Load forecasting method for urban low-voltage distribution network based on PHM

Abstract

This invention relates to the field of distribution network load forecasting technology. Specifically, it relates to a load forecasting method for urban medium- and low-voltage distribution networks based on PHM (Prognostics and Health Management). It includes the following steps: S1, collecting historical load data of the medium- and low-voltage distribution network, simultaneously acquiring PHM data of the medium- and low-voltage distribution network equipment, performing risk health index analysis based on the PHM data, and obtaining the risk health index of the medium- and low-voltage distribution network. This invention employs a hierarchical load splitting strategy of medium-voltage feeder-transformer area-individual user, combined with a two-dimensional user classification system of electricity consumption characteristics and total electricity load, refining load data to the individual user level. Simultaneously, it dynamically sets the data collection duration through load fluctuation frequency and sets a comprehensive fluctuation range based on the statistical characteristics of the load fluctuation range of users of the same category, achieving refined and personalized load forecasting. Compared with existing single-dimensional classification and fixed-range constraints, this significantly improves the accuracy of load forecasting for different types of users.

Description

Technical Field

[0001] This invention relates to the field of distribution network load forecasting technology, and more specifically, to a load forecasting method for urban medium and low voltage distribution networks based on PHM (Programmable Schematic). Background Technology

[0002] As the core hub connecting power production and end users, the operation status of urban medium and low voltage distribution networks directly affects power supply reliability, power quality and grid operation efficiency. Load forecasting is a key technical support for distribution network planning, design, dispatching, operation and fault prevention.

[0003] Existing forecasting methods mostly focus on the correlation analysis between historical load data and external environmental factors, ignoring the constraints of the health status of medium and low voltage distribution network equipment on load carrying capacity. The aging and deterioration of distribution network equipment (such as distribution transformers, ring main units, charging piles, etc.) will directly reduce its safe load limit. The forecast results based solely on electricity consumption patterns may exceed the actual carrying capacity of the equipment, causing risks such as equipment overload and fault tripping, resulting in the forecast results being out of touch with the actual project.

[0004] Secondly, existing technologies lack sufficient granularity in load data breakdown, mostly remaining at the level of medium-voltage feeders or low-voltage distribution areas, without refining to individual power users. Furthermore, the user classification methods are simplistic (mostly based solely on electricity consumption characteristics), making it difficult to accurately capture the load fluctuation characteristics of different users, thus limiting prediction accuracy. Therefore, a load prediction method for urban medium and low voltage distribution networks based on PHM is proposed. Summary of the Invention

[0005] The purpose of this invention is to provide a load forecasting method for urban low-voltage distribution networks based on PHM, so as to solve the problems mentioned in the background art.

[0006] To achieve the above objectives, a load forecasting method for urban medium and low voltage distribution networks based on PHM is provided, including the following steps:

[0007] S1. Collect historical load data of medium and low voltage distribution networks, and at the same time obtain PHM data of medium and low voltage distribution network equipment. Perform risk health index analysis based on PHM data to obtain the risk health index of medium and low voltage distribution networks.

[0008] S2. Extract power users of the medium and low voltage distribution network based on historical load data, and match the historical load data with the power users to obtain the historical load data of each power user. Then sort the historical load data by recording time.

[0009] S3. Based on the load variation rate and recording time of historical load data, select valid time periods for power users, and analyze the load fluctuation range of power users in each time period based on the historical load data sorted by the valid time periods and the recording time to obtain the corresponding load fluctuation range of power users.

[0010] S4. Extract the historical load data of the latest recorded time to classify and match power users. Set the comprehensive fluctuation range corresponding to different categories according to the classification and matching results. Then, predict the predicted load data of power users based on the historical load data. Based on the accuracy of the historical predicted load data of power users, dynamically adjust the predicted load data of power users in the same category in the load fluctuation range and comprehensive fluctuation range.

[0011] S5. Summarize the adjusted forecast load data of all electricity users and convert it into a comprehensive health index. Compare the comprehensive health index with the risk health index. If the comprehensive health index is higher than the risk health index, an early warning will be triggered.

[0012] As a further improvement to this technical solution, in step S1, the historical load data of the medium and low voltage distribution network is collected from the data management section in the same power supply area by the distribution network dispatch center connected to the distribution network, and the PHM data of the medium and low voltage distribution network equipment is obtained at the same time.

[0013] As a further improvement to this technical solution, the risk health index analysis is performed based on PHM data to obtain the risk health index of the medium and low voltage distribution network;

[0014] The process involves first normalizing various PHM data, then using a weighted summation method to calculate the health sub-index of a single device, and finally determining the weight of each device's health sub-index based on the entropy weight method, and finally weighting them to obtain the risk health index of the medium and low voltage distribution network.

[0015] As a further improvement to this technical solution, in S2, the power users of the medium and low voltage distribution network are extracted based on historical load data. According to the distribution network area affiliation, the load data of the medium voltage feeder is broken down to each subordinate distribution transformer area. Then, the total load data of the distribution transformer area is broken down to individual power users according to the user electricity metering ledger, thereby establishing a correlation and matching between historical load data and power users.

[0016] Based on timestamps, the historical load data of each power user are sorted in ascending order from morning to night to form a time-series load dataset.

[0017] As a further improvement to this technical solution, in step S3, load data at adjacent recorded time points in the time-series load dataset are selected to calculate the load variation rate, thereby obtaining the load variation rate for adjacent time periods.

[0018] Set a load fluctuation rate threshold, then compare and filter the load fluctuation rates of adjacent time periods with the load fluctuation rate threshold, and select the time period with the most consecutive load fluctuation rates within the load fluctuation rate threshold range, and take this time period as the valid time period for the power user.

[0019] For each selected valid time period, extract all historical load data of each power user within that time period, statistically obtain the maximum and minimum load values ​​within that time period, and calculate the difference between the maximum and minimum values, which is the load fluctuation range of the corresponding valid time period for that power user.

[0020] As a further improvement to this technical solution, in step S4, multiple power user categories are established, historical load data of the latest recorded time is extracted to classify and match power users, and the electricity consumption nature and load characteristics of power users are extracted from the historical load data. Based on the electricity consumption nature and load characteristics, power users are matched with the corresponding power user categories.

[0021] The latest record time can be set according to the fluctuation frequency of the power user;

[0022] Electricity user categories correspond to various categories with different total electricity loads;

[0023] Based on the category of electricity users, the load fluctuation range of electricity users in the same category is statistically analyzed, and the mean and standard deviation of the load fluctuation range of all users in that category are calculated. Then, the mean ± 2 times the standard deviation is taken as the comprehensive fluctuation range of users in that category.

[0024] As a further improvement to this technical solution, in step S4, a time-series forecasting method is used to predict the load data for each period in the next 24 hours based on the historical load data of each power user after sorting, so as to obtain the initial predicted load data.

[0025] Retrieve historical predicted load data and historical load data of each power user for the same period in recent times to analyze the prediction accuracy and obtain the prediction accuracy of each power user.

[0026] Specifically, if the prediction accuracy is higher than 90%, the overall fluctuation range of the user's category will be widened by 10%.

[0027] When the prediction accuracy is below 80%, the overall fluctuation range will be narrowed by 10%.

[0028] Then, the initial forecast load data is compared with the load fluctuation range of the corresponding effective period and the adjusted comprehensive fluctuation range. If the initial forecast value exceeds any range, it is corrected to the range to obtain the adjusted forecast load data.

[0029] Among these, the predicted load data is limited to not exceeding the comprehensive fluctuation range and the load fluctuation range.

[0030] As a further improvement to this technical solution, in step S5, the adjusted predicted load data of all power users are summarized, and the total predicted load data of each distribution area and the entire distribution network are obtained according to the topological affiliation of medium-voltage feeders and low-voltage distribution areas.

[0031] In the distribution network dispatch center, query the maximum allowable load value of each equipment in the medium and low voltage distribution network, calculate the ratio of the total predicted load data to the maximum allowable load value, and map this ratio to the health index range of [0,1] according to the preset mapping rules. The smaller the ratio, the closer the comprehensive health index is to 1, and vice versa. Finally, the comprehensive health index of the entire distribution network is obtained.

[0032] As a further improvement to this technical solution, the comprehensive health index of the entire power distribution network is compared with the risk health index set in S1.

[0033] When the comprehensive health index exceeds the risk health index, an early warning signal is immediately sent to the distribution network dispatch center.

[0034] If the overall health index does not exceed the risk health index, monitoring will continue.

[0035] Compared with the prior art, the beneficial effects of the present invention are as follows:

[0036] 1. This load forecasting method for urban medium and low voltage distribution networks based on PHM adopts a hierarchical load decomposition strategy of medium-voltage feeder-transformer area-individual user, combined with a two-dimensional user classification system of electricity consumption nature and total electricity load, to refine load data to the individual user level. At the same time, the data collection duration is dynamically set by load fluctuation frequency, and the comprehensive fluctuation range is set based on the statistical characteristics of the load fluctuation range of the same type of user. This achieves refined and personalized load forecasting. Compared with the existing single-dimensional classification and fixed range constraints, it significantly improves the accuracy of load forecasting for different types of users, and is especially suitable for complex distribution network environments where multiple user types coexist.

[0037] 2. In this PHM-based urban low-voltage distribution network load forecasting method, PHM data is incorporated into the load forecasting system. Through normalization processing, weighted summation, and entropy weighting, the health status of equipment is objectively quantified, and a risk health index is constructed. This achieves deep coupling between equipment health carrying capacity and load forecasting, effectively solving the problem that the predicted value exceeds the actual carrying capacity due to the neglect of equipment health constraints in existing technologies. It reduces the risk of equipment overload and fault tripping from the source and ensures the safe operation of the distribution network.

[0038] 3. In this PHM-based urban medium and low voltage distribution network load forecasting method, a dynamic optimization closed loop of forecast accuracy, fluctuation range, and forecast value correction is constructed. The comprehensive fluctuation range of categories is adjusted in real time by adjusting the historical forecast accuracy, and then the individual load fluctuation range of users is combined for dual constraint correction. This ensures that the forecast results not only conform to the historical stable load characteristics of individual users, but also conform to the group fluctuation pattern of users of the same category. It effectively adapts to the dynamic changes in user electricity consumption characteristics and further improves the reliability and adaptability of the forecast results. Attached Figure Description

[0039] Figure 1 This is a flowchart illustrating the load forecasting method for urban medium and low voltage distribution networks based on PHM according to the present invention. Detailed Implementation

[0040] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0041] Please see Figure 1 As shown, the purpose of this embodiment is to provide a load forecasting method for urban medium and low voltage distribution networks based on PHM, including the following steps:

[0042] S1. Collect historical load data of medium and low voltage distribution networks, and simultaneously obtain PHM data of medium and low voltage distribution network equipment. Perform risk health index analysis based on PHM data to obtain the risk health index of medium and low voltage distribution networks. Complete the dual core data collection of load data and equipment health data. At the same time, quantify PHM data into comparable risk health indexes to provide a benchmark threshold for risk assessment of subsequent prediction results. This is the data input and risk calibration layer of the entire method.

[0043] In S1, through the distribution network dispatch center connected to the distribution network, historical load data of medium and low voltage distribution networks are collected from the data management section in the same power supply area, and PHM data of medium and low voltage distribution network equipment are obtained at the same time.

[0044] The core server of the power distribution network dispatch center is connected through a communication interface. The data collection scope is defined as the same power supply area (such as a power supply zone in a certain urban area) to ensure the regional homogeneity of the collected data. Then, the historical load data of medium and low voltage power distribution network for the past 12 months is retrieved from the data management section of the power distribution network dispatch center. Specifically, this includes the active power of medium voltage feeders, the total active load of distribution transformer areas, and the electricity load of individual users. The collection granularity is set to 15 minutes / data entry. The timestamp, meteorological information (temperature, humidity) and holiday markers of each data entry are synchronously associated.

[0045] The equipment status monitoring module of the distribution network dispatch center synchronously acquires PHM data of medium and low voltage distribution network equipment in the same power supply area. The core indicators include top oil temperature, partial discharge, and insulation resistance of medium voltage side distribution transformers, contact temperature and opening and closing mechanical characteristics of ring main units, switch temperature rise and leakage current of low voltage side distribution cabinets, module temperature of charging piles, and inverter operating parameters. The acquisition frequency is matched with the load data at 15 minutes / time.

[0046] Risk health index analysis is performed based on PHM data to obtain the risk health index of medium and low voltage distribution networks;

[0047] The process involves first normalizing various PHM data, then calculating the health sub-index of a single device using a weighted summation method, and finally determining the weights of each device's health sub-index based on the entropy weight method. The weighted average is then used to obtain the risk health index for medium and low voltage distribution networks. The steps are as follows:

[0048] PHM data of different dimensions are mapped to the [0,1] interval to eliminate the difference in dimensions and ensure the rationality of subsequent calculations. Then, for each medium and low voltage distribution network equipment, its core PHM indicators are selected (such as oil temperature, partial discharge, and insulation resistance for distribution transformers). The weighted summation method is used to calculate the health sub-index of a single equipment. The weights are preset to initial values ​​based on the degree of influence of the indicators on equipment failure.

[0049] Based on the health sub-indices of all equipment within the same power supply area, the objective weight of each equipment health sub-index is calculated using the entropy weight method to avoid the bias of subjective weights. The weight allocation is positively correlated with the degree of variation of the equipment health sub-index (the higher the degree of variation, the greater the weight). At the same time, each equipment health sub-index is multiplied by its corresponding entropy weight, and all product results are accumulated to obtain the risk health index of the low-voltage distribution network in the power supply area (value range [0,1], the closer to 1 indicates the lower the risk, and the closer to 0 indicates the higher the risk).

[0050] S2. Based on historical load data, power users in the medium and low voltage distribution network are extracted. At the same time, the historical load data is matched with the power users to obtain the historical load data of each power user. Then, the historical load data is sorted by recording time. The macro-level medium and low voltage distribution network load data is broken down and refined to individual power users. At the same time, a time series dataset is constructed by sorting by time to realize the three-dimensional correlation between load data, users and time, providing refined data support for subsequent load fluctuation analysis and accurate prediction.

[0051] In S2, power users of medium and low voltage distribution networks are extracted based on historical load data. According to the distribution network area affiliation, the load data of medium voltage feeders is broken down to each subordinate distribution transformer area. Then, the total load data of the distribution transformer area is broken down to individual power users according to the user electricity metering ledger, thereby establishing a correlation and matching between historical load data and power users.

[0052] Retrieve the medium and low voltage distribution network topology ledger from the distribution network dispatch center, clarify the list of subordinate distribution transformer areas corresponding to each medium voltage feeder in the same power supply area, and establish a one-to-one correspondence table between medium voltage feeders and distribution transformer areas.

[0053] For the historical total load data of a single medium-voltage feeder, the historical average load ratio of each distribution transformer area under the feeder is calculated. The total load of the feeder is then distributed to each corresponding distribution transformer area according to the ratio, ensuring that the sum of the loads of all transformer areas after distribution is equal to the total load of the feeder. Then, the electricity metering ledgers of each distribution transformer area are retrieved, and the historical electricity consumption ratio of each power user in the area is calculated. The total load data of the distribution transformer area is then split to the corresponding individual power users according to the ratio, ensuring that the sum of the loads of all users after splitting is equal to the total load of the transformer area.

[0054] Using the unique electricity user ID as an index, the split single-user load data is associated with the user's basic information (electricity usage type, metering point location), and invalid data that fails to match is removed to form a user ID-load data association dataset.

[0055] Based on timestamps, the historical load data of each power user are sorted in ascending order from morning to night to form a time-series load dataset.

[0056] Extract the timestamp information of each load data in the associated dataset, convert the timestamp into continuous time series values, and sort the historical load data of each power user in ascending order of time series values ​​from smallest to largest to finally form a continuous time series load dataset for a single user.

[0057] S3. Based on the load variation rate and recording time of historical load data, select valid time periods for power users, and analyze the load fluctuation range of power users in each time period according to the historical load data sorted by the effective time period and the recording time to obtain the load fluctuation range corresponding to the power users; select valid time periods with stable load status, eliminate the interference of abnormal fluctuation periods on load pattern analysis, and quantify the load fluctuation range of each user in the valid time period to provide user-level fluctuation constraint boundaries for the prediction value correction in S4.

[0058] In S3, load data at adjacent recorded time points in the time-series load dataset are selected to calculate the load variation rate, thus obtaining the load variation rate for adjacent time periods.

[0059] For a single user's time-series load dataset, load data from two adjacent time points are selected sequentially in ascending order of time series values ​​and recorded as the load value at the previous time point and the load value at the next time point, forming multiple pairs of adjacent load data.

[0060] Based on each pair of adjacent load data, the load variation rate for the corresponding adjacent time period is calculated. After the variation rate calculation for all adjacent time periods is completed, the load variation rate sequence for that user is formed.

[0061] Set a load fluctuation rate threshold, then compare and filter the load fluctuation rates of adjacent time periods with the load fluctuation rate threshold, and select the time period with the most consecutive load fluctuation rates within the load fluctuation rate threshold range, and take this time period as the valid time period for the power user.

[0062] Based on the load characteristics of different types of users in the distribution network (residential users, industrial and commercial users, and public facility users), differentiated load variation rate threshold ranges are set to ensure that the thresholds are adapted to the users' electricity consumption patterns. Then, the load variation rate sequence of each user is traversed, and the variation rate of each time period is compared with the preset threshold range. Time periods that meet the conditions are marked, and the number of consecutive time periods that meet the conditions is counted. The time period with the most consecutive time periods that meet the threshold range is selected and determined as the valid time period for the power user.

[0063] If there are multiple consecutive time periods with the same number of loads, the time period with the average load closest to the user's average daily load will be selected as the valid time period.

[0064] For each selected valid time period, extract all historical load data of each power user within that time period, statistically obtain the maximum and minimum load values ​​within that time period, and calculate the difference between the maximum and minimum values, which is the load fluctuation range of the corresponding valid time period for that power user.

[0065] S4. Extract historical load data from the latest recorded time to classify and match power users. Based on the classification and matching results, set the comprehensive fluctuation range corresponding to different categories. Then, predict the predicted load data of power users based on historical load data. Based on the accuracy of the historical predicted load data of power users, dynamically adjust the predicted load data of power users in the same category within the load fluctuation range and comprehensive fluctuation range. Combine user load characteristics to achieve classification prediction. Through the dual constraints of individual fluctuation range and group comprehensive fluctuation range, and based on the historical prediction accuracy, the fluctuation range is dynamically optimized. Finally, accurate and reasonable user-level predicted load data is output. This is the core prediction and optimization layer of the whole method.

[0066] In S4, multiple power user categories are established, historical load data of the latest recorded time is extracted to classify and match power users, and the electricity consumption nature and load characteristics of power users are extracted from the historical load data. Based on the electricity consumption nature and load characteristics, power users are matched with the corresponding power user categories.

[0067] The latest record time can be set according to the fluctuation frequency of the power user;

[0068] Electricity user categories correspond to various categories with different total electricity loads;

[0069] Based on the electricity consumption characteristics of users in medium and low voltage distribution networks, a two-level classification system is constructed. The first-level category is divided into three categories according to the nature of electricity consumption: residential users, industrial and commercial users, and public facility users. The second-level category is divided into three subcategories according to the total electricity load size: large, medium, and small, based on the first-level category, thus forming a multi-dimensional classification standard based on the nature of electricity consumption and total load.

[0070] Extract the historical load fluctuation rate sequence of each power user, calculate the user load fluctuation frequency (the number of times the load fluctuation rate exceeds the threshold per unit time), and set the latest record time duration according to the fluctuation frequency differences;

[0071] Among them, users with higher fluctuation frequency (such as industrial and commercial users) have shorter latest record times; users with lower fluctuation frequency (such as residential users) have longer latest record times, ensuring that the extracted data is timely and representative.

[0072] Based on the set latest record time duration, load data for the corresponding time period is extracted from the user time-series load dataset. Two types of features (electricity consumption nature and load characteristics) are extracted simultaneously. The extracted user electricity consumption nature features are matched with the first-level category, and the total electricity load data in the load characteristics are matched with the second-level sub-category to determine the specific category to which each power user belongs, forming a user-category correspondence table.

[0073] Based on the category of electricity users, the load fluctuation range of electricity users in the same category is statistically analyzed, and the mean and standard deviation of the load fluctuation range of all users in that category are calculated. Then, the mean ± 2 times the standard deviation is taken as the comprehensive fluctuation range of users in that category.

[0074] Retrieve valid time-period load fluctuation range data for all electricity users within the same category to form a load fluctuation range dataset for that category. Remove outliers from the dataset (using the Grubbs criterion) to ensure the validity of the statistical sample. Then, based on the cleaned load fluctuation range dataset for that category, calculate the mean and standard deviation of the dataset. Use the statistical method of mean ± 2 times standard deviation to determine the comprehensive fluctuation range for that category of users, as shown in the following formula:

[0075] ;

[0076] Where μ is the average load fluctuation range of users of the same category, n is the number of users of the same category, and ΔP i The effective time period load fluctuation range for the i-th user in the same category;

[0077] ;

[0078] Where σ is the standard deviation of the load fluctuation range of users of the same category;

[0079] ;

[0080] Wherein, ΔP group The comprehensive fluctuation range (kW) for this category of users is an interval value, μ−2σ is the lower limit of the comprehensive fluctuation range, and μ and 2σ are the upper limit of the comprehensive fluctuation range;

[0081] In S4, a time-series forecasting method is used to predict the load data for each period in the next 24 hours based on the historical load data of each power user after sorting. This yields the initial forecast load data.

[0082] The LSTM-Transformer hybrid model was selected as the best time series prediction model. This model takes into account both the long-term time series dependency mining and short-term fluctuation capture capabilities of load data. The historical load data of each power user after sorting was used as the training set. The training set, validation set and test set were divided in a 7:1:2 ratio. The input dimension of the model was set to the historical 24-hour load data, and the output dimension was the load data of each period in the next 24 hours. The model training was completed by minimizing the mean squared error loss function through the Adam optimizer.

[0083] The trained LSTM-Transformer model is deployed to the prediction module of the power distribution network dispatch center. The latest historical time-series load data of each power user is input, and the model predicts the load data for each 15-minute interval within the next 24 hours, obtaining the initial predicted load data for each user. The formula is as follows:

[0084] ;

[0085] Among them, P pred,i Let f be the initial predicted load value for the i-th user during a certain period of the next 24 hours. LSTM−T (·) represents the trained LSTM-Transformer hybrid prediction model, X. i，t-24：t For the i-th user, the historical load time series data for the 24 hours prior to time t;

[0086] Retrieve historical predicted load data and historical load data of each power user for the same period in recent times to analyze the prediction accuracy and obtain the prediction accuracy of each power user.

[0087] Retrieve historical predicted load data sets and historical actual load data sets for each electricity user within the same time period over the past 30 days, and calculate the prediction accuracy for each user using the Mean Absolute Percentage Error (MAPE) formula as follows:

[0088] ;

[0089] Where Acc is the prediction accuracy for the i-th user, m is the number of historical data samples used in the calculation, and P pred,hist,j Let P be the historical predicted load value for the j-th sample. true,hist,j This represents the historical actual load value for the j-th sample.

[0090] Specifically, if the prediction accuracy is higher than 90%, the overall fluctuation range of the user's category will be widened by 10%.

[0091] When the prediction accuracy is below 80%, the overall fluctuation range will be narrowed by 10%.

[0092] Then, the initial forecast load data is compared with the load fluctuation range of the corresponding effective period and the adjusted comprehensive fluctuation range. If the initial forecast value exceeds any range, it is corrected to the range to obtain the adjusted forecast load data.

[0093] Among these, the predicted load data is limited to not exceeding the comprehensive fluctuation range and the load fluctuation range.

[0094] S5. Summarize the adjusted forecast load data of all power users and convert it into a comprehensive health index. Compare the comprehensive health index with the risk health index. If the comprehensive health index is higher than the risk health index, an early warning is triggered. The user-level forecast load is aggregated into the distribution network-level total load and converted into a comprehensive health index that can be compared with the risk health index in S1. Risk assessment and early warning triggering are achieved through the comparison of the two, which is the output of the entire method.

[0095] In S5, the adjusted predicted load data of all power users are summarized, and the total predicted load data of each distribution area and the entire distribution network are obtained according to the topological affiliation of medium-voltage feeders and low-voltage distribution areas.

[0096] In the distribution network dispatch center, query the maximum allowable load value of each equipment area in the medium and low voltage distribution network, calculate the ratio of the total predicted load data to the maximum allowable load value, and map this ratio to the health index range of [0,1] according to the preset mapping rules. The smaller the ratio, the closer the comprehensive health index is to 1, and vice versa. Finally, the comprehensive health index of the entire distribution network is obtained. The steps are as follows:

[0097] The adjusted predicted load data of all power users is retrieved. Based on the distribution network topology, the data is first aggregated at the low-voltage distribution area level, summing the predicted load values ​​of all users within the same distribution area to obtain the total predicted load data for each low-voltage distribution area. Then, the data is aggregated at the medium-voltage feeder level, summing the total predicted load values ​​of all low-voltage distribution areas under the same medium-voltage feeder to obtain the total predicted load data for each medium-voltage feeder. Finally, the total predicted load data of all medium-voltage feeders is summed to obtain the total predicted load data for the entire distribution network.

[0098] In the equipment ledger management module of the distribution network dispatch center, the maximum allowable load value of each low-voltage distribution area, medium-voltage feeder and corresponding equipment of the entire distribution network is retrieved. This value is determined in combination with the equipment nameplate parameters, distribution network operation planning standards and the current health status of the equipment to ensure that the value conforms to the actual safe carrying capacity.

[0099] For the total predicted load data of each level (transformer area, feeder, and whole distribution network), calculate the ratio of it to the maximum allowable load value of the corresponding level. Using a linear mapping rule, map the ratio to the health index range of [0,1]. The smaller the ratio, the closer the comprehensive health index is to 1, which means that the distribution network is in a healthier operating state.

[0100] The larger the ratio, the closer the comprehensive health index is to 0, which means that the risk of overload in the distribution network is higher. The final output is the comprehensive health index of the entire distribution network.

[0101] The overall health index of the entire power distribution network is compared with the risk health index set in S1.

[0102] When the comprehensive health index exceeds the risk health index, an early warning signal is immediately sent to the distribution network dispatch center.

[0103] If the overall health index does not exceed the risk health index, monitoring will continue.

[0104] The foregoing has shown and described the basic principles, main features, and advantages of the present invention. Those skilled in the art should understand that the present invention is not limited to the above embodiments. The embodiments and descriptions in the specification are merely preferred examples and are not intended to limit the invention. Various changes and modifications can be made to the invention without departing from its spirit and scope, and all such changes and modifications fall within the scope of the present invention as claimed. The scope of protection of the present invention is defined by the appended claims and their equivalents.

Claims

1. A load forecasting method for urban medium and low voltage distribution networks based on PHM, characterized in that: Includes the following steps: S1. Collect historical load data of medium and low voltage distribution networks, and at the same time obtain PHM data of medium and low voltage distribution network equipment. Perform risk health index analysis based on PHM data to obtain the risk health index of medium and low voltage distribution networks. S2. Extract power users of the medium and low voltage distribution network based on historical load data, and match the historical load data with the power users to obtain the historical load data of each power user. Then sort the historical load data by recording time. S3. Based on the load variation rate and recording time of historical load data, select valid time periods for power users, and analyze the load fluctuation range of power users in each time period based on the historical load data sorted by the valid time periods and the recording time to obtain the corresponding load fluctuation range of power users. S4. Extract the historical load data of the latest recorded time to classify and match power users. Set the comprehensive fluctuation range corresponding to different categories according to the classification and matching results. Then, predict the predicted load data of power users based on the historical load data. Based on the accuracy of the historical predicted load data of power users, dynamically adjust the predicted load data of power users in the same category in the load fluctuation range and comprehensive fluctuation range. S5. Summarize the adjusted forecast load data of all electricity users and convert it into a comprehensive health index. Compare the comprehensive health index with the risk health index. If the comprehensive health index is higher than the risk health index, an early warning will be triggered.

2. The load forecasting method for urban medium and low voltage distribution networks based on PHM according to claim 1, characterized in that: In S1, the historical load data of the medium and low voltage distribution network is collected from the data management section in the same power supply area by the distribution network dispatch center connected to the distribution network, and the PHM data of the medium and low voltage distribution network equipment is obtained at the same time.

3. The load forecasting method for urban medium and low voltage distribution networks based on PHM according to claim 2, characterized in that: The risk health index analysis is performed based on PHM data to obtain the risk health index of medium and low voltage distribution networks; The process involves first normalizing various PHM data, then using a weighted summation method to calculate the health sub-index of a single device, and finally determining the weight of each device's health sub-index based on the entropy weight method, and finally weighting them to obtain the risk health index of the medium and low voltage distribution network.

4. The urban low-voltage distribution network load forecasting method based on PHM according to claim 1, characterized in that: In S2, power users of the medium and low voltage distribution network are extracted based on historical load data. According to the distribution network area affiliation, the load data of the medium voltage feeder is broken down to each subordinate distribution transformer area. Then, the total load data of the distribution transformer area is broken down to individual power users according to the user electricity metering ledger, thereby establishing a correlation and matching between historical load data and power users. Based on timestamps, the historical load data of each power user are sorted in ascending order from morning to night to form a time-series load dataset.

5. The load forecasting method for urban medium and low voltage distribution networks based on PHM according to claim 4, characterized in that: In step S3, load data at adjacent recorded time points in the time-series load dataset are selected to calculate the load variation rate, thus obtaining the load variation rate for adjacent time periods. Set a load fluctuation rate threshold, then compare and filter the load fluctuation rates of adjacent time periods with the load fluctuation rate threshold, and select the time period with the most consecutive load fluctuation rates within the load fluctuation rate threshold range, and take this time period as the valid time period for the power user. For each selected valid time period, extract all historical load data of each power user within that time period, statistically obtain the maximum and minimum load values ​​within that time period, and calculate the difference between the maximum and minimum values, which is the load fluctuation range of the corresponding valid time period for that power user.

6. The load forecasting method for urban medium and low voltage distribution networks based on PHM according to claim 1, characterized in that: In S4, multiple power user categories are established, historical load data of the latest recorded time is extracted to classify and match power users, and the electricity consumption nature and load characteristics of power users are extracted from the historical load data. Power users are matched with the corresponding power user categories according to the electricity consumption nature and load characteristics. The latest record time can be set according to the fluctuation frequency of the power user; Electricity user categories correspond to various categories with different total electricity loads; Based on the category of electricity users, the load fluctuation range of electricity users in the same category is statistically analyzed, and the mean and standard deviation of the load fluctuation range of all users in that category are calculated. Then, the mean ± 2 times the standard deviation is taken as the comprehensive fluctuation range of users in that category.

7. The load forecasting method for urban medium and low voltage distribution networks based on PHM according to claim 6, characterized in that: In S4, a time-series forecasting method is used to predict the load data for each period in the next 24 hours based on the historical load data of each power user after sorting. This yields the initial predicted load data. Retrieve historical predicted load data and historical load data of each power user for the same period in recent times to analyze the prediction accuracy and obtain the prediction accuracy of each power user. Specifically, if the prediction accuracy is higher than 90%, the overall fluctuation range of the user's category will be widened by 10%. When the prediction accuracy is below 80%, the overall fluctuation range will be narrowed by 10%. Then, the initial forecast load data is compared with the load fluctuation range of the corresponding effective period and the adjusted comprehensive fluctuation range. If the initial forecast value exceeds any range, it is corrected to the range to obtain the adjusted forecast load data. Among these, the predicted load data is limited to not exceeding the comprehensive fluctuation range and the load fluctuation range.

8. The load forecasting method for urban medium and low voltage distribution networks based on PHM according to claim 7, characterized in that: In S5, the adjusted predicted load data of all power users are summarized, and the total predicted load data of each distribution area and the entire distribution network are obtained according to the topological affiliation of medium-voltage feeders and low-voltage distribution areas. In the distribution network dispatch center, query the maximum allowable load value of each equipment in the medium and low voltage distribution network, calculate the ratio of the total predicted load data to the maximum allowable load value, and map this ratio to the health index range of [0,1] according to the preset mapping rules. The smaller the ratio, the closer the comprehensive health index is to 1, and vice versa. Finally, the comprehensive health index of the entire distribution network is obtained.

9. The load forecasting method for urban medium and low voltage distribution networks based on PHM according to claim 8, characterized in that: The comprehensive health index of the entire power distribution network is compared with the risk health index set in S1. When the comprehensive health index exceeds the risk health index, an early warning signal is immediately sent to the distribution network dispatch center. If the overall health index does not exceed the risk health index, monitoring will continue.

Images