Distribution line load prediction and optimal scheduling method and system

Abstract

The invention discloses a distribution line load prediction and optimal scheduling method and system, and relates to the technical field of intelligent scheduling of power systems, and the method comprises the steps: generating a time-aligned multi-source fusion input data set; constructing a mixed time sequence load prediction model, and introducing a weighted quantile loss function in a model training process; constructing a joint probability distribution model of the renewable energy output and demand response participation rate, and sampling joint probability distribution; constructing a rolling time domain power distribution network optimization scheduling model; a two-layer mixed strategy is adopted to deal with uncertainty, an optimization problem is decomposed into a plurality of sub-problems, and an alternating direction multiplier method with adaptive penalty parameters is used for distributed solution. According to the method, a multi-objective optimization scheduling model is established in a rolling time domain, and dynamic closed-loop optimization is realized; and by introducing a two-layer hybrid solving strategy and an ADMM distributed algorithm with an adaptive penalty parameter, the calculation efficiency and expandability are remarkably improved while the global consistency is ensured.

Description

Technical Field

[0001] This invention relates to the field of intelligent dispatching technology for power systems, and in particular to a method and system for predicting and optimizing the load of distribution lines. Background Technology

[0002] With the rapid integration of new loads such as renewable energy, distributed power sources, electric vehicles, and demand response, the operating characteristics of distribution networks are gradually evolving from the traditional "unidirectional power flow and deterministic loads" to a complex structure characterized by "multi-source heterogeneity, bidirectional power flow, and strong randomness." Traditional distribution network dispatching models mainly rely on empirical rules or deterministic optimization methods, which are insufficient to effectively cope with the random fluctuations in load and renewable energy output. This leads to problems such as voltage exceeding limits on distribution lines, increased active power losses, and frequent triggering of equipment operating constraints. Existing technologies for load forecasting in distribution networks mainly include time series analysis, neural network forecasting, and grey forecasting. These methods have a certain accuracy in short-term forecasting, but most only focus on a single time scale or a single forecasting index, ignoring the non-stationarity of load characteristics and the correlation at multiple time levels. At the same time, they do not adequately characterize the correlation between renewable energy output (such as photovoltaic and wind power) and user response behavior, resulting in forecast results that cannot fully reflect system uncertainties. At the optimization dispatching level, existing distribution network dispatching methods mostly adopt a centralized solution framework, which models and solves the system's power flow constraints, equipment operating constraints, and economic objectives in a unified manner. This approach is feasible in small and medium-sized systems, but in smart distribution networks containing a large number of distributed energy sources and multi-layered control nodes, it suffers from high computational complexity, heavy data communication pressure, and difficulty in balancing privacy protection. Furthermore, traditional deterministic scheduling models fail to adequately consider the impact of prediction errors and scenario fluctuations, easily leading to a mismatch between the scheduling plan and the actual operating state.

[0003] In recent years, academia and industry have gradually introduced the concept of rolling time-domain optimization and distributed solution algorithms to achieve dynamic scheduling and hierarchical coordinated control of power distribution networks. However, existing solutions generally suffer from the following shortcomings:

[0004] The rolling optimization process did not fully incorporate the uncertain distribution characteristics of load forecasting, resulting in a disconnect between forecasting and scheduling; the multi-objective optimization lacked a unified weighting and dynamic weight adjustment mechanism, making it difficult to simultaneously consider economy, security, and user comfort; and the distributed solution process failed to achieve adaptive coordination among sub-networks, leading to slow convergence speed or inconsistent solutions.

[0005] Therefore, there is an urgent need for a distribution line load forecasting and optimal scheduling method that can jointly consider load forecasting results, uncertainty modeling and distributed solution mechanism in the rolling time domain, so as to achieve high-precision forecasting, high-robust scheduling and real-time collaborative optimization of complex distribution network systems. Summary of the Invention

[0006] In view of the above-mentioned problems, the present invention is proposed.

[0007] Therefore, the problem to be solved by this invention is how to achieve high-precision load forecasting and optimized scheduling in the rolling time domain for distribution lines under the conditions of high proportion of renewable energy access and high uncertainty of load-side behavior, so as to take into account the economy, safety and reliability of system operation.

[0008] To solve the above-mentioned technical problems, the present invention provides the following technical solution: In a first aspect, embodiments of the present invention provide a method for load prediction and optimized scheduling of power distribution lines, including collecting and storing data related to the operation of the power distribution network, performing data quality processing, and generating a time-aligned multi-source fusion input dataset; Using a multi-source fusion input dataset as input, a hybrid time-series load prediction model is constructed. During the model training process, a weighted quantile loss function is introduced to optimize the prediction accuracy of the prediction mean and multiple quantiles. Based on the uncertainty characteristics of distributed energy monitoring data and meteorological elements in the data related to distribution network operation, a joint probability distribution model of renewable energy output and demand response participation rate is constructed. A distribution-based scenario generation method is used to sample the joint probability distribution and generate multiple sets of renewable energy output and demand response participation rate scenarios in the future time domain. A rolling time-domain distribution network optimization scheduling model is constructed, and the distribution network operation performance is jointly optimized in the form of a weighted multi-objective function within the prediction time range of the rolling time domain. A two-layer hybrid strategy is adopted to handle uncertainty. The optimization problem is decomposed into multiple subproblems and solved in a distributed manner using the alternating direction multiplier method with adaptive penalty parameters.

[0009] As a preferred embodiment of the power distribution line load prediction and optimal scheduling method of the present invention, wherein: Hybrid time-series load forecasting models include: A spatial encoder based on graph neural networks is used to extract spatial features of the topology of power distribution lines and the electrical relationships between nodes. A time-series decoder based on an improved self-attention mechanism is used to model the temporal dependency of load change characteristics at different time scales. An uncertainty estimation module based on Monte Carlo Dropout and deep belief networks is used to model the probability distribution of the prediction results and output the load probability distribution P(L) at each prediction time t. t ).

[0010] As a preferred embodiment of the power distribution line load prediction and optimal scheduling method of the present invention, wherein:

[0011] A weighted quantile loss function is introduced to optimize the prediction accuracy of the predicted mean and multiple quantiles, including:

[0012] The weighted quantile loss function is applied to different quantiles q∈ The prediction results of {0.1, 0.5, 0.9} are jointly optimized, and the result is in the following form:

[0013] ;

[0014] in, Here, y represents the weighted quantile loss function, and y represents the actual load value. This is the predicted value corresponding to quantile q. where is the weighting coefficient for each quantile q. ( ) represents the quantile loss function;

[0015] The definition of the quantile loss function is:

[0016] ;

[0017] in, This represents the prediction error;

[0018] During model training, a dynamic weighting mechanism is employed, which dynamically adjusts the weighting based on the quantile error on the validation set. The forecast mean and interval uncertainty are jointly optimized.

[0019] As a preferred embodiment of the distribution line load forecasting and optimal scheduling method of the present invention, the method includes: constructing a joint probability distribution model of renewable energy output and demand response participation rate, comprising:

[0020] Marginal probability models were established for photovoltaic power output, wind power output, and demand response rate, respectively. Photovoltaic and wind power output were fitted with Gaussian mixture distributions to capture their asymmetry under extreme weather conditions. The demand response participation rate was modeled using either a truncated normal distribution or a log-normal distribution based on historical response probability characteristics.

[0021] To characterize the dependencies between different variables, a joint distribution based on the Copula function is constructed:

[0022] ;

[0023] Among them, C( ) represents the correlation function. For the joint cumulative distribution function, Let be the edge cumulative distribution function of photovoltaic power output. Let be the edge cumulative distribution function of wind power. Let be the marginal cumulative distribution function of the demand response participation rate. , and It is a random variable.

[0024] As a preferred embodiment of the power distribution line load prediction and optimal scheduling method of the present invention, wherein: a distribution-based scenario generation method is used to sample the joint probability distribution, including:

[0025] The Copula function model is used to model the nonlinear correlation between meteorological variables and renewable energy output, or the Gaussian mixture model is used to fit and sample the output distribution characteristics under different operating conditions.

[0026] In each generated scenario, the probability of occurrence of each generated scenario is determined. The multiple sets of generated renewable energy output and demand response participation rate scenarios, along with the probability of occurrence of each generated scenario, are output together as input to the subsequent optimization scheduling model to characterize the uncertainty of renewable energy output and demand response.

[0027] As a preferred embodiment of the distribution line load forecasting and optimal scheduling method of the present invention, the method includes: constructing a rolling time-domain distribution network optimal scheduling model, and jointly optimizing the distribution network operation performance in the form of a weighted multi-objective function within the prediction time range of the rolling time domain, including:

[0028] The scheduling time is divided into continuous rolling periods, each corresponding to a prediction window. Within each period, optimization calculations are performed based on the latest observation data and prediction results, and the first decision at the current moment is executed.

[0029] After each execution, the scrolling window moves forward, data is collected again, forecasts and scenarios are updated, and the optimization execution is repeated to achieve closed-loop control; probabilistic load forecasting is combined with renewable energy output and demand response scenarios to form multiple sets of operating scenarios.

[0030] As a preferred embodiment of the power distribution line load forecasting and optimal scheduling method of the present invention, the method employs a two-layer hybrid strategy to handle uncertainties, decomposes the optimization problem into multiple sub-problems, and uses the alternating direction multiplier method with adaptive penalty parameters for distributed solution, including:

[0031] By introducing risk budget constraints to key constraints, the original probabilistic constraints are transformed into solvable deterministic constraints.

[0032] For the short-term fine-tuning stage in the rolling time domain, the model predictive control concept is adopted. Each optimization step starts from the current observation and dynamically adjusts the electromechanical resource output, energy storage charging and discharging, and demand response in combination with the weighted period of multiple scenarios.

[0033] The decomposed subproblems are solved iteratively using ADMM, with each subnet independently calculating its local optimum to ensure constraint satisfaction. Convergence and consistency correction are performed during iterations using boundary variables, and Lagrange multipliers and penalty terms are used to guide the solutions of each subnet towards global consistency. ADMM incorporates constraints into Lagrangian form and adds a quadratic penalty term to form an enhanced Lagrangian function. ;

[0034] in, To enhance the Lagrange function, Let m be the local objective function of subnet m. For Lagrange multipliers, This is the coefficient for secondary penalties. For globally coordinated boundary variables, Let m be the set of local decision variables for subnet m. Let m be the boundary variable of subnet m.

[0035] Secondly, embodiments of the present invention provide a power distribution line load forecasting and optimized scheduling system, comprising:

[0036] The data acquisition and preprocessing module is used to collect and store data related to the operation of the power distribution network, perform data quality processing, and generate a time-aligned multi-source fusion input dataset.

[0037] The load forecasting module is used to construct a hybrid time-series load forecasting model using a multi-source fusion input dataset as input. During the model training process, a weighted quantile loss function is introduced to optimize the forecasting mean and the forecasting accuracy of multiple quantiles.

[0038] The demand generation module is used to construct a joint probability distribution model of renewable energy output and demand response participation rate based on the uncertainty characteristics of distributed energy monitoring data and meteorological elements in the data related to distribution network operation. It adopts a distribution-based scenario generation method to sample the joint probability distribution and generate multiple sets of renewable energy output and demand response participation rate scenarios in the future time domain.

[0039] The scheduling model building module is used to construct a rolling time domain distribution network optimization scheduling model, which jointly optimizes the operation performance of the distribution network in the form of a weighted multi-objective function within the prediction time range of the rolling time domain.

[0040] The strategy generation module is used to handle uncertainty using a two-layer hybrid strategy, decomposing the optimization problem into multiple subproblems and solving them in a distributed manner using the alternating direction multiplier method with adaptive penalty parameters.

[0041] Thirdly, embodiments of the present invention provide a computer device, including a memory and a processor, wherein the memory stores a computer program, wherein: when the computer program instructions are executed by the processor, they implement the steps of the power distribution line load prediction and optimization scheduling method as described in the first aspect of the present invention.

[0042] Fourthly, embodiments of the present invention provide a computer-readable storage medium having a computer program stored thereon, wherein: when the computer program instructions are executed by a processor, they implement the steps of the power distribution line load prediction and optimization scheduling method as described in the first aspect of the present invention.

[0043] The beneficial effects of this invention are as follows: This invention establishes a multi-objective optimization scheduling model in the rolling time domain, taking into account active power loss, node voltage deviation, operating cost and risk penalty, to achieve dynamic closed-loop optimization of prediction and scheduling; and by introducing a two-layer hybrid solution strategy and an ADMM distributed algorithm with adaptive penalty parameters, it significantly improves computational efficiency and scalability while ensuring global consistency. Under the conditions of high proportion of renewable energy access and load fluctuation, it achieves high-precision prediction, robust scheduling and intelligent collaborative control of the distribution network, thereby improving the safety, economy and reliability of system operation. Attached Figure Description

[0044] To more clearly illustrate the technical solutions of the embodiments of the present invention, the accompanying drawings used in the description of the embodiments will be briefly introduced below. Obviously, the accompanying drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0045] Figure 1 A flowchart of a method for load forecasting and optimal scheduling of power distribution lines; Figure 2 This is a schematic diagram of the structure of a power distribution line load forecasting and optimization dispatching system; Figure 3 A schematic diagram of a dielectric structure for a method and system of load forecasting and optimal scheduling of power distribution lines; Figure 4 This diagram shows the computational equipment for a method and system for predicting and optimizing the load of power distribution lines.

[0046] In the diagram: 30, optical disc; 40, computing device; 401, processing unit; 402, system memory; 403, bus; 404, external device; 405, I / O interface; 406, network adapter; 4021, RAM; 4022, cache memory; 4023, ROM; 4024, program module; 4025, utility. Detailed Implementation

[0047] To make the above-mentioned objects, features and advantages of the present invention more apparent and understandable, the specific embodiments of the present invention will be described in detail below with reference to the accompanying drawings.

[0048] Many specific details are set forth in the following description in order to provide a full understanding of the invention. However, the invention may also be practiced in other ways different from those described herein, and those skilled in the art can make similar extensions without departing from the spirit of the invention. Therefore, the invention is not limited to the specific embodiments disclosed below.

[0049] Secondly, the term "one embodiment" or "embodiment" as used herein refers to a specific feature, structure, or characteristic that may be included in at least one implementation of the present invention. The phrase "in one embodiment" appearing in different places in this specification does not necessarily refer to the same embodiment, nor is it a single or selective embodiment that is mutually exclusive with other embodiments.

[0050] Example 1

[0051] Reference Figures 1-4 This is the first embodiment of the present invention, which provides a method for load prediction and optimized scheduling of power distribution lines, including: S1: Collect and store data related to the operation of the power distribution network, perform data quality processing, and generate a time-aligned multi-source fusion input dataset.

[0052] Furthermore, time-series interpolation or autoregressive interpolation based on historical patterns is used to recover data lost due to sensor interruptions; sliding window statistics and Z-score detection algorithms are used to identify abrupt changes or out-of-limit values, and abnormal data affected by faults or communication noise are removed or corrected; based on the master station clock and the timestamps of each terminal device, data of different sampling periods (1min, 5min, 15min) are interpolated, resampled, and time-aligned; and various voltage, current, power, temperature, wind speed, and other data are standardized and converted according to unified engineering units.

[0053] Furthermore, after data cleaning, the data is fused and features are constructed based on time, space, and statistical characteristics: historical averages, volatility, intraday cycle features, and weekly cycle features are extracted from the load curve; weather elements and calendar features are encoded as external input variables; the operating status and power distribution of feeders, transformers, and distributed energy nodes are mapped to a unified node attribute vector through a topology matrix; and the min-max normalization or Z-score standardization methods are used to adjust each feature value to a unified dimension range to prevent model training deviation.

[0054] S2: Using a multi-source fusion input dataset as input, a hybrid time-series load prediction model is constructed. During the model training process, a weighted quantile loss function is introduced to optimize the prediction accuracy of the prediction mean and multiple quantiles.

[0055] Furthermore, the hybrid time-series load forecasting model includes: a spatial encoder based on graph neural networks for extracting spatial features of the distribution line topology and electrical connections between nodes; a time-series decoder based on an improved self-attention mechanism for time-series dependency modeling of load change characteristics at different time scales; and an uncertainty estimation module based on Monte Carlo Dropout and deep belief networks for probabilistic distribution modeling of the forecast results, outputting the load probability distribution P(L) at each forecast time t. t ).

[0056] Furthermore, the weighted quantile loss function is applied to different quantiles q∈ The prediction results of {0.1, 0.5, 0.9} are jointly optimized, and the result is in the following form:

[0057] ;

[0058] in, Here, y represents the weighted quantile loss function, and y represents the actual load value. This is the predicted value corresponding to quantile q. where is the weighting coefficient for each quantile q. ( ) represents the quantile loss function;

[0059] The definition of the quantile loss function is:

[0060] ;

[0061] in, This represents the prediction error;

[0062] During model training, a dynamic weighting mechanism is employed, which dynamically adjusts the weighting based on the quantile error on the validation set. The forecast mean and interval uncertainty are jointly optimized.

[0063] Furthermore, the spatial encoder performs neighborhood aggregation operations on each node through multi-layer graph convolution operations to extract spatially relevant features:

[0064] ;

[0065] in, Let i be the feature vector of node i in the (l+1)th layer. For the adjacency weight, σ( Let be a non-linear activation function, and let N(i) be the set of neighbors of node i. For nodes The feature vector of the l-th layer, For bias terms;

[0066] Through multi-level iterations, the node-level spatial embedding representation H is obtained. spatial It is used to reflect the structural characteristics and power distribution patterns of the power distribution network.

[0067] The temporal decoder uses node-level spatial embedding to represent H spatial Using the load sequence within the historical time window as input, an improved self-attention mechanism is employed for time dependency modeling. In traditional attention mechanisms, long-term dependencies are easily overwhelmed by local features. Therefore, this invention performs multi-scale normalization on the attention weights and introduces a time decay factor γ, making the model more sensitive to recent changes while maintaining a weaker focus on trends in distant timeframes.

[0068] ;

[0069] in, Let k be the attention weights from time t. K and K are the query matrix and key matrix, respectively. Let be the query vector at time t. Let K be the key vector at time k. Let be the key vector of the nth sequence index, and γ be the time decay coefficient. Let k be the dimension of the key vector, and n be all possible historical time indices;

[0070] After multi-head attention and feedforward network layer calculations, the temporal feature representation H is obtained. temporal It can accurately capture load change patterns at different time scales, including periodic trends and sudden disturbances.

[0071] At the output end, the sample distribution is further density estimated to obtain the load probability distribution P(L) at each prediction time t. t ), and calculate the confidence interval.

[0072] Furthermore, to enhance the model's adaptability to different load scenarios, a dynamic weighting mechanism is introduced:

[0073] Initial stage: Let w 0.1 =0.3,w 0.5 =0.4,w 0.9 =0.3; After each training round, calculate the mean absolute percentage error of each quantile on the validation set; if the error of high quantiles is large, automatically increase w. 0.9 ;

[0074] If the error at lower quantiles is large, then increase w. 0.1 This enhances robustness in low-load areas.

[0075] S3: Based on the uncertainty characteristics of distributed energy monitoring data and meteorological elements in the distribution network operation data, a joint probability distribution model of renewable energy output and demand response participation rate is constructed. A distribution-based scenario generation method is used to sample the joint probability distribution and generate multiple sets of renewable energy output and demand response participation rate scenarios in the future time domain.

[0076] Furthermore, marginal probability models were established for photovoltaic power output, wind power output, and demand response rate, respectively. Photovoltaic and wind power output were fitted with Gaussian mixture distributions to capture their asymmetry under extreme weather conditions. The demand response participation rate was modeled using either a truncated normal distribution or a log-normal distribution based on historical response probability characteristics.

[0077] To characterize the dependencies between different variables, a joint distribution based on the Copula function is constructed:

[0078] ;

[0079] Among them, C( ) represents the correlation function. For the joint cumulative distribution function, Let be the edge cumulative distribution function of photovoltaic power output. Let be the edge cumulative distribution function of wind power. Let be the marginal cumulative distribution function of the demand response participation rate. , and It is a random variable.

[0080] Furthermore, the Copula function model is used to model the nonlinear correlation between meteorological variables and renewable energy output, or the Gaussian mixture model is used to fit and sample the output distribution characteristics under different operating conditions.

[0081] In each generated scenario, the probability of occurrence of each generated scenario is determined. The generated sets of multiple renewable energy output and demand response participation rate scenarios and the probability of occurrence of each scenario are output together as input to the subsequent distribution network optimization scheduling model to characterize the uncertainty of renewable energy output and demand response.

[0082] It should be noted that by normalizing each power generation output to [0,1] according to its rated capacity, the demand response rate R... tMaintain a proportionality (0-1); if there are a large number of 0s or 1s, mark them as special samples in advance. Perform detrending and deseasonalization on the time series: if intraday and seasonality are very significant, model using time periods (hours) or remove the seasonal term using regression before building the marginal distribution; truncate or model sparse extreme values ​​separately. Divide the data into training, validation, and test sets, and remove or correct outliers.

[0083] If the demand response rate R t Most values ​​are continuously distributed in (0,1), and there are no large numbers of 0s or 1s. Use a Beta distribution, or a log-normal or truncated normal distribution. If a large number of values ​​are 0 (no response) or 1 (full response), use a zero-inflation model: [The text then abruptly shifts to a different topic:] ...the demand response rate R... t The model is considered a mixture of point-mass and continuous distributions, with the remainder following a truncated normal. Steps: Estimate the existence of zero- and one-inflation; if it is a simple continuous distribution, fit the truncated normal or log-normal parameters using MLE, and save the CDF and inverse CDF. If it is a mixture model, estimate the parameters using EM or by directly optimizing the mixture's log-likelihood.

[0084] Furthermore, historical data from similar time periods or under similar meteorological conditions are used as samples. The number of components K is selected using BIC and AIC or cross-validation, and the Gaussian Mixture Model (GMM) is fitted using the EM algorithm to obtain component weights, mean, and variance. If multiple grid connection points are being jointly modeled, a GMM can be fitted to each point individually, and then Copula can be used to couple the dependencies between different points; alternatively, a GMM can be directly fitted to the aggregated output. First, components are randomly selected according to their weights, and then output values ​​are sampled from those Gaussian components; if there are physical boundaries (e.g., non-negative, upper limit), the sampling results are truncated or resampled.

[0085] Set up to generate N sample trajectories (e.g., N=5000-20000), each trajectory being a sequence of PV, WIND, and DR values ​​for the entire prediction window. If an independent sampling method is used, time correlation can be recovered through residual autoregression or bootstrap. Extract representative feature vectors for each trajectory, such as peak value, peak location, mean, standard deviation, energy, maximum drop rate, correlation index, etc.; or directly use the entire trajectory vector. For each cluster c, select the cluster center or cluster centroid as a set of multiple renewable energy output and demand response participation rate scenarios. The probability of occurrence for each generated scenario is calculated based on the proportion of samples within the cluster.

[0086] S4: Construct a rolling time-domain distribution network optimization scheduling model, and jointly optimize the distribution network operation performance in the form of a weighted multi-objective function within the predicted time range of the rolling time domain.

[0087] Furthermore, the scheduling time is divided into continuous rolling periods, each corresponding to a prediction window. Within each period, optimization calculations are performed based on the latest observation data and prediction results, and the first decision at the current moment is executed.

[0088] After each execution, the scrolling window moves forward, data is collected again, forecasts and scenarios are updated, and the optimization execution is repeated to achieve closed-loop control; probabilistic load forecasting is combined with renewable energy output and demand response scenarios to form multiple sets of operating scenarios.

[0089] It should be noted that the daily scheduling cycle is divided into several consecutive rolling periods, for example, one rolling period every 15 minutes. For each rolling period k, the corresponding prediction time range of the rolling time domain is defined, which can be 1-4 hours; the rolling step size is determined to ensure that the output of each optimization only executes the first step decision. At the beginning of the rolling period k, the latest observation data is collected, including real-time load of the distribution network, renewable energy output, meteorological information, demand response participation rate and available capacity, switch status and topology and other equipment parameters; the data is preprocessed: missing value imputation, outlier removal, normalization, and a time-aligned multi-source input dataset is generated; the time-aligned multi-source input dataset is input into the hybrid time series prediction model, and the load probability distribution P(L) at each prediction time is output. t The GNN spatial encoder extracts line topology and electrical connections between nodes; the self-attention decoder models temporal dependencies; Monte Carlo Dropout and deep belief network estimate uncertainty; and a weighted quantile loss function is used to optimize the predicted mean and quantiles, enhancing the accuracy of peak and off-peak load predictions. The output load prediction is the probability distribution at each time point, which can be used for risk assessment in subsequent scheduling.

[0090] S5: A two-layer hybrid strategy is adopted to handle uncertainty. The optimization problem is decomposed into multiple subproblems and solved in a distributed manner using the alternating direction multiplier method with adaptive penalty parameters.

[0091] Furthermore, risk budget constraints are introduced for key constraints, transforming the original probabilistic constraints into solvable deterministic constraints; for the short-term fine-tuning stage in the rolling time domain, the model predictive control concept is adopted, with each optimization step starting from the current observation and dynamically adjusting the electromechanical resource output, energy storage charging and discharging, and demand response in combination with the weighted period of multiple scenarios.

[0092] The decomposed subproblems are solved iteratively using ADMM, with each subnet independently calculating its local optimum to ensure constraint satisfaction. Convergence and consistency correction are performed during iterations using boundary variables, and Lagrange multipliers and penalty terms are used to guide the solutions of each subnet towards global consistency. ADMM incorporates constraints into Lagrangian form and adds a quadratic penalty term to form an enhanced Lagrangian function. ;

[0093] in, To enhance the Lagrange function, Let m be the local objective function of subnet m. For Lagrange multipliers, This is the coefficient for secondary penalties. For globally coordinated boundary variables, Let m be the set of local decision variables for subnet m. Let m be the boundary variable of subnet m.

[0094] It should be noted that the boundary variables of all subnets The system converges to form a globally coordinated boundary variable z, and performs consistency correction in each iteration. A weighted convergence strategy is employed to dynamically adjust the influence weights of each subnetwork on the global boundary variable based on subnetwork importance or the sensitivity of key grid nodes, improving global scheduling accuracy. Each subnetwork independently solves the enhanced Lagrangian function, enabling parallel computation and significantly reducing overall solution time. Combining dynamic convergence of boundary variables with Lagrange multiplier adjustment achieves distributed iterative solution while ensuring global constraint satisfaction, suitable for large-scale distribution network rolling scheduling. Using probabilistic load forecasting, renewable energy output, and demand response scenarios as inputs, and combining this with the aforementioned distributed ADMM solution, each round of rolling time-domain optimization finds local optima in multiple scenarios, while ensuring global consistency through boundary variables and the Lagrange multiplier mechanism. Integrating distributed ADMM with probabilistic rolling optimization improves the robustness and economy of the distribution network under load fluctuations, DER output uncertainty, and demand response participation.

[0095] Example 2

[0096] This embodiment is the second embodiment of the present invention, which provides a method for predicting and optimizing the scheduling of power distribution line loads. In order to verify the beneficial effects of the present invention, a simulation experiment is conducted for scientific demonstration.

[0097] Set up a medium-voltage distribution network, including 3 substations, 12 feeders, and a total of 120 nodes, equipped with smart meters, distributed energy monitoring devices, meteorological monitoring equipment, and historical archives.

[0098] Training strategy: Weighted quantile loss function, quantile set ={0.1,0.5,0.9}, high quantile weighting is used to improve the accuracy of peak load prediction; low quantile weighting is used to enhance the robustness of low-valley load prediction, and the load probability distribution for the next 24 hours is output.

[0099] Sampling and generating 50 sets of renewable energy output and demand response participation rate scenarios for the next 24 hours, and the probability of occurrence for each generated scenario;

[0100] Table 1 Data on Renewable Energy and Demand Response Scenarios

[0101]

[0102] The simulation uses a rolling time period every hour and executes the first optimization decision every minute. It employs scenario trees and risk budgeting to transform probabilistic constraints into deterministic constraints. Based on model predictive control, it starts with current observations and combines weighted multi-scenario or worst-case scenario solutions. The simulation conditions include a 24-hour time range and a 1-minute time step.

[0103] Load forecasting and optimized scheduling are updated in real time and solved in parallel using distributed computing in subnets.

[0104] Table 2 Comparison of Simulation Results

[0105]

[0106] Table 2 illustrates the following: Active power loss in distribution lines is reduced by 12%. This invention's method, through probabilistic load forecasting and multi-scenario power flow optimization, achieves more balanced node power distribution, reduces feeder current during peak loads, and decreases line losses. The number of node voltage exceedances is reduced by 80%. In traditional methods, randomly fluctuating photovoltaic output easily causes voltage fluctuations. This invention introduces confidence interval scheduling and risk budget constraints to effectively suppress voltage exceedances, significantly improving the voltage stability and operational safety of the distribution network. DER scheduling costs are reduced by 13.7%. This invention, through joint optimization of DER output and demand response participation rate, enables energy storage devices to charge during low-price periods and discharge during high-price periods, resulting in more economical overall scheduling. Demand response penalty costs are reduced by 40%. Traditional deterministic strategies lead to user response exceeding or falling short of limits. This invention uses scenario generation and response probability distribution fitting to make scheduling signals more consistent with actual responsiveness. Solution time is reduced by 64%. This invention uses ADMM distributed solution, splitting the global problem into six subnets for parallel computation. Adaptive penalty terms adjust the convergence speed, reducing the average solution time per step from 2.5 seconds to 0.9 seconds.

[0107] After introducing the method of exemplary embodiments of the present invention, the following references are made. Figure 2 An exemplary embodiment of the present invention provides a power distribution line load forecasting and optimized scheduling system, the system comprising:

[0108] The data acquisition and preprocessing module is used to collect and store data related to the operation of the power distribution network, perform data quality processing, and generate a time-aligned multi-source fusion input dataset.

[0109] The load forecasting module is used to construct a hybrid time-series load forecasting model using a multi-source fusion input dataset as input. During the model training process, a weighted quantile loss function is introduced to optimize the forecasting mean and the forecasting accuracy of multiple quantiles.

[0110] The demand generation module is used to construct a joint probability distribution model of renewable energy output and demand response participation rate based on the uncertainty characteristics of distributed energy monitoring data and meteorological elements in the data related to distribution network operation. It adopts a distribution-based scenario generation method to sample the joint probability distribution and generate multiple sets of renewable energy output and demand response participation rate scenarios in the future time domain.

[0111] The scheduling model building module is used to construct a rolling time domain distribution network optimization scheduling model, which jointly optimizes the operation performance of the distribution network in the form of a weighted multi-objective function within the prediction time range of the rolling time domain.

[0112] The strategy generation module is used to handle uncertainty using a two-layer hybrid strategy, decomposing the optimization problem into multiple subproblems and solving them in a distributed manner using the alternating direction multiplier method with adaptive penalty parameters.

[0113] After introducing the method and system of exemplary embodiments of the present invention, the following references are made. Figure 3 A computer-readable storage medium according to exemplary embodiments of the present invention will be described, please refer to... Figure 3 The computer-readable storage medium shown is an optical disc 30, on which a computer program (i.e., a program product) is stored. When the computer program is run by a processor, it will implement the steps described in the above method implementation method. The specific implementation method of each step will not be repeated here.

[0114] It should be noted that examples of the computer-readable storage medium may also include, but are not limited to, phase-change memory (PRAM), static random access memory (SRAM), dynamic random access memory (DRAM), other types of random access memory (RAM), read-only memory (ROM), electrically erasable programmable read-only memory (EEPROM), flash memory or other optical and magnetic storage media, which will not be elaborated here.

[0115] After introducing the methods and media of exemplary embodiments of the present invention, the following references are made. Figure 4 A computing device for power distribution line load prediction and optimized scheduling according to an exemplary embodiment of the present invention.

[0116] Figure 4 A block diagram is shown of an exemplary computing device 40 suitable for implementing embodiments of the present invention. The computing device 40 may be a computer system or a server. Figure 4 The computing device 40 shown is merely an example and should not impose any limitation on the functionality and scope of use of the embodiments of the present invention.

[0117] like Figure 4As shown, the components of computing device 40 may include, but are not limited to: one or more processors or processing units 401, system memory 402, and bus 403 connecting different system components (including system memory 402 and processing unit 401).

[0118] The computing device 40 typically includes a variety of computer system readable media. These media can be any available media that can be accessed by the computing device 40, including volatile and non-volatile media, and removable and non-removable media.

[0119] System memory 402 may include computer system readable media in the form of volatile memory, such as at least one of random access memory (RAM) 4021 or cache memory 4022. Computing device 40 may further include other removable / non-removable, volatile / non-volatile computer system storage media. By way of example only, ROM 4023 may be used to read and write non-removable, non-volatile magnetic media (…). Figure 4 (Not shown in the image, usually referred to as "hard drive"). Although not shown in... Figure 4 The diagram illustrates that disk drives for reading and writing to removable non-volatile disks (e.g., "floppy disks") and optical disc drives for reading and writing to removable non-volatile optical discs (e.g., CD-ROMs, DVD-ROMs, or other optical media) can be provided. In these cases, each drive can be connected to bus 403 via one or more data media interfaces. System memory 402 may include at least one program product having a set (e.g., at least one) of program modules configured to perform the functions of the embodiments of the present invention.

[0120] A program / utility 4025 having a set (at least one) of program modules 4024 may be stored, for example, in system memory 402, and such program modules 4024 include, but are not limited to, an operating system, one or more application programs, other program modules, and program data, each or some combination of these examples may include an implementation of a network environment. Program modules 4024 typically perform the functions and / or methods described in the embodiments of the present invention.

[0121] The computing device 40 can also communicate with one or more external devices 404 (such as a keyboard, pointing device, display, etc.). This communication can be performed via the input / output (I / O) interface 405. Furthermore, the computing device 40 can also communicate with one or more networks (e.g., local area networks (LANs), wide area networks (WANs), and / or public networks, such as the Internet) via a network adapter 406. Figure 4 As shown, network adapter 406 communicates with other modules of computing device 40 (such as processing unit 401) via bus 403. It should be understood that, although... Figure 4 As not shown, it can be used in conjunction with computing device 40 with other hardware and / or software modules.

[0122] The processing unit 401 executes various functional applications and data processing by running programs stored in the system memory 402.

[0123] Those skilled in the art will understand that, for the sake of convenience and brevity, the specific working processes of the systems, devices, and units described above can be referred to the corresponding processes in the foregoing method embodiments, and will not be repeated here.

[0124] In the several embodiments provided by this invention, it should be understood that the disclosed systems, apparatuses, and methods can be implemented in other ways. The apparatus embodiments described above are merely illustrative. For example, the division of units is only a logical functional division, and in actual implementation, there may be other division methods. Furthermore, multiple units or components may be combined or integrated into another system, or some features may be ignored or not executed. Additionally, the coupling or direct coupling or communication connection shown or discussed may be through some communication interfaces; the indirect coupling or communication connection between devices or units may be electrical, mechanical, or other forms.

[0125] It should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and are not intended to limit it. Although the present invention has been described in detail with reference to preferred embodiments, those skilled in the art should understand that modifications or equivalent substitutions can be made to the technical solutions of the present invention without departing from the spirit and scope of the technical solutions of the present invention, and all such modifications or substitutions should be covered within the scope of the claims of the present invention.

Claims

1. A method for load forecasting and optimal scheduling of power distribution lines, characterized in that, include: Collect and store data related to the operation of the power distribution network, perform data quality processing, and generate a time-aligned multi-source fusion input dataset; Using a multi-source fusion input dataset as input, a hybrid time-series load prediction model is constructed. During the model training process, a weighted quantile loss function is introduced to optimize the prediction accuracy of the prediction mean and multiple quantiles. Based on the uncertainty characteristics of distributed energy monitoring data and meteorological elements in the data related to distribution network operation, a joint probability distribution model of renewable energy output and demand response participation rate is constructed. A distribution-based scenario generation method is used to sample the joint probability distribution and generate multiple sets of renewable energy output and demand response participation rate scenarios in the future time domain. A rolling time-domain distribution network optimization scheduling model is constructed, and the distribution network operation performance is jointly optimized in the form of a weighted multi-objective function within the prediction time range of the rolling time domain. A two-layer hybrid strategy is adopted to handle uncertainty. The optimization problem is decomposed into multiple subproblems and solved in a distributed manner using the alternating direction multiplier method with adaptive penalty parameters.

2. The method for load forecasting and optimized scheduling of power distribution lines as described in claim 1, characterized in that, The hybrid time-series load forecasting model includes: A spatial encoder based on graph neural networks is used to extract spatial features of the topology of power distribution lines and the electrical relationships between nodes. A time-series decoder based on an improved self-attention mechanism is used to model the temporal dependency of load change characteristics at different time scales. An uncertainty estimation module based on Monte Carlo Dropout and deep belief networks is used to model the probability distribution of the prediction results and output the load probability distribution P(L) at each prediction time t. t ).

3. The method for load forecasting and optimized scheduling of power distribution lines as described in claim 1, characterized in that: The introduction of a weighted quantile loss function to optimize the prediction accuracy of the predicted mean and multiple quantiles includes: The weighted quantile loss function is applied to different quantiles q∈ The prediction results of {0.1, 0.5, 0.9} are jointly optimized, and the weighted quantile loss function is in the form of: ； in, Here, y represents the weighted quantile loss function, and y represents the actual load value. This is the predicted value corresponding to quantile q. where is the weighting coefficient for each quantile q. ( ) represents the quantile loss function; The definition of the quantile loss function is: ； in, This represents the prediction error; During model training, a dynamic weighting mechanism is employed, which dynamically adjusts the weighting based on the quantile error on the validation set. The forecast mean and interval uncertainty are jointly optimized.

4. The method for load forecasting and optimized scheduling of power distribution lines as described in claim 1, characterized in that: The construction of the joint probability distribution model for renewable energy output and demand response participation rate includes: Marginal probability models were established for photovoltaic power output, wind power output, and demand response rate, respectively. Photovoltaic and wind power output were fitted with Gaussian mixture distributions to capture their asymmetry under extreme weather conditions. The demand response participation rate was modeled using either a truncated normal distribution or a log-normal distribution based on historical response probability characteristics. To characterize the dependencies between different variables, a joint distribution based on the Copula function is constructed: ； Among them, C( ) represents the correlation function. For the joint cumulative distribution function, Let be the edge cumulative distribution function of photovoltaic power output. Let be the edge cumulative distribution function of wind power. Let be the marginal cumulative distribution function of the demand response participation rate. , and It is a random variable.

5. The method for load forecasting and optimized scheduling of power distribution lines as described in claim 1, characterized in that, The method for generating scenes based on distributions, which samples the joint probability distribution, includes: The Copula function model is used to model the nonlinear correlation between meteorological variables and renewable energy output, or the Gaussian mixture model is used to fit and sample the output distribution characteristics under different operating conditions. In each generated scenario, the probability of occurrence of each generated scenario is determined. The multiple sets of generated renewable energy output and demand response participation rate scenarios, along with the probability of occurrence of each generated scenario, are output together as input to the subsequent distribution network optimization and scheduling model to characterize the uncertainty of renewable energy output and demand response.

6. The method for load forecasting and optimized scheduling of power distribution lines as described in claim 1, characterized in that: The aforementioned construction of a rolling time-domain distribution network optimization scheduling model, within the prediction time range of the rolling time domain, jointly optimizes the distribution network operation performance in the form of a weighted multi-objective function, including: The scheduling time is divided into continuous rolling periods, each rolling period corresponds to a prediction window. Within each rolling period, optimization calculations are performed based on the latest observation data and prediction results, and the first step decision at the current moment is executed. After each execution, the scrolling window moves forward, data is collected again, forecasts and scenarios are updated, and the optimization execution is repeated to achieve closed-loop control; probabilistic load forecasting is combined with renewable energy output and demand response scenarios to form multiple sets of operating scenarios.

7. The method for load forecasting and optimized scheduling of power distribution lines as described in claim 1, characterized in that: The method employs a two-layer hybrid strategy to handle uncertainty, decomposing the optimization problem into multiple sub-problems and using the alternating direction multiplier method with adaptive penalty parameters for distributed solution, including: By introducing risk budget constraints to key constraints, the original probabilistic constraints are transformed into solvable deterministic constraints. For the short-term fine-tuning stage in the rolling time domain, the model predictive control concept is adopted. Each optimization step starts from the current observation and dynamically adjusts the electromechanical resource output, energy storage charging and discharging, and demand response in combination with the weighted period of multiple scenarios. The decomposed subproblems are solved iteratively using ADMM, with each subnet independently calculating its local optimum to ensure constraint satisfaction. Convergence and consistency correction are performed during iterations using boundary variables, and Lagrange multipliers and penalty terms are used to guide the solutions of each subnet towards global consistency. ADMM incorporates constraints into Lagrangian form and adds a quadratic penalty term to form an enhanced Lagrangian function. ; in, To enhance the Lagrange function, Let m be the local objective function of subnet m. For Lagrange multipliers, This is the coefficient for secondary penalties. For globally coordinated boundary variables, Let m be the set of local decision variables for subnet m. Let m be the boundary variable of subnet m.

8. A power distribution line load forecasting and optimized dispatching system, characterized in that, include: The data acquisition and preprocessing module is used to collect and store data related to the operation of the power distribution network, perform data quality processing, and generate a time-aligned multi-source fusion input dataset. The load forecasting module is used to construct a hybrid time-series load forecasting model using a multi-source fusion input dataset as input. During the model training process, a weighted quantile loss function is introduced to optimize the forecasting mean and the forecasting accuracy of multiple quantiles. The demand generation module is used to construct a joint probability distribution model of renewable energy output and demand response participation rate based on the uncertainty characteristics of distributed energy monitoring data and meteorological elements in the data related to distribution network operation. It adopts a distribution-based scenario generation method to sample the joint probability distribution and generate multiple sets of renewable energy output and demand response participation rate scenarios in the future time domain. The scheduling model building module is used to construct a rolling time domain distribution network optimization scheduling model, which jointly optimizes the operation performance of the distribution network in the form of a weighted multi-objective function within the prediction time range of the rolling time domain. The strategy generation module is used to handle uncertainty using a two-layer hybrid strategy, decomposing the optimization problem into multiple subproblems and solving them in a distributed manner using the alternating direction multiplier method with adaptive penalty parameters.

9. A computer device comprising a memory and a processor, wherein the memory stores a computer program, characterized in that: When the processor executes the computer program, it implements the steps of the power distribution line load prediction and optimal scheduling method according to any one of claims 1 to 7.

10. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by the processor, it implements the steps of the power distribution line load prediction and optimal scheduling method according to any one of claims 1 to 7.

Images