Layered heavy overload risk early warning method and device for power distribution network, terminal equipment and storage medium

Abstract

The invention provides a layered heavy overload risk early warning method and device for a power distribution network, terminal equipment and a storage medium, and belongs to the technical field of electric power. The method comprises the following steps: predicting a first prediction load of a transformer substation through an operation feature vector used for representing a load fluctuation trend and a power generation fluctuation trend of the power distribution network; the method comprises the following steps: determining a first overload risk of each transformer substation, predicting a second predicted load of each line according to an operation feature vector and the first overload risk, determining a second overload risk of each line, predicting a third predicted load of each transformer according to the operation feature vector and the second overload risk, and determining a second overload risk of each line. Determining the third overload risk of each transformer; and finally, according to the third overload risk, the cascade overload probability is calculated. And generating a heavy overload risk early warning signal of the power distribution network according to the cascade overload probability. Therefore, the problem of low accuracy of the current heavy overload risk prediction result is solved.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of electric power, and in particular to a layered overload risk early warning method and device for a power distribution network, a terminal device and a storage medium. BACKGROUND

[0002] With the large-scale access of new elements with high uncertainty such as distributed photovoltaic, wind power, electric vehicle charging piles in the power distribution network, the load characteristics of the power grid are undergoing profound changes. The traditionally relatively stable load curve now presents stronger randomness, intermittency and volatility, which significantly increases the overload risk of key devices such as distribution transformers and feeders, and has become a core hidden danger directly threatening the safe and stable operation of regional power grids. Under this background, accurate prediction of power grid device load is undoubtedly a key technology for actively identifying and preventing overload accidents.

[0003] Currently, the industry generally relies on time series prediction models based on supervised learning, such as long short-term memory networks (LSTM) or time convolution networks (TCN). The core idea of this type of technology path is to collect the historical operation data of the target device (such as a transformer or a line) as the model input, train the model to learn the static mapping relationship from the historical sequence to the future value, and finally output the load prediction value of a specific period in the future, and judge the overload risk by comparing with the rated capacity threshold of the device.

[0004] However, this traditional paradigm has a fundamental limitation: its prediction perspective is isolated and local. It processes the prediction target as an independent time series signal, while generally ignoring the influence of the operating state and topological association of devices at various levels from distributed power sources (generation end) to user loads (consumption end) on the overall overload risk. The existing method ignores the dynamic interaction mechanism of the power distribution network across levels and devices, resulting in low accuracy of the current overload risk prediction results. SUMMARY

[0005] The present application provides a layered overload risk early warning method and device for a power distribution network, a terminal device and a storage medium, which can solve the problem of low accuracy of overload risk prediction results in the prior art.

[0006] An embodiment of the present application provides a layered overload risk early warning method for a power distribution network, comprising: obtaining an operating feature vector for representing load fluctuation trends and power generation fluctuation trends of the power distribution network and a topological structure of the power distribution network under a plurality of preset time periods; predicting a first predicted load of each substation in the power distribution network according to the operating feature vector, and determining a first overload risk of each substation according to the first predicted load; According to the topology structure, determine the transformer connected to each line, and according to the second overload risk of each transformer corresponding to the line and the operation characteristic vector, predict the third predicted load of each transformer, and according to the third predicted load, determine the third overload risk of each transformer; According to the topology structure, determine the transformer connected to each line, and according to the second overload risk of each transformer corresponding to the line and the operation characteristic vector, predict the third predicted load of each transformer, and according to the third predicted load, determine the third overload risk of each transformer; According to the third overload risk, calculate the cascade overload probability for characterizing the failure of the power distribution network caused by the overload event; According to the cascade overload probability, generate the overload risk warning signal of the power distribution network.

[0007] Further, the operation characteristic vector for characterizing the load fluctuation trend and the power generation fluctuation trend of the power distribution network under a plurality of preset time periods comprises: Obtain the device parameters of the power distribution network, the historical operation data and the historical environment data under a plurality of historical time nodes; According to the historical operation data, the historical environment data and the device parameters, time alignment and space alignment are performed to construct a space-time correlation matrix; According to the preset time period and the space-time correlation matrix, extract the operation characteristic vector for characterizing the load fluctuation trend and the power generation fluctuation trend of the power distribution network under each time period.

[0008] Further, after obtaining the operation characteristic vector for characterizing the load fluctuation trend and the power generation fluctuation trend of the power distribution network under a plurality of preset time periods and the topology structure of the power distribution network, it further comprises: Obtain the preset prediction time, the preset plurality of candidate load prediction models and the historical prediction error rate of each candidate load prediction model; According to the operation characteristic vector, the historical prediction error rate and the prediction time, respectively construct the first state sequence of the transformer level, the second state sequence of the line level and the third state sequence of the transformer level; With the maximum of prediction accuracy, calculation efficiency and performance stability as the goal, according to the first state sequence, the second state sequence and the third state sequence, select the first load prediction model for predicting the first predicted load, the second load prediction model for predicting the second predicted load and the third load prediction model for predicting the third predicted load from the candidate load prediction model.

[0009] Further, the obtaining of the preset prediction time, the preset plurality of candidate load prediction models and the historical prediction error rates of the candidate load prediction models comprises: obtaining historical prediction data of the preset candidate load prediction models; wherein the historical prediction data comprises a historical operation feature vector, a historical load prediction result and an error rate of the historical load prediction result; calculating the similarity of the operation feature vector and the historical operation feature vector, and taking the historical prediction data corresponding to a plurality of historical operation feature vectors with a similarity greater than a preset threshold as target historical prediction data of each candidate load prediction model; calculating the historical prediction error rate of each candidate load prediction model according to the error rates in the plurality of target historical prediction data.

[0010] Further, the prediction of the second predicted load of each line according to the first overload risk of the substation corresponding to each line and the operation feature vector, and the determination of the second overload risk of each line according to the second predicted load, comprises: inputting the operation feature vector into a second load prediction model to enable the second load prediction model to output the second predicted load of each line, and calculating the initial second overload risk of each line segment according to the second predicted load; weighting and summing the initial second overload risk and the related first overload risk to determine the second overload risk of each line.

[0011] Further, the prediction of the third predicted load of each transformer according to the second overload risk of the line corresponding to each transformer and the operation feature vector, and the determination of the third overload risk of each transformer according to the third predicted load, comprises: inputting the operation feature vector into a third load prediction model to enable the third load prediction model to output the third predicted load of each line, and calculating the initial third overload risk of each line segment according to the third predicted load; weighting and summing the initial third overload risk and the related second overload risk to determine the third overload risk of each line.

[0012] An embodiment of the present application also provides a layered overload risk early warning device of a power distribution network, comprising: a data acquisition module configured to acquire an operation feature vector for representing load fluctuation trend and power generation fluctuation trend of the power distribution network and a topology structure of the power distribution network under a plurality of preset time periods; a first prediction module configured to predict a first predicted load of each substation in the power distribution network according to the operation feature vector, and determine a first overload risk of each substation according to the first predicted load. a second prediction module configured to determine substations connected by each line according to the topology, predict a second predicted load of each line according to the first overload risk of the substation corresponding to each line and the operation feature vector, and determine a second overload risk of each line according to the second predicted load; a third prediction module configured to determine lines connected by each transformer according to the topology, predict a third predicted load of each transformer according to the second overload risk of the line corresponding to each transformer and the operation feature vector, and determine a third overload risk of each transformer according to the third predicted load; a risk assessment module configured to calculate a cascading overload probability for representing the existence of an overload event of the power distribution network according to the third overload risk; a risk warning module configured to generate an overload risk warning signal of the power distribution network according to the cascading overload probability.

[0013] Further, the data acquisition module acquires an operation feature vector for representing a load fluctuation trend and a power generation fluctuation trend of the power distribution network in a preset time period, including: acquiring device parameters, historical operation data and historical environmental data of the power distribution network at a plurality of historical time nodes; aligning the historical operation data, the historical environmental data and the device parameters in time and space, and constructing a time-space correlation matrix; extracting the operation feature vector for representing the load fluctuation trend and the power generation fluctuation trend of the power distribution network in each time period according to the preset time period and the time-space correlation matrix.

[0014] The application further provides a terminal device, including: one or more processors; a memory coupled to the processor, configured to store one or more programs; When the one or more programs are executed by the one or more processors, the one or more processors implement the layered overload risk warning method of the power distribution network as described in the above embodiments.

[0015] The application further provides a computer readable storage medium having a computer program stored thereon, and the computer program is executed by a processor to implement the layered overload risk warning method of the power distribution network as described in the above embodiments.

[0016] The application has the following beneficial effects: The application provides a layered overload risk early warning method, device, terminal equipment and storage medium of a power distribution network. BRIEF DESCRIPTION OF DRAWINGS

[0017] In order to more clearly illustrate the technical solutions of the present application, the following will briefly introduce the drawings needed to be used in the embodiments. Obviously, the drawings in the following description are only some embodiments of the present application, and other drawings can also be obtained by those skilled in the art without any creative effort.

[0018] Figure 1 is a flow diagram of a layered overload risk early warning method of a power distribution network provided by an embodiment of the present application; Figure 2 is a structural diagram of a layered overload risk early warning device of a power distribution network provided by an embodiment of the present application; Figure 3 is a structural diagram of a terminal equipment provided by an embodiment of the present application; Figure 4 is a load prediction result of experimental data of an embodiment of the present application; Figure 5 is a layered overload risk prediction result of experimental data of an embodiment of the present application. DETAILED DESCRIPTION

[0019] In order to make the objects, technical solutions and advantages of the present application clearer, the following will clearly and completely describe the technical solutions in the present application in conjunction with the accompanying drawings of the embodiments of the present application. Obviously, the described embodiments are only a part of the embodiments of the present application, rather than all the embodiments of the present application. Based on the embodiments in the present application, all other embodiments obtained by those of ordinary skill in the art without creative work fall within the scope of protection of the present application.

[0020] Unless otherwise defined, all technical and scientific terms used herein have the same meaning as commonly understood by one of ordinary skill in the art to which this application belongs; the terminology used herein is for the purpose of describing particular embodiments only and is not intended to be limiting of this application; the terms "comprising," "comprises" and "including" and "has" and any variations thereof used herein are intended to cover a non-exclusive inclusion.

[0021] In the description of the embodiments of the present application, the technical terms "first", "second", etc. are only used to distinguish different objects, and cannot be understood as indicating or implying relative importance or implicitly indicating the number, specific order or primary and secondary relationship of the indicated technical features. In the description of the embodiments of the present application, the meaning of "a plurality of" is two or more, unless otherwise explicitly and specifically limited.

[0022] Reference herein to "an embodiment" means that a particular feature, structure, or characteristic described in connection with the embodiment can be included in at least one embodiment of the present application. The appearance of the phrase in various places in the specification does not necessarily all refer to the same embodiment, nor is it necessarily independent or alternative embodiments to each other. Those skilled in the art explicitly and implicitly understand that the embodiments described herein can be combined with other embodiments.

[0023] In the description of the embodiments of the present application, the term "and / or" is only a description of the association relationship of the associated objects, which means that there can be three relationships, for example, A and / or B, which can represent the three cases of A alone, A and B together, and B alone. In addition, the character " / " in this paper generally represents that the front and rear associated objects are a "or" relationship.

[0024] In the description of the embodiments of the present application, the term "a plurality of" refers to two or more (including two), and similarly, "a plurality of groups" refers to two or more groups (including two groups), and "a plurality of pieces" refers to two or more pieces (including two pieces).

[0025] In the description of the embodiments of the present application, unless otherwise explicitly specified and limited, the technical terms "mounting", "connection", "connecting", "fixing" and the like should be understood in a broad sense, for example, can be fixed connection, can also be detachable connection, or integral; can be mechanical connection, can also be electrical connection; can be direct connection, can also be indirect connection through an intermediate medium, can be internal communication of two elements or interaction relationship of two elements. For those skilled in the art, the specific meaning of the above terms in the embodiments of the present application can be understood according to the specific circumstances.

[0026] Referring to Figure 1 To solve the problems in the prior art, an embodiment of the present application provides a layered overload risk early warning method for a power distribution network, comprising: S1, obtaining an operation feature vector for representing load fluctuation trend and power generation fluctuation trend of the power distribution network and a topology structure of the power distribution network under a plurality of preset time periods; Preferably, the operation feature vector for representing the load fluctuation trend and the power generation fluctuation trend of the power distribution network under the plurality of preset time periods comprises: obtaining device parameters of the power distribution network, historical operation data and historical environment data at a plurality of historical time nodes; performing time alignment and space alignment according to the historical operation data, the historical environment data and the device parameters to construct a space-time correlation matrix; extracting the operation feature vector for representing the load fluctuation trend and the power generation fluctuation trend of the power distribution network under each time period according to the plurality of preset time periods and the space-time correlation matrix.

[0027] In a preferred embodiment of the present application, through a power data center and the like platform, a plurality of source heterogeneous data such as power distribution network framework topology data (containing line impedance, transformer capacity and the like parameters), device account information (containing device commissioning time, maintenance record), metering operation data (containing real-time load rate of main transformer, feeder and transformer), distributed power supply data (containing output prediction of photovoltaic and wind power) and meteorological data (containing temperature, illumination, wind speed) are accessed.

[0028] Among them: topology data: including the network connection relationship of station, line and transformer three levels, stored in the form of graph database, supporting fast query and hierarchical association; device parameters: including the capacity, voltage level, operation life and the like information of main transformer, feeder, distribution transformer and the like devices; operation data: including load power, voltage, current, power factor and the like, the sampling frequency is set to 1 hour / time; distributed power supply data: including photovoltaic output, wind power output and the like, the sampling frequency is 1 hour / time; meteorological data: including wind speed, illumination intensity, temperature, humidity and the like, transmitted to an edge computing node or a cloud server in real time by using an industrial Internet of Things protocol (such as MQTT / OPC UA).

[0029] It should be noted that, in order to ensure the accuracy of subsequent calculation, the embodiment adopts a statistical method to detect outliers of the original load and output data. For any observation value at any time , if it deviates from the sliding mean value by more than a set threshold (such as 3 times the standard deviation ), it is determined to be an outlier: For the data points determined to be outliers, linear interpolation or adjacent window average can be used for repair.

[0030] Further, the accessed data is uniformly stored in a time series database, and is preliminarily aligned based on timestamps to construct a space topology + time series spatiotemporal correlation matrix. In order to meet the high-precision requirement of the prediction model on the time series data, the feature data set used for model training and prediction needs to be further fine-tuned for timestamp alignment. Different sources of data (such as metering data, meteorological data) may have different sampling frequencies (such as 15 minutes / time and 1 hour / time) or data missing, resulting in incomplete time axis. The following operations are performed to ensure accurate matching of data in the time dimension: A unified standard time axis is constructed , with 1 hour as the sampling step, as the baseline time series for the prediction task.

[0031] For data with a sampling frequency higher than 1 hour, down-sampling (such as taking the average value or interpolation) is used to unify it to 1 hour step; for missing values, linear interpolation or prediction method based on adjacent data is used for completion.

[0032] For data with timestamp deviation, re-alignment is performed based on the time axis to ensure that all feature data have corresponding values at each time point. This step ensures that the data input into the prediction model is highly consistent in the time dimension, improving the stability of model training and the accuracy of prediction results.

[0033] Further, in order to eliminate the influence of different dimensions, normalization is performed on various types of data. The minimum-maximum normalization method is used to convert the original data to standard values within the interval : ; wherein , are the minimum and maximum values in the original data .

[0034] Finally, at the time series level, multi-dimensional features of daily, weekly, and monthly scales are extracted. Let the sequence be , then the window size is​​ Under the condition of ; ; where is the moving average value at time , representing the average load or output level within the window size , and is the difference feature at time , representing the change amount of adjacent time point observations, , are the observations at time and , respectively. These features help capture trends, periodicity, and volatility when inputted into the prediction model subsequently.

[0035] It can be understood that the daily scale is used to capture short-period load fluctuations and is suitable for sudden load and short-term new energy change prediction, and the feature vector includes the following components: load and output features: based on the extracted moving average (window size , corresponding to 24 hours) and difference features , reflecting the hourly load or new energy output change. Time features: including periodic encoding of time stamps (such as sine and cosine encoding of hours ) to capture the periodicity within the day (such as morning and evening peak). Weather features: weather data (such as temperature, light intensity, wind speed), extract weather indicators related to short-period load (such as daily average temperature, hourly light change rate). Equipment status features: including real-time load rates at station level, line level, and transformer level (such as main transformer load rate, feeder load rate, transformer load rate), reflecting the running state of equipment at short time scale.

[0036] The weekly scale is used to capture medium-period regularity and reflect weekday / weekend patterns, and the feature vector includes the following components: load and output features: based on the extracted moving average (window size , corresponding to 7 days), reflecting the average trend of load within the week; difference features (based on daily scale difference aggregation, such as daily difference mean), capturing the fluctuation regularity within the week. Time features: including weekly time encoding (such as one-hot encoding of the day of the week, or sine / cosine encoding Meteorological features: extract weekly-scale meteorological statistics (e.g., weekly average temperature, weekly maximum light intensity, weekly average wind speed) to reflect the influence of weather on the load pattern within a week. Operational pattern features: include the weekly average load rate at the station level, line level, and transformer level, as well as the load rate comparison index between weekdays and weekends (e.g., the difference between the average load rate on weekdays and the average load rate on weekends).

[0037] monthly scale to capture long-term trends and reflect seasonal changes, the feature vector includes the following components: load and output features: based on the extracted moving average (window size corresponding to 30 days), reflecting the long-term trend of monthly load or output; difference features (based on weekly-scale difference aggregation, such as weekly difference average), capturing the trend changes within the month. Time features: include intra-month time encoding (e.g., month one-hot encoding or sine / cosine encoding ) to capture seasonal regularity. Meteorological features: extract monthly-scale meteorological statistics (e.g., monthly average temperature, monthly average light intensity, monthly average wind speed) to reflect the influence of seasonal weather changes on the load. Long-term trend features: include the monthly average load rate at the station level, line level, and transformer level, as well as the trend slope of the monthly load rate (e.g., the load rate change rate calculated by linear regression).

[0038] Each scale sequence can extract local and global features through convolution or attention mechanism, which can be expressed as: ; where is the operational feature vector at scale , is the candidate prediction model corresponding to the time scale, is the sliding window size.

[0039] Preferably, after obtaining the operational feature vectors representing the load fluctuation trend and the power generation fluctuation trend of the power distribution network and the topology of the power distribution network at a preset number of time periods, the method further comprises: S11, obtaining a preset prediction time, a preset number of candidate load prediction models, and historical prediction error rates of each candidate load prediction model; Preferably, the method of obtaining a preset prediction time, a preset number of candidate load prediction models, and historical prediction error rates of each candidate load prediction model comprises: obtain historical prediction data of preset candidate load prediction models; wherein the historical prediction data comprises a historical operation feature vector, a historical load prediction result, and an error rate of the historical load prediction result; calculate a similarity of the operation feature vector and the historical operation feature vector, and take historical prediction data corresponding to several historical operation feature vectors with a similarity greater than a preset threshold as target historical prediction data of each candidate load prediction model; and calculate a historical prediction error rate of each candidate load prediction model according to error rates in the target historical prediction data.

[0040] In a preferred embodiment of the present application, several candidate load prediction models of different structures (LSTM, Transformer, and GRU) are pre-trained and stored in a model pool. : ; Further, by calculating the similarity of the operation feature vector and the historical operation feature vector, the target historical prediction data of each candidate load prediction model in a similar power distribution network operation scenario as the present is screened, and the historical prediction error rate of each candidate load prediction model in the similar power distribution network operation scenario as the present is determined.

[0041] The historical prediction error rate is obtained by assigning weights to error rates of each target historical prediction data according to the similarity, and then calculating a weighted sum of the error rates as the historical prediction error rate of each candidate load prediction model.

[0042] S12, according to the operation feature vector, the historical prediction error rate, and the prediction time, a first state sequence of a substation level, a second state sequence of a line level, and a third state sequence of a transformer level are respectively constructed; S13, with the goal of maximizing prediction accuracy, calculation efficiency, and performance stability, according to the first state sequence, the second state sequence, and the third state sequence, a first load prediction model for predicting a first prediction load, a second load prediction model for predicting a second prediction load, and a third load prediction model for predicting a third prediction load are respectively selected from the candidate load prediction models.

[0043] In a preferred embodiment of the present application, a model selection network is constructed, and a dynamic selection model learning “period + feature → optimal model” is trained by designing states (current period, level feature, historical error), actions (selecting a model in the model pool), and rewards (minimizing prediction error + calculation efficiency + result stability).

[0044] state sequence : ; wherein is a multi-scale feature vector, which is defined as is a historical prediction error rate, is a prediction time, such as a daily, weekly, or monthly load, represents a level, which is represented by a discrete value indicating the level of the current prediction task.

[0045] Further, the action selects a prediction model from the model pool; the reward is defined to maximize the comprehensive maximization of prediction accuracy, computational efficiency, and result stability, and the reward function is defined to minimize the prediction error, prediction variance, and computational overhead: ; wherein , is a weight coefficient, is a prediction error, is a prediction result variance, used to measure stability, is the computational overhead of the model , which is defined as: ; wherein , is a weight coefficient, is the inference time, calculated as , wherein is the actual inference time, is the preset maximum allowed inference time (e.g., 1 second). is the computational resource occupancy rate, calculated as , wherein is the actual resource occupancy, is the maximum available resource amount.

[0046] The agent strategy optimization formula is: ; wherein is the agent strategy, is a discount factor that controls the influence of future rewards, is the prediction time step length.

[0047] After training is completed, the model selection network will select the first load prediction model, the second load prediction model, and the third load prediction model when selecting a model, with the goal of maximizing prediction accuracy, computational efficiency, and performance stability.

[0048] ​S2, predicting first predicted load of each substation in the power distribution network according to the operation feature vector, and determining first overload risk of each substation according to the first predicted load; In a preferred embodiment of the present application, the operation feature vector is input into a first load prediction model to obtain a first predicted load , and the upper limit of capacity of each substation is determined according to the first predicted load . In contrast, the overload ratio is calculated as a station-level risk indicator . ; When , , the risk =0. When , , the risk increases linearly according to the overload ratio.

[0049] S3, determining substation connected to each line according to the topology, and predicting second predicted load of each line according to the first overload risk of the substation corresponding to each line and the operation feature vector, and determining second overload risk of each line according to the second predicted load. Preferably, the second predicted load of each line is predicted according to the first overload risk of the substation corresponding to each line and the operation feature vector, and the second overload risk of each line is determined according to the second predicted load, comprising: inputting the operation feature vector into a second load prediction model, so that the second load prediction model outputs the second predicted load of each line, and calculating the initial second overload risk of each line according to the second predicted load; and weighting and summing the initial second overload risk and the related first overload risk to determine the second overload risk of each line.

[0050] In a preferred embodiment of the present application, the second predicted load: ; wherein, the operation feature vector is , the second load prediction model is . Further, the initial second overload risk of the line is : ; wherein, the upper limit of capacity of the line is .

[0051] When the line substation connected, the second overload risk is: wherein, is a station→line level risk transfer weight coefficient.

[0052] S4, according to the topology, determining the line connected by each transformer, and according to the second overload risk of the line corresponding to each transformer and the operation feature vector, predicting the third predicted load of each transformer, and according to the third predicted load, determining the third overload risk of each transformer; Preferably, according to the second overload risk of the line corresponding to each transformer and the operation feature vector, predicting the third predicted load of each transformer, and according to the third predicted load, determining the third overload risk of each transformer, comprising: inputting the operation feature vector into a third load prediction model, so that the third load prediction model outputs the third predicted load of each line, and according to the third predicted load, calculating the initial third overload risk of each line segment; the initial third overload risk and the related second overload risk are weighted and summed to determine the third overload risk of each line.

[0053] In a preferred embodiment of the present application, the third predicted load and the initial third overload risk: wherein, is the capacity upper limit of the transformer , the third load prediction model , the third predicted load , the initial third overload risk.

[0054] When the transformer is connected with the line , the third overload risk : .

[0055] wherein is a line→transformer level risk transfer coefficient.

[0056] S5, according to the third overload risk, calculating the cascade overload probability for representing the fault caused by the overload event of the power distribution network; S6, according to the cascade overload probability, generating the overload risk warning signal of the power distribution network. ​​​​

[0057] In a preferred embodiment of the present application, the cascading overload probability is calculated according to the risk of each level after topological mapping : ; Finally, according to matching with the preset threshold of each level risk warning signal, the matching successful overload risk warning signal is sent to the operation and maintenance personnel.

[0058] Finally, in this embodiment, the prediction results and risk indicators of each level are sorted into a report, including: periodic risk: station, line, and transformer level risks are summarized according to different time scales such as day, week, and month; cascading warning: cascading overload probability is calculated to generate a warning level; decision reference: operation adjustment suggestions are provided, such as adjusting load distribution, hierarchical control, equipment maintenance plan, etc.; visual output: risk heat map, load curve, cascading warning chart, etc.

[0059] To verify the present application, a typical feeder of a city-level distribution network in southern China and the distribution transformers under its jurisdiction are selected as the verification object, and actual collected load data is used for example analysis. The method of the present application is based on Python 3.10, TensorFlow 2.9, and Gurobi 10.0.1 for algorithm implementation and verification. The load prediction results are as shown in Figure 4 , the prediction results of each level overload risk are as shown in Figure 5 , and the final prediction results are as shown in the following table: Referring to Figure 2 , an embodiment of the present application provides a layered overload risk warning device for a distribution network, comprising: a data acquisition module configured to acquire an operation feature vector representing load fluctuation trend and power generation fluctuation trend of the distribution network and a topological structure of the distribution network in a preset time period; a first prediction module configured to predict a first predicted load of each transformer substation in the distribution network according to the operation feature vector, and determine a first overload risk of each transformer substation according to the first predicted load; a second prediction module configured to determine transformer substations connected by each line according to the topological structure, predict a second predicted load of each line according to the first overload risk of the transformer substations corresponding to each line and the operation feature vector, and determine a second overload risk of each line according to the second predicted load; a third prediction module, configured to determine lines connected to each transformer according to the topology structure, and predict a third predicted load of each transformer according to the second overload risk of the line corresponding to each transformer and the operation feature vector, and determine a third overload risk of each transformer according to the third predicted load; a risk assessment module, configured to calculate a cascading overload probability for representing the existence of an overload event of the power distribution network according to the third overload risk; a risk warning module, configured to generate an overload risk warning signal of the power distribution network according to the cascading overload probability.

[0060] Further, the data acquisition module acquires an operation feature vector for representing load fluctuation trend and power generation fluctuation trend of the power distribution network in a preset time period, including: acquiring device parameters of the power distribution network, historical operation data and historical environment data at a plurality of historical time nodes; aligning the historical operation data, the historical environment data and the device parameters in time and space, and constructing a time-space correlation matrix; extracting the operation feature vector for representing the load fluctuation trend and the power generation fluctuation trend of the power distribution network in each time period according to the preset time period and the time-space correlation matrix.

[0061] It can be understood that the above device embodiment is corresponding to the method embodiment of the application, which can realize the power distribution network hierarchical overload risk warning method provided by any one of the above method embodiments.

[0062] It should be noted that the device embodiments described above are only schematic, and part or all of the modules can be selected to achieve the purpose of the embodiment. In addition, the connection relationship between the modules in the device embodiment provided by the application indicates that there is a communication connection between them, which can be realized as one or more communication buses or signal lines. Those skilled in the art can understand and implement without creative labor.

[0063] Referring to Figure 3 The application also provides a terminal device, including: one or more processors; a memory coupled to the processor, configured to store one or more programs; When the one or more programs are executed by the one or more processors, the one or more processors implement the power distribution network hierarchical overload risk warning method as described above.

[0064] The processor is configured to control overall operations of the terminal device to complete all or part of the steps of the power grid hierarchical overload risk early warning method described above. The memory is configured to store various types of data to support the operation of the terminal device, which can include, for example, instructions for any application or method operating on the terminal device, and application-related data. The memory can be implemented by any type of volatile or non-volatile storage device or a combination thereof, such as static random access memory (SRAM), electrically erasable programmable read-only memory (EEPROM), erasable programmable read-only memory (EPROM), programmable read-only memory (PROM), read-only memory (ROM), magnetic memory, flash memory, magnetic disk or optical disk.

[0065] In an exemplary embodiment, the terminal device can be implemented by one or more application specific integrated circuits (ASICs), digital signal processors (DSPs), digital signal processing devices (DSPDs), programmable logic devices (PLDs), field programmable gate arrays (FPGAs), controllers, microcontrollers, microprocessors or other electronic elements, for executing a power grid hierarchical overload risk early warning method as described in any of the above embodiments and achieving the technical effects consistent with the above method.

[0066] In another exemplary embodiment, a computer-readable storage medium including a computer program is also provided, which, when executed by a processor, implements the steps of a power grid hierarchical overload risk early warning method as described in any of the above embodiments. For example, the computer-readable storage medium can be the above-mentioned memory including the computer program, which can be executed by the processor of the terminal device to complete a power grid hierarchical overload risk early warning method as described in any of the above embodiments and achieve the technical effects consistent with the above method.

[0067] The above is the preferred embodiment of the present application, it should be noted that for those skilled in the art, without departing from the principles of the present application, can also make several improvements and refinements, these improvements and refinements are also considered the scope of protection of the present application.

Claims

1. A hierarchical heavy overload risk early warning method for power distribution networks, characterized in that, include: Obtain the operating feature vectors and the topology of the distribution network to characterize the load fluctuation trend and the power generation fluctuation trend of the distribution network under a preset number of time periods. Based on the operating feature vector, the first predicted load of each substation in the distribution network is predicted, and based on the first predicted load, the first major overload risk of each substation is determined. Based on the topology, the substations connected to each line are determined, and based on the first overload risk of the substations corresponding to each line and the operating feature vector, the second predicted load of each line is predicted, and based on the second predicted load, the second overload risk of each line is determined. Based on the topology, the lines connected to each transformer are determined, and based on the second overload risk of the lines corresponding to each transformer and the operating feature vector, the third predicted load of each transformer is predicted, and based on the third predicted load, the third overload risk of each transformer is determined. Based on the third overload risk, calculate the cascade overload probability used to characterize the fault caused by a heavy overload event in the distribution network; Based on the cascaded overload probability, a heavy overload risk warning signal for the distribution network is generated.

2. The hierarchical heavy overload risk early warning method for a distribution network as described in claim 1, characterized in that, The acquisition of operational feature vectors, used to characterize the load fluctuation trend and power generation fluctuation trend of the distribution network under a preset number of time periods, includes: Acquire equipment parameters, historical operating data at several historical time points, and historical environmental data of the power distribution network; Based on historical operational data, historical environmental data, and equipment parameters, time and space alignment are performed to construct a spatiotemporal correlation matrix; Based on the preset time period and spatiotemporal correlation matrix, operational feature vectors are extracted for each time period to characterize the load fluctuation trend and power generation fluctuation trend of the power distribution network.

3. The hierarchical heavy overload risk early warning method for a distribution network as described in claim 2, characterized in that, After obtaining the operating feature vectors and the topology of the distribution network, which characterize the load fluctuation trend and generation fluctuation trend of the distribution network under several preset time periods, the process also includes: Obtain the preset prediction time, several preset candidate load prediction models, and the historical prediction error rate of each candidate load prediction model; Based on the operational feature vector, the historical prediction error rate, and the prediction time, a first state sequence at the substation level, a second state sequence at the line level, and a third state sequence at the transformer level are constructed respectively. With the goal of maximizing prediction accuracy, computational efficiency, and performance stability, a first load prediction model for predicting the first predicted load, a second load prediction model for predicting the second predicted load, and a third load prediction model for predicting the third predicted load are selected from the candidate load prediction models based on the first state sequence, the second state sequence, and the third state sequence.

4. The hierarchical heavy overload risk early warning method for a distribution network as described in claim 3, characterized in that, The acquisition of the preset prediction time, a number of preset candidate load prediction models, and the historical prediction error rate of each candidate load prediction model includes: Obtain historical prediction data of a preset candidate load prediction model; wherein, the historical prediction data includes: historical operating feature vectors, historical load prediction results, and error rates of historical load prediction results; Calculate the similarity between the running feature vector and the historical running feature vector, and use the historical prediction data corresponding to several historical running feature vectors with similarity greater than a preset threshold as the target historical prediction data for each candidate load prediction model. Calculate the historical prediction error rate of each candidate load prediction model based on the error rate in the historical prediction data of several targets.

5. A hierarchical heavy overload risk early warning method for a distribution network as described in claim 4, characterized in that, The step of predicting the second predicted load of each line based on the first overload risk of the substation corresponding to each line and the operating characteristic vector, and determining the second overload risk of each line based on the second predicted load, includes: The running feature vector is input into the second load prediction model so that the second load prediction model outputs the second predicted load of each line, and the initial second heavy overload risk of each line segment is calculated based on the second predicted load. The initial second-level overload risk and the related first-level overload risk are weighted and summed to determine the second-level overload risk of each line.

6. The hierarchical heavy overload risk early warning method for a distribution network as described in claim 5, characterized in that, The step of predicting the third predicted load of each transformer based on the second overload risk of the line corresponding to each transformer and the operating characteristic vector, and determining the third overload risk of each transformer based on the third predicted load, includes: The running feature vector is input into the third load prediction model so that the third load prediction model outputs the third predicted load of each line, and the initial third overload risk of each line segment is calculated based on the third predicted load. The initial third overload risk and the related second overload risk are weighted and summed to determine the third overload risk of each line.

7. A hierarchical heavy overload risk early warning device for a power distribution network, characterized in that, include: The data acquisition module is used to acquire the operating feature vectors and the topology of the distribution network, which characterize the load fluctuation trend and the power generation fluctuation trend of the distribution network under a preset number of time periods. The first prediction module is used to predict the first predicted load of each substation in the distribution network based on the operating feature vector, and to determine the first heavy overload risk of each substation based on the first predicted load. The second prediction module is used to determine the substations connected to each line according to the topology, and to predict the second predicted load of each line according to the first overload risk of the substations corresponding to each line and the operating feature vector, and to determine the second overload risk of each line according to the second predicted load. The third prediction module is used to determine the lines connected to each transformer based on the topology, and to predict the third predicted load of each transformer based on the second overload risk of the lines corresponding to each transformer and the operating feature vector, and to determine the third overload risk of each transformer based on the third predicted load. The risk assessment module is used to calculate the cascaded overload probability, which characterizes the existence of a major overload event in the distribution network, based on the third overload risk. The risk warning module is used to generate a heavy overload risk warning signal for the distribution network based on the cascaded overload probability.

8. A hierarchical heavy overload risk early warning device for a power distribution network as described in claim 7, characterized in that, The data acquisition module acquires operational feature vectors, used to characterize the load fluctuation trend and power generation fluctuation trend of the distribution network under several preset time periods, including: Acquire equipment parameters, historical operating data at several historical time points, and historical environmental data of the power distribution network; Based on historical operational data, historical environmental data, and equipment parameters, time and space alignment are performed to construct a spatiotemporal correlation matrix; Based on the preset time period and spatiotemporal correlation matrix, operational feature vectors are extracted for each time period to characterize the load fluctuation trend and power generation fluctuation trend of the power distribution network.

9. A terminal device, characterized in that, include: One or more processors; A memory, coupled to the processor, for storing one or more programs; When the one or more programs are executed by the one or more processors, the one or more processors implement a hierarchical heavy overload risk early warning method for a power distribution network as described in any one of claims 1-6.

10. A storage medium having a computer program stored thereon, characterized in that, When the computer program is executed by the processor, it implements a hierarchical heavy overload risk early warning method for a power distribution network as described in any one of claims 1-6.

Images