the target transformer area by using the acquired data; constructing a plurality of to-be-selected load prediction models by using the obtained data, and calculating sudden change response capability analysis results corresponding to the prediction models; screening the prediction model according to each analysis result to obtain an optimal load prediction model; and predicting the load of the target transformer area by using the optimal load prediction model. According to the method, the appropriate load prediction model can be selected according to the characteristics of different transformer areas, and the problem that a unified mode prediction method in the prior art is difficult to adapt to the diversity of the transformer areas, and consequently the transformer area load prediction accuracy is poor is solved.

Description

Technical Field

[0001] This disclosure relates to the field of power system technology, and more specifically, to a microservice-based method and apparatus for predicting transformer area load. Background Technology

[0002] With the widespread application of distributed photovoltaic (PV) power generation technology in power systems, load forecasting technology for distributed PV distribution areas is also constantly developing.

[0003] In the early stages of load forecasting technology development, traditional methods based on statistical analysis and empirical formulas were primarily employed. These methods typically relied on historical load data and simple meteorological factors for forecasting. In recent years, with the rise of artificial intelligence and machine learning technologies, numerous load forecasting methods based on these technologies have emerged. For example, neural network models are used, trained on large amounts of historical data, to learn patterns in load changes and achieve forecasting.

[0004] However, with the increasing number and wide distribution of distributed photovoltaic (PV) power stations, their operating characteristics are affected by various factors such as geographical location, sunlight conditions, and equipment performance, resulting in significant differences in load characteristics across different distribution areas. Traditional load forecasting techniques typically use the same forecasting model for multiple distribution areas. This uniform forecasting method is ill-suited to the diversity of distribution areas and fails to deeply understand the characteristics of different distribution areas, thus resulting in poor accuracy in load forecasting. Summary of the Invention

[0005] In view of the above situation, this disclosure provides a microservice-based method and apparatus for predicting transformer load, which aims to solve the above problems or at least partially solve the above problems.

[0006] In a first aspect, embodiments of this disclosure provide a microservice-based method for predicting transformer load. The method is implemented using a microservice architecture, which includes a data management microservice and a load prediction matching microservice. The method includes:

[0007] By calling the data management microservice, the deployment data of each photovoltaic power station in the target area is obtained, and the operation data and weather data of the power station under typical weather conditions are extracted.

[0008] Using the power plant deployment data, as well as the power plant operation data and power plant weather data under typical weather conditions, the target power area is subjected to time series stability analysis, spatial coupling analysis and environmental sensitivity analysis, and the corresponding analysis results are obtained.

[0009] Based on the power plant operation data and power plant weather data under the typical weather conditions, multiple candidate load prediction models are constructed, and the analysis results of the sudden change response capability corresponding to each candidate load prediction model are calculated. Among them, the multiple candidate load prediction models are deployed to the microservice center through containerized deployment, including: time series prediction model, feature interaction prediction model and comprehensive prediction model.

[0010] By calling the load forecasting matching microservice, the multiple candidate load forecasting models are screened based on the analysis results and the sudden change response capability analysis results to obtain the optimal load forecasting model for the target transformer area.

[0011] Using the optimal load forecasting model, the power station deployment data, historical operation data, and weather data of each photovoltaic power station in the target area are processed to obtain the load forecasting data of the target area.

[0012] Secondly, this disclosure also provides a microservice-based transformer load forecasting device, which is implemented based on a microservice architecture, including a data management microservice and a load forecasting matching microservice; the device includes:

[0013] The analysis module is used to acquire power station deployment data and extract power station operation data and weather data under typical weather conditions for each photovoltaic power station in the target area by calling the data management microservice. Using the power station deployment data, power station operation data, and power station weather data under typical weather conditions, the module performs time-series stability analysis, spatial coupling analysis, and environmental sensitivity analysis on the target area, obtaining corresponding analysis results. Based on the power station operation data and power station weather data under typical weather conditions, multiple candidate load prediction models are constructed, and the sudden change response capability analysis results corresponding to each candidate load prediction model are calculated. These multiple candidate load prediction models are deployed to the microservice center using a containerized deployment method, including: a time-series prediction model, a feature interaction prediction model, and a comprehensive prediction model.

[0014] The selection module is used to filter the multiple candidate load prediction models by calling the load prediction matching microservice, based on the analysis results and the sudden change response capability analysis results, to obtain the optimal load prediction model for the target transformer area.

[0015] The forecasting module is used to process the power plant deployment data, historical operation data and weather data of each photovoltaic power plant in the target area using the optimal load forecasting model to obtain the load forecasting data of the target area.

[0016] By employing the above technical solutions, the microservice-based load forecasting method and apparatus provided in this disclosure, through the construction of a microservice center including photovoltaic power plant data management microservices and load forecasting matching microservices, and by deploying the time-series forecasting model, feature interaction forecasting model, and comprehensive forecasting model to the microservice center using a containerized deployment approach, achieves modularity and flexible deployment of the system. This facilitates independent maintenance and updates of each microservice and model, improves the system's scalability and maintainability, and effectively addresses different business needs and changes. Simultaneously, by acquiring power plant deployment data and typical weather data for each photovoltaic power plant in the target area, analysis of time-series stability, spatial coupling, and environmental sensitivity is conducted. Multiple candidate load forecasting models, including the time-series forecasting model, feature interaction forecasting model, and comprehensive forecasting model, are constructed and their sudden change response capabilities are analyzed. Based on the analysis results, the optimal load forecasting model is selected, and finally, the model is used to process the data to obtain load forecasting data. As can be seen, this solution achieves the goal of selecting appropriate prediction models based on the characteristics of different distribution areas. It solves the problem that the existing unified prediction method is difficult to adapt to the diversity of distribution areas, cannot deeply explore the characteristics of different distribution areas, and cannot make targeted predictions based on the characteristics of different distribution areas, resulting in poor accuracy of distribution area load prediction. It provides a more reliable basis for the stable operation, rational scheduling, and optimized allocation of power resources of distributed photovoltaic systems, thereby improving the economic benefits and operating efficiency of the entire distributed photovoltaic system within the distribution area.

[0017] The above description is merely an overview of the technical solution disclosed herein. In order to better understand the technical means of this disclosure and to implement it in accordance with the contents of the specification, and to make the above and other objects, features and advantages of this disclosure more apparent and understandable, specific embodiments of this disclosure are described below. Attached Figure Description

[0018] The accompanying drawings, which are included to provide a further understanding of this disclosure and form part of this disclosure, illustrate exemplary embodiments of the present disclosure and are used to explain the disclosure, but do not constitute an undue limitation of the disclosure. In the drawings:

[0019] Figure 1 A flowchart illustrating the microservice-based transformer load prediction method provided in this embodiment of the present disclosure is shown.

[0020] Figure 2 A schematic diagram of the structure of the microservice-based transformer load prediction device provided in this embodiment of the present disclosure is shown.

[0021] Figure 3 A schematic diagram of the structure of an electronic device provided in an embodiment of this disclosure is shown. Detailed Implementation

[0022] To make the objectives, technical solutions, and advantages of this disclosure clearer, the technical solutions of this disclosure will be clearly and completely described below in conjunction with specific embodiments and corresponding drawings. Obviously, the described embodiments are only a part of the embodiments of this disclosure, and not all of them. All other embodiments obtained by those skilled in the art based on the embodiments of this disclosure without creative effort are within the scope of protection of this disclosure.

[0023] It should be noted that similar labels and letters in the following figures indicate similar items. Therefore, once an item is defined in one figure, it does not need to be further defined and explained in subsequent figures.

[0024] It should be noted that the terms "first," "second," etc., in the specification, claims, and accompanying drawings of this disclosure are used to distinguish similar objects and are not necessarily used to describe a specific order or sequence. It should be understood that such use can be interchanged where appropriate so that the embodiments of this disclosure described herein can be implemented in orders other than those illustrated or described herein. Furthermore, the term "comprising" and its variations should be interpreted as open-ended terms meaning "including but not limited to."

[0025] As mentioned earlier, the number of distributed photovoltaic (PV) power stations is constantly increasing and their distribution is widespread. Their operating characteristics are affected by various factors such as geographical location, sunlight conditions, and equipment performance, resulting in significant differences in load characteristics between different distribution areas. Traditional load forecasting techniques generally use the same forecasting model for multiple distribution areas. This uniform forecasting method is difficult to adapt to the diversity of distribution areas and cannot deeply explore the characteristics of different distribution areas to make targeted forecasts, resulting in poor accuracy of distribution area load forecasts.

[0026] Based on this, the present invention proposes a microservice-based method and apparatus for predicting transformer load. The following detailed description of the present invention will be provided through specific embodiments.

[0027] To facilitate understanding of this embodiment, a detailed description of the microservice-based distribution transformer load forecasting method disclosed in this disclosure is provided first. The execution entity of the microservice-based distribution transformer load forecasting method provided in this disclosure is generally a computer device with certain computing capabilities. This computer device may include, for example, a terminal device, a server, or other processing devices. The terminal device may be a user equipment (UE), a mobile device, a user terminal, or a terminal, etc. In some possible implementations, this microservice-based distribution transformer load forecasting method can be implemented by the processor calling computer-readable instructions stored in memory.

[0028] Figure 1This illustration shows a flowchart of a microservice-based transformer load forecasting method provided in an embodiment of this disclosure. Figure 1 As can be seen, the embodiments disclosed herein are based on a microservice architecture, which includes a data management microservice and a load prediction and matching microservice, and includes at least steps S101-S105:

[0029] S101: By calling the data management microservice, for each photovoltaic power station in the target area, obtain the power station deployment data, and extract the power station operation data and power station weather data under typical weather conditions;

[0030] S102: Using the power plant deployment data, as well as the power plant operation data and power plant weather data under typical weather conditions, perform time series stability analysis, spatial coupling degree analysis and environmental sensitivity analysis on the target power station area respectively, and obtain the corresponding analysis results;

[0031] S103: Based on the power plant operation data and power plant weather data under the typical weather conditions, construct multiple candidate load prediction models and calculate the sudden change response capability analysis results corresponding to each candidate load prediction model; wherein, the multiple candidate load prediction models are deployed to the microservice center through containerized deployment, including: time series prediction model, feature interaction prediction model and comprehensive prediction model.

[0032] S104: By calling the load prediction matching microservice, the multiple candidate load prediction models are screened according to the analysis results and the sudden change response capability analysis results to obtain the optimal load prediction model for the target transformer area;

[0033] S105: Using the optimal load forecasting model, the power station deployment data, historical operation data and weather data of each photovoltaic power station in the target area are processed to obtain the load forecasting data of the target area.

[0034] Understandably, the microservice architecture also includes the interface gateway microservice. As the entry point of the microservice center, the interface gateway microservice is responsible for handling the routing of external requests, load balancing, and communication management between services, ensuring smooth data interaction and call processes between various microservices and the prediction model.

[0035] As can be seen, this embodiment of the disclosure achieves modularity and flexible deployment of the system by building a microservice center that includes microservices such as photovoltaic power plant data management and load forecasting matching, and by deploying the time-series forecasting model, feature interaction forecasting model, and comprehensive forecasting model to the microservice center using a containerized deployment method. This facilitates independent maintenance and updates of each microservice and model, improves the system's scalability and maintainability, and effectively addresses different business needs and changes. Simultaneously, by acquiring power plant deployment data and typical weather data for each photovoltaic power plant in the target area, analysis of time-series stability, spatial coupling, and environmental sensitivity is conducted. Multiple candidate load forecasting models, including the time-series forecasting model, feature interaction forecasting model, and comprehensive forecasting model, are constructed and their abrupt change response capabilities are analyzed. Based on the analysis results, the optimal load forecasting model is selected, and finally, this model is used to process the data to obtain load forecasting data. As can be seen, this solution achieves the goal of selecting appropriate prediction models based on the characteristics of different distribution areas. It solves the problem that the existing unified prediction method is difficult to adapt to the diversity of distribution areas, cannot deeply explore the characteristics of different distribution areas, and cannot make targeted predictions based on the characteristics of different distribution areas, resulting in poor accuracy of distribution area load prediction. It provides a more reliable basis for the stable operation, rational scheduling, and optimized allocation of power resources of distributed photovoltaic systems, thereby improving the economic benefits and operating efficiency of the entire distributed photovoltaic system within the distribution area.

[0036] The following provides a detailed explanation of S101-S105.

[0037] Regarding S101-S102 above:

[0038] Power plant operation data reflects the operating status of the power plant, including but not limited to: power generation and output of the photovoltaic power plant. Power plant deployment data includes the layout and structure information of the power plant, including but not limited to: the location, scale, and layout of the photovoltaic power plant. Power plant weather data includes but is not limited to: weather type, temperature, humidity, light intensity, wind speed, irradiance, and sunshine duration.

[0039] Typical weather conditions include, for example, sunny, cloudy, overcast, and rainy weather. Using typical features instead of preprocessed data simplifies the data processing and improves the efficiency of load forecasting model building.

[0040] During implementation, data cleaning, spatiotemporal alignment, and feature encoding can be performed on power plant operation data, power plant deployment data, and power plant weather data, and the preprocessed data can replace the original data. Among these processes, data cleaning removes noise and errors from the data; spatiotemporal alignment ensures consistency of data from different sources in both time and space; and feature encoding transforms the data into a form suitable for subsequent model processing. These preprocessing operations improve data quality and usability, laying the foundation for subsequent analysis and prediction.

[0041] In some embodiments, timing stability analysis is performed on the target station area to obtain corresponding analysis results, including:

[0042] For each type of typical weather, take it as the target typical weather and perform the following steps:

[0043] The time series data consisting of power plant operation data and power plant weather data under the target typical weather conditions are subjected to stationarity check to obtain stationarity check results; autocorrelation analysis is performed to obtain autocorrelation analysis results; and fluctuation analysis is performed to obtain fluctuation analysis results.

[0044] Based on the stationarity check results, the autocorrelation analysis results, and the fluctuation analysis results, the target time series stability analysis results are obtained.

[0045] Based on the time series stability analysis results of each target, the time series stability analysis results of the target station area are determined.

[0046] For example, if there were 18 sunny days, 10 cloudy days, and 3 rainy days in the past month, then the corresponding time series data can be generated based on the power plant operation data and power plant weather data under each typical weather condition, resulting in 3 time series data: sunny day time series data, cloudy day time series data, and rainy day time series data.

[0047] Stationarity tests, autocorrelation analyses, and volatility analyses were performed on the sunny time series data, yielding results of stationarity, strong correlation, and low volatility, respectively, thus indicating that the target time series stability analysis result is stable. Similarly, stationarity tests, autocorrelation analyses, and volatility analyses were performed on the cloudy time series data, yielding results of stationarity, moderate correlation, and moderate volatility, respectively, thus indicating that the target time series stability analysis result is generally stable. Finally, stationarity tests, autocorrelation analyses, and volatility analyses were performed on the overcast time series data, yielding results of non-stationarity, weak correlation, and strong volatility, respectively, thus indicating that the target time series stability analysis result is unstable.

[0048] Finally, based on the temporal stability analysis results of each target, the temporal stability analysis result of the target station area is determined. For example, if the temporal stability analysis result corresponding to a typical weather condition is unstable, then the temporal stability analysis result of the target station area is determined to be unstable.

[0049] It should be noted that the stationarity verification method, autocorrelation analysis method, and fluctuation analysis method in this embodiment are existing technologies and will not be described in detail here.

[0050] This embodiment uses each typical weather condition as a target typical weather condition and performs stationarity verification, autocorrelation analysis, and fluctuation analysis on the time series data composed of the corresponding power plant operation data and weather data. This yields the time series stability analysis results for each target typical weather condition, and comprehensively determines the time series stability analysis results for the target power distribution area. This approach can accurately capture the time series characteristics of data under different weather patterns, thereby accurately determining the time series stability of the target power distribution area. It provides a scientific basis for the dynamic matching of subsequent prediction models and effectively improves the pertinence and accuracy of load prediction for complex power distribution areas.

[0051] Spatial coupling analysis results can reveal the degree of spatial mutual influence between various power stations in the target area. Specifically, in some embodiments, the power station deployment data includes the location information and terrain information of each photovoltaic power station; spatial coupling analysis is performed on the target area to obtain the corresponding analysis results, including:

[0052] A spatial weight matrix is ​​constructed based on the location information and terrain information of each of the aforementioned locations; the spatial weight matrix is ​​used to represent the spatial proximity or spatial interaction relationship between each of the photovoltaic power stations.

[0053] Using the operating data of each power station and the spatial weight matrix, spatial autocorrelation analysis and local spatial correlation analysis are performed on the target transformer area to obtain the corresponding analysis results.

[0054] The spatial coupling degree of the target station area is calculated based on the results of spatial autocorrelation analysis and local spatial correlation analysis.

[0055] The spatial coupling degree analysis result is determined based on the spatial coupling degree and the preset coupling degree threshold.

[0056] For example, assume the target area has three photovoltaic power stations: A, B, and C, with location coordinates: A(0,0), B(2,0), C(0,3) (unit: km); terrain information: there is a small hill between A and C, with an elevation difference of 100 meters, resulting in approximately 50% shading; the area between B and C is a plain with no shading. Power station operating data: daily power generation is A = 100 kWh, B = 120 kWh, and C = 80 kWh.

[0057] First, calculate the reciprocal of the distance between power stations, and use it as the initial weight. Specifically:

[0058]

[0059] Then, the weights are adjusted based on terrain information. Specifically: Since there is 50% occlusion between A and C, the weight is adjusted to 0.33 × 50% = 0.165; since there is no occlusion between B and C, the weight remains at 0.28; and since there is no terrain influence between A and B, the weight remains at 0.5. Finally, the spatial weight matrix is:

[0060]

[0061] Next, using the operational data and spatial weight matrix of each power station, and employing the global Moran's index, spatial autocorrelation analysis was performed on each photovoltaic power station to obtain the corresponding analysis results. Specifically, the spatial autocorrelation analysis steps are as follows:

[0062] 1. Calculate the average and deviation of power generation.

[0063] Average power generation: 100 kWh;

[0064] Deviation values: A = -20, B = +20, C = -20.

[0065] 2. Calculate the numerator.

[0066]

[0067] Substituting the values, we get the numerator as -492.

[0068] 3. Calculate the denominator term

[0069]

[0070] ∑∑W ij =0.5 + 0.165 + 0.5 + 0.28 + 0.165 + 0.28 = 1.89

[0071] 4. Global Moran Index

[0072]

[0073] Using operational data and spatial weight matrices from each power station, and employing the local Moran's index, a local spatial correlation analysis is performed on each photovoltaic power station to obtain the corresponding analysis results. In specific implementation, the local Moran's index formula is:

[0074]

[0075] Substituting the inputs of power plants A, B, and C into the local Moran exponent formula, we obtain:

[0076] IA = (100-100)×[W AB ×(120-100)+W AC [×(80-100)]=0

[0077] I B =20×[W AB ×(20)+W AC [×(-20)]=-312

[0078] I C = (80-100)×[W CA ×(100-100)+W CB [×(120-100)}=-112

[0079] Then, based on the spatial autocorrelation analysis results and the local spatial correlation analysis results, the spatial coupling degree of the target station area is calculated. Specifically:

[0080] Absolute value of the global exponent: |I| = 0.64;

[0081] Local exponential absolute mean: (|I A |+|I B |+|I C |) / 3=(0+312+112) / 3≈141.33;

[0082] The final spatial coupling degree is:

[0083] Here, we assume a global ratio of 70% and a local ratio of 30%, with 500 being a preset normalization constant. Finally, the spatial coupling degree analysis result is determined based on the spatial coupling degree and a preset coupling degree threshold. The coupling degree threshold can be set according to actual needs, and this embodiment does not limit it. For example, if the coupling degree threshold is 0.5, and the spatial coupling degree is greater than 0.5, the spatial coupling degree analysis result is high coupling degree; otherwise, the spatial coupling degree analysis result is low coupling degree.

[0084] In some embodiments, environmental sensitivity analysis is performed on the target transformer area to obtain corresponding analysis results, including:

[0085] Using power plant operation data and power plant weather data under various typical weather conditions, a multiple linear regression model was fitted to obtain the power plant weather data, which includes a variety of environmental factors.

[0086] Calculate the standard deviation of each environmental factor and power plant operating data;

[0087] Each of the aforementioned environmental factors is taken as a target environmental factor. The coefficient corresponding to the target environmental factor is multiplied by the standard deviation of the target environmental factor and divided by the standard deviation of the power plant operating data to obtain the sensitivity coefficient corresponding to the target environmental factor.

[0088] Based on the sensitivity coefficients described above, the comprehensive sensitivity coefficient is calculated.

[0089] Based on the comprehensive sensitivity coefficient and the preset sensitivity threshold, the environmental sensitivity analysis results of the target transformer area are determined.

[0090] In this embodiment, for example, daily power generation and power plant weather data under typical weather conditions over the past year can be obtained. For example, the power plant weather data includes light intensity X1, temperature X2, and humidity X3.

[0091] Then, using the power plant's weather data as the independent variable and daily power generation as the dependent variable, a multiple linear regression model was fitted, for example: Power Generation = 30 + 0.15X1 - 1.8X2 + 0.05X3. Next, the standard deviations of each variable were calculated, resulting in σ(X1) = 200 W / m', σ(X2) = 10℃, σ(X3) = 20%, and σ(Y) = 40 kWh. The sensitivity coefficients of each factor were then calculated.

[0092] illumination:

[0093] temperature:

[0094] humidity:

[0095] Finally, the average value of all sensitivity coefficients can be used as the comprehensive sensitivity coefficient. Alternatively, a weighted average of the sensitivity coefficients can be performed to obtain the comprehensive sensitivity coefficient. The weights of each sensitivity coefficient can be set according to actual needs, and this embodiment does not impose limitations on this.

[0096] The sensitivity threshold can be set according to actual needs, and this embodiment does not limit it. For example, the sensitivity coefficient is 0.7. If the overall sensitivity coefficient is greater than 0.7, the environmental sensitivity analysis result is that the environment is sensitive; otherwise, it is that the environment is not sensitive.

[0097] Regarding the above S103:

[0098] Specifically, in some embodiments, the step of constructing multiple candidate load prediction models based on power plant operation data and power plant weather data under typical weather conditions, and calculating the sudden change response capability analysis results corresponding to each candidate load prediction model, includes:

[0099] Extract routine operation data and sudden change operation data from power plant operation data and power plant weather data under the aforementioned typical weather conditions;

[0100] Using the aforementioned routine operating data, each of the selected load prediction models is trained.

[0101] Each of the selected load prediction models is used to process the power plant weather data in the sudden change operation data to obtain the corresponding sudden change load prediction results;

[0102] Based on the predicted sudden load and the power plant operation data in the sudden operation data, the analysis results of the sudden response capability of each of the selected load prediction models are determined.

[0103] For example, routine operation data can be extracted from the past year's operating data and weather data of each photovoltaic power station under typical weather conditions, and used as a training dataset. This data reflects the operating patterns of the power station under normal weather conditions. Using the routine operation data, initial time-series prediction models, feature interaction prediction models, and comprehensive prediction models can be trained respectively to obtain the candidate load prediction models.

[0104] Next, extract abrupt change operation data from power plant operation data under typical weather conditions and power plant weather data. Identify data on three sudden events encountered by the target transformer area in the past year. The first was a summer rainstorm lasting two hours, during which solar radiation intensity plummeted, resulting in a significant drop in power output. The second was a winter strong wind that affected the normal operation of the photovoltaic panels, causing large fluctuations in power output, lasting three hours. Data from one hour before and one hour after each sudden event is used as the abrupt change operation data, which includes the same parameters as the training dataset.

[0105] Next, the power plant weather data from the sudden change in operation data were processed using time-series forecasting models, feature interaction forecasting models, and integrated forecasting models to obtain corresponding sudden change in load forecasting results. For example, for the first sudden rainstorm, the time-series forecasting model, based on the time series patterns of previous normal operation data, predicted that the power output would drop to 40% of the normal level 30 minutes after the start of the rainstorm; the feature interaction forecasting model predicted that the power output would drop to 35% of the normal level; and the integrated forecasting model predicted that the power output would drop to 20% of the normal level. Similar predictions were made by the three models for the data from the two sudden changes, yielding their respective load forecasting results.

[0106] The load forecast results of each model are compared with the actual power generation data under sudden changes to evaluate the prediction accuracy of each model. Based on the prediction accuracy, the sudden change response capability analysis result of each model is determined. For example, the sudden change response capability analysis result of the model with the highest accuracy is determined to be strong; the sudden change response capability analysis result of the model with the lowest accuracy is determined to be weak; and the sudden change response capability analysis result of the model with medium accuracy is determined to be average.

[0107] Regarding S104 above:

[0108] In some embodiments, the step of screening the plurality of candidate load forecasting models based on the analysis results and the mutation response capability analysis results to obtain the optimal load forecasting model for the target power distribution area includes:

[0109] Based on the analysis results of each mutation response capability, the matching score of each of the selected load prediction models is initialized;

[0110] If the environmental sensitivity analysis result indicates that the system is sensitive to the environment, then the matching score of the comprehensive prediction model is updated using a preset first score.

[0111] If the time series stability analysis result is unstable, the matching score of the time series prediction model is updated using a preset second score.

[0112] If the spatial coupling degree analysis result is high, then the matching score of the feature interaction prediction model is updated using a preset third score.

[0113] The model with the highest matching score is selected as the optimal load prediction model for the target transformer area.

[0114] For example, suppose the environmental sensitivity analysis result is "sensitive to the environment"; the temporal stability analysis result is "unstable"; and the spatial coupling analysis result is "high spatial coupling". The comprehensive prediction model is sensitive to sudden changes; the temporal prediction model is insensitive to sudden changes; and the feature interaction prediction model is sensitive to sudden changes. Preset score settings: Preset first score (bonus points for the comprehensive model when environmentally sensitive): 10 points; Preset second score (bonus points for the temporal model when temporally unstable): 8 points; Preset third score (bonus points for the feature interaction model when spatial coupling is high): 12 points. Here, the preset scores can be set as needed, and this embodiment does not limit this.

[0115] Specifically, firstly, the matching scores of each model are initialized based on the results of the mutation response capability analysis: the comprehensive prediction model is sensitive to mutations and has an initial matching score of 5; the time series prediction model is not sensitive to mutations and has an initial matching score of 2; and the feature interaction prediction model is sensitive to mutations and has an initial matching score of 5.

[0116] Then, the environmental sensitivity analysis result indicates sensitivity to the environment, so a preset first score of 10 points is added to the comprehensive prediction model, updating the comprehensive prediction model's matching score to 5 + 10 = 15 points. The time series stability analysis result indicates instability, so a preset second score of 8 points is added to the time series prediction model, updating the time series prediction model's score to 2 + 8 = 10 points. The spatial coupling degree analysis result indicates high spatial coupling, so a preset third score of 12 points is added to the feature interaction prediction model. The feature interaction prediction model's score is updated to 5 + 12 = 17 points. The final scores are: Comprehensive prediction model: 15 points; Time series prediction model: 10 points; Feature interaction prediction model: 17 points. Comparing the matching scores of each model, the feature interaction prediction model has the highest score (17 points), therefore, the feature interaction prediction model is selected as the optimal load prediction model for the target transformer area.

[0117] This embodiment constructs a scientific model selection mechanism by combining the analysis results of the target transformer area's environmental sensitivity, temporal stability, and spatial coupling, as well as the analysis results of the sudden change response capability of each candidate load forecasting model. First, the matching score of each model is initialized based on its sudden change response capability. Then, the matching score of each model is updated specifically according to the characteristics of the transformer area using preset scores. By comprehensively evaluating the matching degree of each model under different dimensions, the model with the highest score is finally selected as the optimal load forecasting model, effectively improving the adaptability of the model to the load characteristics of the target transformer area, thereby significantly improving the accuracy and reliability of load forecasting.

[0118] Regarding the above S105:

[0119] In some embodiments, if the optimal load forecasting model is the time-series forecasting model; the step of using the optimal load forecasting model to process the power station deployment data, historical operation data, and weather data of each photovoltaic power station in the target area to obtain the load forecasting data of the target area includes:

[0120] The load sequence composed of the historical operating data of the power plant is decomposed to obtain the trend term, seasonal term, and residual term;

[0121] Based on the trend item, a trend prediction model is constructed; and the trend prediction value is calculated using the trend prediction model.

[0122] Using the dynamic Fourier coefficient correction algorithm, seasonal forecast values ​​are calculated based on the seasonal terms and the day-ahead weather forecast data;

[0123] Calculate the average value of the residual terms and use the average value as the predicted residual value;

[0124] The load forecast data is calculated based on the trend forecast, the seasonal forecast, and the residual forecast.

[0125] In this embodiment, the load sequence, composed of historical operating data of the power plant, is first decomposed to obtain a trend term, a seasonal term, and a residual term. Specifically, a historical output curve can be generated first; then, based on the historical output curve, the window duration corresponding to seasonal changes is adjusted to decompose the load sequence into a trend term, a seasonal term, and a residual term. The trend term reflects the overall trend of load change over time, the seasonal term reflects the periodic variation of load in different seasons, and the residual term includes other random fluctuations that cannot be explained by the trend term and the seasonal term. This decomposition method helps to more clearly analyze the components of the load, so that different forecasting methods can be used for different parts, thereby improving the accuracy of forecasting.

[0126] Then, based on the trend term, a trend prediction model is constructed; and using the trend prediction model, the trend prediction value is calculated. Here, the trend prediction model can be, for example, a constrained linear regression model, where the constraint condition is that the maximum power does not exceed the inverter's rated capacity. The inverter's rated capacity limits the maximum power generation capacity of the photovoltaic power plant. Considering this constraint condition when predicting the trend component ensures that the prediction results conform to actual physical limitations, improving the rationality and reliability of the prediction.

[0127] Next, using a dynamic Fourier coefficient correction algorithm, the amplitude of the seasonal component is calculated based on the seasonal term and the day-ahead weather forecast data. The day-ahead weather forecast data provides weather information for the next day, and weather conditions have a significant impact on the output of photovoltaic power plants. By combining weather forecasts and dynamically adjusting the amplitude of the seasonal component, the model can better adapt to weather changes and more accurately predict load changes under different weather conditions. Specifically, the calculation steps for the amplitude of the seasonal component are as follows: extract typical features corresponding to the weather forecast; calculate the basic Fourier coefficients based on the seasonal term; correct the basic Fourier coefficients according to the typical features; and output the corrected seasonal component amplitude corresponding to the weather type.

[0128] Finally, the average value of the residual terms is calculated and used as the residual prediction value. Based on the trend prediction value, seasonal prediction value, and residual prediction value, load forecast data is calculated. Specifically, the trend prediction value, seasonal prediction value, and residual prediction value can be weighted and superimposed to generate a load forecast curve, thus obtaining the load forecast data. In implementation, the weights can be set according to actual needs; this embodiment does not impose any limitations on this. This weighted superposition method comprehensively considers the prediction results of different components of the load, enabling a more comprehensive and accurate reflection of the load changes of distributed photovoltaic power stations, ultimately yielding a load forecast curve that conforms to the actual situation.

[0129] This embodiment decomposes the load sequence composed of historical operating data of photovoltaic power plants into trend components, seasonal components, and residual components, respectively capturing the overall trend of load change, periodic patterns, and random fluctuations. Then, an adaptation method is adopted for different components, which can accurately characterize the complex change pattern of photovoltaic load in the time dimension, significantly improving the accuracy, timeliness, and reliability of load forecasting for the target distribution area.

[0130] In some embodiments, the power plant weather data includes solar irradiance, and the power plant deployment data includes the total area of ​​the photovoltaic array and the module conversion efficiency; if the optimal load forecasting model is the feature interaction forecasting model, the process of using the optimal load forecasting model to process the power plant deployment data, historical operation data, and weather data of each photovoltaic power plant in the target area to obtain the load forecasting data of the target area includes:

[0131] For any target power station in the target area, a photovoltaic module diagram structure is generated based on the three-dimensional point cloud data of the photovoltaic modules in the target power station;

[0132] Based on the photovoltaic module diagram and the power plant weather data, calculate the inter-module shading matrix;

[0133] Calculate the occlusion area and occlusion time based on the inter-component shadow occlusion matrix;

[0134] The load forecast data is calculated based on the light intensity, the total area of ​​the photovoltaic array, the conversion efficiency of the modules, the shading area, and the shading time.

[0135] In this embodiment, 3D point cloud data can accurately represent the position and shape of photovoltaic modules in space. Converting it into a graph structure can better describe the spatial relationships and connection methods between modules. This graph structure can intuitively reflect the layout characteristics of the photovoltaic power station and provide an effective data structure for subsequent analysis of the interactions between modules.

[0136] For example, two photovoltaic modules A and B are deployed on a residential rooftop, with 3D point cloud coordinates of (0,0,3) and (2.5,0,3) respectively. Each module measures 2m × 1m, faces due south, and has a tilt angle of 30°. The total area of ​​the photovoltaic array is 4m². 2 The component conversion efficiency is 20%. At 10:00 AM on a certain day, the light intensity was 700 W / m². 2 Solar altitude angle 30°, azimuth angle 160°.

[0137] First, based on the 3D point cloud data, extract geometric features such as component height and spacing to generate a photovoltaic module diagram structure containing nodes (components A and B) and edges (spacing 2.5m, azimuth angle 0°).

[0138] Based on the graph structure edge attributes and the sun's position, the shadow length is calculated. Combined with the component width, it is determined that the shadow of component A covers 60% of the width of component B. An occlusion matrix is ​​generated for this time. Then, the occlusion periods of 1 hour each from 10:00-11:00 and 14:00-15:00 are statistically analyzed, with an average occlusion area of ​​1.2m per hour. 2 .

[0139] Finally, the daily load forecast of the target power plant is calculated by multiplying the irradiance, total area of ​​the photovoltaic array, and module conversion efficiency by the shading area and shading time, and by correcting the effective power generation area or by weighting by shading time.

[0140] This embodiment converts the three-dimensional point cloud data of the photovoltaic modules of the target power station into a graph structure, calculates the shading matrix between modules by combining the weather data of the power station, and further obtains the shading area and shading time. Finally, it integrates key data such as light intensity, total area of ​​photovoltaic array, and module conversion efficiency, and comprehensively considers the combined impact of photovoltaic module spatial layout, environmental factors and module characteristics on power generation. It effectively captures the shading relationship between modules and its dynamic changes, thereby significantly improving the accuracy and reliability of photovoltaic power generation load prediction for the target area.

[0141] In some embodiments, the integrated forecasting model includes a Transformer module and an XGBoost module; if the optimal load forecasting model is the integrated forecasting model, the process of using the optimal load forecasting model to process the power station deployment data, historical operation data, and weather data of each photovoltaic power station in the target area to obtain the load forecasting data of the target area includes:

[0142] Using the Transformer module, feature extraction and feature fusion are performed on the acquired power plant deployment data, historical power plant operation data, and power plant weather data to obtain a fused feature vector;

[0143] The fused feature vector is processed using the XGBoost module to output the load prediction result.

[0144] In this embodiment, during implementation, a comprehensive prediction model can be trained first. Specifically, an initial comprehensive prediction model and dataset are obtained, and the dataset is divided into a training set, a validation set, and a test set.

[0145] The initial integrated prediction model consists of a Transformer module and an XGBoost module. The dataset contains samples that do not change over time, such as power plant deployment data describing the physical attributes and geographical location of the power plants, as well as time-series features such as historical operating data and weather data arranged in chronological order. The historical operating data can cover indicators such as power generation and equipment parameters, while the weather data can include data such as solar intensity and temperature. In addition, the samples also need to be labeled with the predicted load or power generation value corresponding to a specific future time window.

[0146] The training set is used for initial training of the initial comprehensive prediction model, enabling the model to learn the mapping relationship between features and loads in the sample data. The validation set is used to monitor the model's performance during training, adjusting model parameters to prevent overfitting and ensuring good performance on unseen data. The test set is a dataset independent of the training and validation sets, used for evaluation to accurately reflect the model's performance in real-world applications. If the evaluation results on the test set are unsatisfactory—i.e., the model's prediction error is large or other evaluation metrics fail to meet expectations—it indicates potential problems with the model, such as insufficient model complexity, inappropriate data feature selection, overfitting, or underfitting. To address these issues, the model's structure and parameters can be further adjusted, or the data can be reprocessed and feature-engineered to continuously optimize the model and improve its prediction accuracy and stability.

[0147] XGBoost is a high-efficiency machine learning model based on the gradient boosting algorithm. During training, it builds a predictive model by iteratively fitting the residuals of the previous iteration. Evaluation metrics (such as mean squared error and mean absolute error) on the validation set provide feedback on the model's performance. Based on this feedback, the parameters of the XGBoost model (such as learning rate, tree depth, and minimum weight of leaf nodes) are adjusted to ensure the model fits the training data well and maintains good generalization ability on the validation data, thereby continuously optimizing the model's performance.

[0148] After a series of training, optimization, and evaluation processes, the comprehensive prediction model has learned the effective relationship between data features and load. At this point, the acquired distributed photovoltaic power station operation data, deployment data, and weather data are used to extract features through the Transformer module. These extracted features are then input into the optimized XGBoost module. The model processes the data based on the learned patterns and rules, ultimately outputting a prediction result for the distributed photovoltaic power station load.

[0149] During implementation, after obtaining the prediction results from the optimal load forecasting model, they can be compared with actual load data to calculate the error. Based on the error feedback, it can be determined whether the currently recommended forecasting model is suitable. If the error is large, it indicates that the current model may not be well adapted to the data, and the model recommendation can be optimized. The recommendation strategy can be adjusted, other models can be considered, or the parameters of the current model can be adjusted to improve the model's prediction accuracy.

[0150] Those skilled in the art will understand that in the above-described method of the specific embodiments, the order in which the steps are written does not imply a strict execution order, but constitutes no limitation on the implementation process. The specific execution order of each step should be determined by its function and possible internal logic.

[0151] It should be noted that in practical applications, all the above-described possible implementation methods can be combined in any way to form possible embodiments of this disclosure, and will not be described in detail here. The information (including but not limited to device information, user information, etc.) and data (including but not limited to data used for analysis, storage, and display) involved in this application are all information and data authorized by the user or fully authorized by all parties. The software tools or components appearing in the embodiments of this disclosure are merely illustrative examples and do not represent actual use.

[0152] Based on the same concept, this disclosure also provides a microservice-based transformer load prediction device, which corresponds one-to-one with the microservice-based transformer load prediction method in the above embodiments. Figure 2 A schematic diagram of the structure of the microservice-based transformer load forecasting device provided in this embodiment is shown. See also: Figure 2 As shown, the microservice-based transformer load forecasting device 200 provided in this embodiment is implemented based on a microservice architecture, which includes a data management microservice and a load forecasting and matching microservice; the device includes:

[0153] The analysis module 201 is used to obtain power station deployment data and extract power station operation data and power station weather data under typical weather conditions for each photovoltaic power station in the target area by calling the data management microservice; using the power station deployment data, power station operation data and power station weather data under typical weather conditions, the module performs time series stability analysis, spatial coupling analysis and environmental sensitivity analysis on the target area to obtain corresponding analysis results; based on the power station operation data and power station weather data under typical weather conditions, the module constructs multiple candidate load prediction models and calculates the sudden change response capability analysis results corresponding to each candidate load prediction model; wherein, the multiple candidate load prediction models are deployed to the microservice center through containerized deployment, including: time series prediction model, feature interaction prediction model and comprehensive prediction model;

[0154] Selection module 202 is used to filter the multiple candidate load prediction models by calling the load prediction matching microservice, based on the analysis results and the sudden change response capability analysis results, to obtain the optimal load prediction model for the target transformer area.

[0155] The prediction module 203 is used to process the power plant deployment data, historical operation data and weather data of each photovoltaic power plant in the target area using the optimal load prediction model to obtain the load prediction data of the target area.

[0156] In some embodiments, in the above-described apparatus, the analysis module 201, when performing time-series stability analysis on the target power station area and obtaining corresponding analysis results, is configured to: take each typical weather condition as a target typical weather condition and perform the following steps: perform stationarity verification on the time-series data composed of power station operation data and power station weather data under the target typical weather condition to obtain stationarity verification results; perform autocorrelation analysis to obtain autocorrelation analysis results; perform fluctuation analysis to obtain fluctuation analysis results; obtain target time-series stability analysis results based on the stationarity verification results, the autocorrelation analysis results, and the fluctuation analysis results; and determine the time-series stability analysis results of the target power station area based on each of the target time-series stability analysis results.

[0157] In some embodiments, in the above-described apparatus, the power plant deployment data includes the location information and terrain information of each photovoltaic power plant; the analysis module 201, when performing spatial coupling analysis on the target area to obtain corresponding analysis results, is used to: construct a spatial weight matrix based on the location information and terrain information; the spatial weight matrix is ​​used to represent the spatial proximity or spatial interaction relationship between each photovoltaic power plant; perform spatial autocorrelation analysis and local spatial correlation analysis on the target area using the operating data of each power plant and the spatial weight matrix to obtain corresponding analysis results; calculate the spatial coupling degree of the target area based on the spatial autocorrelation analysis results and the local spatial correlation analysis results; and determine the spatial coupling degree analysis result based on the spatial coupling degree and a preset coupling degree threshold.

[0158] In some embodiments, in the above-described apparatus, the analysis module 201, when performing environmental sensitivity analysis on the target power station area and obtaining corresponding analysis results, is configured to: fit a multiple linear regression model using power station operation data and power station weather data under various typical weather conditions; the power station weather data includes multiple environmental factors; calculate the standard deviation of each environmental factor and the power station operation data; take each environmental factor as a target environmental factor, multiply the coefficient corresponding to the target environmental factor by the standard deviation of the target environmental factor, and divide by the standard deviation of the power station operation data to obtain the sensitivity coefficient corresponding to the target environmental factor; calculate a comprehensive sensitivity coefficient based on each sensitivity coefficient; and determine the environmental sensitivity analysis result of the target power station area based on the comprehensive sensitivity coefficient and a preset sensitivity threshold.

[0159] In some embodiments, in the above-described apparatus, the analysis module 201, when constructing multiple candidate load prediction models based on power plant operation data and power plant weather data under typical weather conditions, and calculating the corresponding sudden change response capability analysis results for each candidate load prediction model, is configured to: extract regular operation data and sudden change operation data from the power plant operation data and power plant weather data under typical weather conditions; train each candidate load prediction model using the regular operation data; process the power plant weather data in the sudden change operation data using each candidate load prediction model to obtain the corresponding sudden change load prediction result; and determine the sudden change response capability analysis result for each candidate load prediction model based on the sudden change load prediction result and the power plant operation data in the sudden change operation data.

[0160] In some embodiments, in the above-described apparatus, the selection module 202 is configured to: initialize the matching score of each of the selected load prediction models based on the results of each mutation response capability analysis; if the environmental sensitivity analysis result indicates sensitivity to the environment, update the matching score of the comprehensive prediction model using a preset first score; if the time series stability analysis result indicates instability, update the matching score of the time series prediction model using a preset second score; if the spatial coupling degree analysis result indicates high spatial coupling, update the matching score of the feature interaction prediction model using a preset third score; and select the model with the highest matching score as the optimal load prediction model for the target transformer area.

[0161] In some embodiments, in the above apparatus, if the optimal load forecasting model is the time-series forecasting model; the forecasting module 203 is configured to: decompose the load sequence composed of the historical operating data of the power plant to obtain a trend term, a seasonal term, and a residual term; construct a trend forecasting model based on the trend term; calculate a trend forecast value using the trend forecasting model; calculate a seasonal forecast value based on the seasonal term and the day-ahead weather forecast data using a dynamic Fourier coefficient correction algorithm; calculate the average value of the residual term and use the average value as the residual forecast value; and calculate the load forecasting data based on the trend forecast value, the seasonal forecast value, and the residual forecast value.

[0162] In some embodiments, in the above-described apparatus, the power plant weather data includes solar irradiance, and the power plant deployment data includes the total area of ​​the photovoltaic array and the module conversion efficiency. If the optimal load prediction model is the feature interaction prediction model, the prediction module 203 is configured to: generate a photovoltaic module diagram structure based on the three-dimensional point cloud data of the photovoltaic modules in the target power plant for any target power plant in the target area; calculate the inter-module shading matrix based on the photovoltaic module diagram structure and the power plant weather data; calculate the shading area and shading time based on the inter-module shading matrix; and calculate the load prediction data based on the solar irradiance, the total area of ​​the photovoltaic array, the module conversion efficiency, the shading area, and the shading time.

[0163] In some embodiments, in the above-described apparatus, the integrated forecasting model includes a Transformer module and an XGBoost module; if the optimal load forecasting model is the integrated forecasting model, the forecasting module 203 is used to: use the Transformer module to perform feature extraction and feature fusion on the acquired power plant deployment data, historical power plant operation data, and power plant weather data to obtain a fused feature vector; and use the XGBoost module to process the fused feature vector and output the load forecasting result.

[0164] This invention provides a microservice-based load forecasting device for transformer substations. By establishing a microservice center comprising microservices such as photovoltaic power plant data management and load forecasting matching, and by deploying time-series forecasting models, feature-interaction forecasting models, and integrated forecasting models to the microservice center using containerization, the system achieves modularity and flexible deployment. This facilitates independent maintenance and updates of each microservice and model, improves system scalability and maintainability, and effectively addresses diverse business needs and changes. Simultaneously, by acquiring power plant deployment data and typical weather data for each photovoltaic power plant in the target substation area, the device conducts analysis on time-series stability, spatial coupling, and environmental sensitivity. It also constructs multiple candidate load forecasting models, including time-series forecasting, feature-interaction forecasting, and integrated forecasting models, and analyzes their ability to respond to sudden changes. Based on the analysis results, the optimal load forecasting model is selected, and finally, this model is used to process the data to obtain load forecast data. As can be seen, this solution achieves the goal of selecting appropriate prediction models based on the characteristics of different distribution areas. It solves the problem that the existing unified prediction method is difficult to adapt to the diversity of distribution areas, cannot deeply explore the characteristics of different distribution areas, and cannot make targeted predictions based on the characteristics of different distribution areas, resulting in poor accuracy of distribution area load prediction. It provides a more reliable basis for the stable operation, rational scheduling, and optimized allocation of power resources of distributed photovoltaic systems, thereby improving the economic benefits and operating efficiency of the entire distributed photovoltaic system within the distribution area.

[0165] Specific limitations regarding the microservice-based distribution transformer load forecasting device can be found in the limitations of the microservice-based distribution transformer load forecasting method described above, and will not be repeated here. Each module in the aforementioned microservice-based distribution transformer load forecasting device can be implemented entirely or partially through software, hardware, or a combination thereof. These modules can be embedded in or independent of the processor in a computer device in hardware form, or stored in the memory of a computer device in software form, so that the processor can call and execute the corresponding operations of each module.

[0166] Figure 3 A schematic diagram of the structure of an electronic device provided in an embodiment of this disclosure is shown. Figure 3 As shown, at the hardware level, this electronic device includes a processor, and optionally also includes an internal bus, a network interface, and memory. The memory may include main memory, such as high-speed random-access memory (RAM), or it may include non-volatile memory, such as at least one disk drive. Of course, this electronic device may also include other hardware required for other business operations.

[0167] The processor, network interface, and memory can be interconnected via an internal bus, which can be an ISA (Industry Standard Architecture) bus, a PCI (Peripheral Component Interconnect) bus, or an EISA (Extended Industry Standard Architecture) bus, etc. The bus can be divided into address bus, data bus, control bus, etc. For ease of representation, Figure 3 The symbol is represented by a single double-headed arrow, but this does not mean that there is only one bus or one type of bus.

[0168] Memory is used to store programs. Specifically, programs may include program code, which includes computer operation instructions. Memory may include main memory and non-volatile memory, and provides instructions and data to the processor.

[0169] The processor reads the corresponding computer program from non-volatile memory into memory and then runs it, forming a microservice-based area load forecasting device at the logical level. The processor executes the program stored in memory and specifically performs the aforementioned methods.

[0170] The processor may be an integrated circuit chip with signal processing capabilities. In implementation, the steps of the above methods can be completed by integrated logic circuits in the processor's hardware or by instructions in software form. The processor can be a general-purpose processor, including a Central Processing Unit (CPU), a Network Processor (NP), etc.; it can also be a Digital Signal Processor (DSP), an Application Specific Integrated Circuit (ASIC), a Field-Programmable Gate Array (FPGA), or other programmable logic devices, discrete gate or transistor logic devices, or discrete hardware components. It can implement or execute the methods, steps, and logic block diagrams disclosed in the embodiments of this disclosure. The general-purpose processor can be a microprocessor or any conventional processor. The steps of the methods disclosed in the embodiments of this disclosure can be directly embodied in the execution of a hardware decoding processor, or executed by a combination of hardware and software modules in the decoding processor. The software modules can reside in random access memory, flash memory, read-only memory, programmable read-only memory, electrically erasable programmable memory, registers, or other mature storage media in the art. The storage medium is located in the memory, and the processor reads the information in the memory and, in conjunction with its hardware, completes the steps of the above method.

[0171] The electronic device can execute the microservice-based distribution area load forecasting method provided in several embodiments of this disclosure, and is implemented as a microservice-based distribution area load forecasting device. Figure 2 The functions of the embodiments shown are not described in detail here.

[0172] This disclosure also proposes a computer-readable storage medium that stores one or more programs, the programs including instructions that, when executed by an electronic device including multiple applications, enable the electronic device to perform the microservice-based transformer load forecasting method provided in several embodiments of this disclosure.

[0173] Those skilled in the art will understand that embodiments of this disclosure can be provided as methods, systems, or computer program products. Therefore, this disclosure can take the form of a completely hardware embodiment, a completely software embodiment, or an embodiment combining software and hardware aspects. Furthermore, this disclosure can take the form of a computer program product embodied on one or more computer-usable storage media (including, but not limited to, disk storage, CD-ROM, optical storage, etc.) containing computer-usable program code.

[0174] This disclosure is described with reference to flowchart illustrations and / or block diagrams of methods, apparatus (systems), and computer program products according to embodiments of this disclosure. It will be understood that each block of the flowchart illustrations and / or block diagrams, and combinations of blocks in the flowchart illustrations and / or block diagrams, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, special-purpose computer, embedded processor, or other programmable data processing apparatus to produce a machine, such that the instructions, which execute via the processor of the computer or other programmable data processing apparatus, create a machine for implementing the flowchart illustrations and / or block diagrams. Figure 1 One or more processes and / or boxes Figure 1 A device that provides the functions specified in one or more boxes.

[0175] These computer program instructions may also be stored in a computer-readable storage medium that can direct a computer or other programmable data processing device to function in a particular manner, such that the instructions stored in the computer-readable storage medium produce an article of manufacture including instruction means, which are implemented in a process Figure 1 One or more processes and / or boxes Figure 1 The function specified in one or more boxes.

[0176] These computer program instructions may also be loaded onto a computer or other programmable data processing equipment to cause a series of operational steps to be performed on the computer or other programmable equipment to produce a computer-implemented process, thereby providing instructions that execute on the computer or other programmable equipment for implementing the process. Figure 1 One or more processes and / or boxes Figure 1 The steps of the function specified in one or more boxes.

[0177] In a typical configuration, a computing device includes one or more processors (CPU), input / output interfaces, network interfaces, and memory.

[0178] Memory may include non-persistent storage in computer-readable media, such as random access memory (RAM) and / or non-volatile memory, such as read-only memory (ROM) or flash RAM. Memory is an example of computer-readable media.

[0179] Computer-readable media include both permanent and non-permanent, removable and non-removable media that can store information by any method or technology. Information can be computer-readable instructions, data structures, modules of programs, or other data. Examples of computer storage media include, but are not limited to, phase-change memory (PRAM), static random access memory (SRAM), dynamic random access memory (DRAM), other types of random access memory (RAM), read-only memory (ROM), electrically erasable programmable read-only memory (EEPROM), flash memory or other memory technologies, CD-ROM, digital versatile optical disc (DVD) or other optical storage, magnetic tape, magnetic tape, disk storage or other magnetic storage devices, or any other non-transferable medium that can be used to store information accessible by a computing device. As defined herein, computer-readable media does not include transient computer-readable media, such as modulated data signals and carrier waves.

[0180] It should also be noted that the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such a process, method, article, or apparatus. Without further limitation, an element defined by the phrase "comprising one..." does not exclude the presence of other identical elements in the process, method, article, or apparatus that includes said element.

[0181] Those skilled in the art will understand that embodiments of this disclosure can be provided as methods, systems, or computer program products. Therefore, this disclosure can take the form of a completely hardware embodiment, a completely software embodiment, or an embodiment combining software and hardware aspects. Furthermore, this disclosure can take the form of a computer program product embodied on one or more computer-usable storage media (including, but not limited to, disk storage, CD-ROM, optical storage, etc.) containing computer-usable program code.

[0182] The above are merely embodiments of this disclosure and are not intended to limit the scope of this disclosure. Various modifications and variations can be made to this disclosure by those skilled in the art. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of this disclosure should be included within the scope of the claims of this disclosure.

Claims

1. A microservice-based method for predicting transformer load, characterized in that, The method is implemented based on a microservice architecture, which includes a data management microservice and a load prediction and matching microservice; the method includes: By calling the data management microservice, the deployment data of each photovoltaic power station in the target area is obtained, and the operation data and weather data of the power station under typical weather conditions are extracted. Using the power plant deployment data, as well as the power plant operation data and power plant weather data under typical weather conditions, the target power area is subjected to time series stability analysis, spatial coupling analysis and environmental sensitivity analysis, and the corresponding analysis results are obtained. Based on the power plant operation data and power plant weather data under the typical weather conditions, multiple candidate load prediction models are constructed, and the analysis results of the sudden change response capability corresponding to each candidate load prediction model are calculated. Among them, the multiple candidate load prediction models are deployed to the microservice center through containerized deployment, including: time series prediction model, feature interaction prediction model and comprehensive prediction model. By calling the load forecasting matching microservice, the multiple candidate load forecasting models are screened based on the analysis results and the sudden change response capability analysis results to obtain the optimal load forecasting model for the target transformer area. Using the optimal load forecasting model, the power station deployment data, historical operation data, and weather data of each photovoltaic power station in the target area are processed to obtain the load forecasting data of the target area.

2. The method according to claim 1, characterized in that, A time-series stability analysis was performed on the target station area, and the corresponding analysis results were obtained, including: For each type of typical weather, take it as the target typical weather and perform the following steps: The time series data consisting of power plant operation data and power plant weather data under the target typical weather conditions are subjected to stationarity check to obtain stationarity check results; autocorrelation analysis is performed to obtain autocorrelation analysis results; and fluctuation analysis is performed to obtain fluctuation analysis results. Based on the stationarity check results, the autocorrelation analysis results, and the fluctuation analysis results, the target time series stability analysis results are obtained. Based on the time series stability analysis results of each target, the time series stability analysis results of the target station area are determined.

3. The method according to claim 1, characterized in that, The power station deployment data includes the location information and terrain information of each photovoltaic power station; Spatial coupling analysis was performed on the target station area to obtain the corresponding analysis results, including: A spatial weight matrix is ​​constructed based on the location information and terrain information of each of the aforementioned locations; the spatial weight matrix is ​​used to represent the spatial proximity or spatial interaction relationship between each of the photovoltaic power stations. Using the operating data of each power station and the spatial weight matrix, spatial autocorrelation analysis and local spatial correlation analysis are performed on the target transformer area to obtain the corresponding analysis results. The spatial coupling degree of the target station area is calculated based on the results of spatial autocorrelation analysis and local spatial correlation analysis. The spatial coupling degree analysis result is determined based on the spatial coupling degree and the preset coupling degree threshold.

4. The method according to claim 1, characterized in that, An environmental sensitivity analysis was performed on the target transformer area, and the corresponding analysis results were obtained, including: Using power plant operation data and power plant weather data under various typical weather conditions, a multiple linear regression model was fitted to obtain the power plant weather data, which includes a variety of environmental factors. Calculate the standard deviation of each environmental factor and power plant operating data; Each of the aforementioned environmental factors is taken as a target environmental factor. The coefficient corresponding to the target environmental factor is multiplied by the standard deviation of the target environmental factor and divided by the standard deviation of the power plant operating data to obtain the sensitivity coefficient corresponding to the target environmental factor. Based on the sensitivity coefficients described above, the comprehensive sensitivity coefficient is calculated. Based on the comprehensive sensitivity coefficient and the preset sensitivity threshold, the environmental sensitivity analysis results of the target transformer area are determined.

5. The method according to claim 1, characterized in that, The process involves constructing multiple candidate load forecasting models based on power plant operation data and weather data under typical weather conditions, and calculating the sudden change response capability analysis results corresponding to each candidate load forecasting model, including: Extract routine operation data and sudden change operation data from power plant operation data and power plant weather data under the aforementioned typical weather conditions; Using the aforementioned routine operating data, each of the selected load prediction models is trained. Each of the selected load prediction models is used to process the power plant weather data in the sudden change operation data to obtain the corresponding sudden change load prediction results; Based on the predicted sudden load and the power plant operation data in the sudden operation data, the analysis results of the sudden response capability of each of the selected load prediction models are determined.

6. The method according to any one of claims 1-5, characterized in that, The process of filtering the multiple candidate load forecasting models based on the analysis results and the analysis results of the sudden change response capabilities to obtain the optimal load forecasting model for the target power distribution area includes: Based on the analysis results of each mutation response capability, the matching score of each of the selected load prediction models is initialized; If the environmental sensitivity analysis result indicates that the system is sensitive to the environment, then the matching score of the comprehensive prediction model is updated using a preset first score. If the time series stability analysis result is unstable, the matching score of the time series prediction model is updated using a preset second score. If the spatial coupling degree analysis result is high, then the matching score of the feature interaction prediction model is updated using a preset third score. The model with the highest matching score is selected as the optimal load prediction model for the target transformer area.

7. The method according to claim 1, characterized in that, If the optimal load forecasting model is the time-series forecasting model; the process of using the optimal load forecasting model to process the power station deployment data, historical operation data, and weather data of each photovoltaic power station in the target area to obtain the load forecasting data of the target area includes: The load sequence composed of the historical operating data of the power plant is decomposed to obtain the trend term, seasonal term, and residual term; Based on the trend item, a trend prediction model is constructed; and the trend prediction value is calculated using the trend prediction model. Using the dynamic Fourier coefficient correction algorithm, seasonal forecast values ​​are calculated based on the seasonal terms and the day-ahead weather forecast data; Calculate the average value of the residual terms and use the average value as the predicted residual value; The load forecast data is calculated based on the trend forecast, the seasonal forecast, and the residual forecast.

8. The method according to claim 1, characterized in that, The power station weather data includes solar irradiance, and the power station deployment data includes the total area of ​​the photovoltaic array and the module conversion efficiency. If the optimal load forecasting model is the feature interaction forecasting model, the optimal load forecasting model is used to process the power station deployment data, historical operation data, and weather data of each photovoltaic power station in the target area to obtain the load forecasting data for the target area, including: For any target power station in the target area, a photovoltaic module diagram structure is generated based on the three-dimensional point cloud data of the photovoltaic modules in the target power station; Based on the photovoltaic module diagram and the power plant weather data, calculate the inter-module shading matrix; Calculate the occlusion area and occlusion time based on the inter-component shadow occlusion matrix; The load forecast data is calculated based on the light intensity, the total area of ​​the photovoltaic array, the conversion efficiency of the modules, the shading area, and the shading time.

9. The method according to claim 1, characterized in that, The integrated forecasting model includes a Transformer module and an XGBoost module; if the optimal load forecasting model is the integrated forecasting model, the optimal load forecasting model is used to process the power station deployment data, historical operation data, and weather data of each photovoltaic power station in the target area to obtain the load forecasting data of the target area, including: Using the Transformer module, feature extraction and feature fusion are performed on the acquired power plant deployment data, historical power plant operation data, and power plant weather data to obtain a fused feature vector; The fused feature vector is processed using the XGBoost module to output the load prediction result.

10. A microservice-based transformer area load forecasting device, characterized in that, The device is implemented based on a microservice architecture, which includes a data management microservice and a load prediction and matching microservice; the device includes: The analysis module is used to acquire power station deployment data and extract power station operation data and weather data under typical weather conditions for each photovoltaic power station in the target area by calling the data management microservice. Using the power station deployment data, power station operation data, and power station weather data under typical weather conditions, the module performs time-series stability analysis, spatial coupling analysis, and environmental sensitivity analysis on the target area, obtaining corresponding analysis results. Based on the power station operation data and power station weather data under typical weather conditions, multiple candidate load prediction models are constructed, and the sudden change response capability analysis results corresponding to each candidate load prediction model are calculated. These multiple candidate load prediction models are deployed to the microservice center using a containerized deployment method, including: a time-series prediction model, a feature interaction prediction model, and a comprehensive prediction model. The selection module is used to filter the multiple candidate load prediction models by calling the load prediction matching microservice, based on the analysis results and the sudden change response capability analysis results, to obtain the optimal load prediction model for the target transformer area. The forecasting module is used to process the power plant deployment data, historical operation data and weather data of each photovoltaic power plant in the target area using the optimal load forecasting model to obtain the load forecasting data of the target area.

Images