A power demand uncertainty analysis method and system based on multi-scenario simulation

Abstract

A method and system for analyzing power demand uncertainty based on multi-scenario simulation is disclosed. The method first collects multi-source heterogeneous operating data of a novel power system to construct a power demand analysis dataset. Multiple types of uncertainty influencing factors are extracted and classified from this dataset. A quantitative model is used to mine three types of features: distribution, deviation, and correlation. A multi-scenario system of baseline, extreme, and coupled scenarios is established. Then, a dynamic scenario simulation model is used to obtain power demand simulation results under different scenarios. Uncertainty assessment and sensitivity analysis are combined to screen key influencing factors. Finally, based on the simulation data and target factors, a multi-scenario power demand uncertainty assessment report is obtained. This invention, by integrating multi-source heterogeneous data, classifying and quantifying various uncertainty influencing factors, and conducting multi-scenario simulations, comprehensively characterizes the uncertainty characteristics of power demand and identifies core influencing factors, thereby improving the accuracy of power demand analysis results and the safe operation of the power system.

Description

Technical Field

[0001] This invention relates to uncertainty analysis methods, belonging to the field of power demand analysis, and particularly to a method and system for power demand uncertainty analysis based on multi-scenario simulation. Background Technology

[0002] Uncertainty analysis of electricity demand is the premise and foundation for the planning and operation of new power systems. Its core is to identify the uncertain factors affecting electricity demand, quantify the degree of impact of uncertainty on electricity demand, predict the fluctuation range of electricity demand under different scenarios, and provide a scientific basis for grid planning, energy storage configuration, and dispatch optimization.

[0003] Existing power demand uncertainty analysis systems mostly rely on statistical analysis methods based on historical data or fixed-parameter prediction models to analyze and assess power demand. These methods typically select historical typical operating data to build empirical models and perform simulations under a few preset operating conditions. Because the processing of input data mainly depends on the continuation of historical patterns and static parameter settings, and because different sources of influence are not differentiated during the analysis process, the model's responsiveness to changes in operating conditions is limited. In actual power grid operation, when load structure changes, meteorological conditions fluctuate, or power output changes, if the aforementioned power demand uncertainty analysis system is still used, the analysis results will deviate significantly from actual demand. This can lead to over- or under-capacity configurations during the power grid planning stage, and amplified load forecasting deviations and delayed scheduling strategy adjustments during the dispatching and operation stage. This can further cause power shortages in local areas, significantly reducing the safety of the new power system operation. Therefore, there is an urgent need for a more accurate power demand uncertainty analysis method based on multi-scenario simulation to address the aforementioned shortcomings of existing technologies. Summary of the Invention

[0004] The purpose of this invention is to overcome the above-mentioned defects and problems in the prior art and to provide a highly accurate method and system for analyzing power demand uncertainty based on multi-scenario simulation.

[0005] To achieve the above objectives, the technical solution of the present invention is: a method for analyzing power demand uncertainty based on multi-scenario simulation, comprising:

[0006] Acquire multi-source heterogeneous operational data to characterize changes in electricity demand in the target new power system, and construct a power demand analysis dataset;

[0007] Candidate uncertainty factors were extracted and classified based on the electricity demand analysis dataset; the classified candidate uncertainty factors include climate uncertainty factors, energy uncertainty factors, and load uncertainty factors.

[0008] The candidate uncertainty influencing factors are quantified based on the uncertainty quantification model, and a multi-scenario system is constructed based on the quantification results. The quantification results include normal distribution characteristics, abnormal deviation characteristics, and correlation structure characteristics. The multi-scenario system includes a baseline scenario, an extreme scenario, and a coupled scenario.

[0009] The quantitative results are input into the dynamic scenario simulation model to obtain the dynamic simulation results of the power demand of the target new power system under multiple scenario systems.

[0010] Based on the dynamic simulation results of electricity demand, candidate uncertainty influencing factors are screened through uncertainty assessment and sensitivity correlation analysis to obtain target uncertainty influencing factors that meet the preset assessment conditions.

[0011] Based on the dynamic simulation results of electricity demand and the factors affecting the uncertainty of the target, a multi-scenario electricity demand uncertainty assessment report is constructed.

[0012] Optionally, the step of acquiring multi-source heterogeneous operational data to characterize changes in the power demand of the target new power system and constructing a power demand analysis dataset specifically includes:

[0013] Acquire electricity demand data and data on uncertain influencing factors as multi-source heterogeneous operation data; the electricity demand data includes historical electricity demand data and real-time electricity demand data; the data on uncertain influencing factors includes climate factor data, energy factor data, load structure data, and power grid operation data;

[0014] Data quality optimization processing was performed on the multi-source heterogeneous operation data, and correlation analysis was used to remove redundant data in the multi-source heterogeneous operation data to obtain a set of power demand analysis data.

[0015] Optionally, the step of performing quantification processing on the classified candidate uncertainty influencing factors based on the uncertainty quantification model specifically includes:

[0016] Based on the analytic hierarchy process combined with the entropy weight method, and by integrating subjective and objective weights, a comprehensive weight is constructed and weights are assigned to the candidate uncertainty influencing factors after classification.

[0017] An uncertainty quantification model is constructed for each different type of candidate uncertainty influencing factor; wherein, the types of the candidate uncertainty influencing factors include random factors, fuzzy factors, and nonlinear correlation factors.

[0018] Optionally, the construction of a multi-scenario system based on the quantification results specifically includes:

[0019] A baseline scenario is constructed based on the normal distribution characteristics in the quantification results; the baseline scenario is a scenario in which the uncertain influencing factors are in the normal fluctuation range and the target new power system maintains a stable power operation state;

[0020] Extreme scenarios are constructed based on the abnormal deviation characteristics in the quantification results; the extreme scenarios are those in which a single uncertain influencing factor is in an abnormal fluctuation range, causing the target new power system to be unable to maintain a stable operating state.

[0021] Coupled scenarios are constructed based on the correlation structure features in the quantification results; the coupled scenario is a synergistic scenario formed by superimposing or combining multiple extreme scenarios according to preset correlation constraints.

[0022] Scenario parameters were set for the baseline scenario, extreme scenario, and coupled scenario, and a multi-scenario system was constructed.

[0023] Optionally, the step of inputting the quantification results into the dynamic scenario simulation model to obtain the dynamic simulation results of the power demand of the target new power system under multiple scenario frameworks specifically includes:

[0024] A dynamic scenario simulation model is constructed based on an improved long short-term memory network and combined with quantification results;

[0025] The quantitative results are input into the dynamic scenario simulation model, and the output is the dynamic simulation results of the power demand of the target new power system under multiple scenarios. The dynamic simulation results of power demand include the time-series variation curve of power demand, fluctuation range, peak load and valley load.

[0026] Optionally, based on the dynamic simulation results of electricity demand, candidate uncertainty influencing factors are screened through uncertainty assessment and sensitivity correlation analysis to obtain target uncertainty influencing factors that meet preset assessment conditions, specifically including:

[0027] Calculate the fluctuation range of electricity demand under various scenarios; the fluctuation range includes the maximum fluctuation range and the average fluctuation range.

[0028] Based on the dynamic simulation results of electricity demand, a probability distribution curve of electricity demand is constructed;

[0029] Based on the fluctuation amplitude value and probability distribution curve, the candidate uncertainty influencing factors of each scenario are classified into risk levels; the risk levels include high risk level, medium risk level and low risk level.

[0030] Based on the risk level, sensitivity quantification is performed on the candidate uncertainty influencing factors to obtain the sensitivity coefficient corresponding to each candidate uncertainty influencing factor.

[0031] Candidate uncertainty influencing factors are ranked and screened based on sensitivity coefficients, and those that meet the preset sensitivity threshold conditions are selected as target uncertainty influencing factors.

[0032] Optionally, after constructing a multi-scenario electricity demand uncertainty assessment report based on the dynamic simulation results of electricity demand and the influencing factors of target uncertainty, the report further includes:

[0033] The report on uncertainty assessment of multi-scenario electricity demand is compared and analyzed with actual electricity demand operation data, and the deviation analysis results are output.

[0034] Based on the results of the deviation analysis, the model parameters of the uncertainty quantification model, the scenario parameters of the multi-scenario system, and the simulation parameters of the dynamic scenario simulation model are subjected to collaborative optimization.

[0035] A power demand uncertainty analysis system based on multi-scenario simulation, which applies the above-mentioned method, the system comprising:

[0036] The data acquisition module is used to acquire multi-source heterogeneous operational data to characterize changes in the power demand of the target new power system and to construct a power demand analysis dataset.

[0037] The uncertainty influencing factor classification module is used to extract and classify candidate uncertainty influencing factors based on the electricity demand analysis data set; the classified candidate uncertainty influencing factors include climate uncertainty factors, energy uncertainty factors, and load uncertainty factors;

[0038] The multi-scenario system construction module is used to perform quantification processing on the classified candidate uncertainty influencing factors based on the uncertainty quantification model, and construct a multi-scenario system based on the quantification results; the quantification results include normal distribution characteristics, abnormal deviation characteristics, and correlation structure characteristics; the multi-scenario system includes a baseline scenario, an extreme scenario, and coupled scenarios;

[0039] The dynamic simulation module is used to input the quantification results into the dynamic scenario simulation model to obtain the dynamic simulation results of the power demand of the target new power system under multiple scenario systems.

[0040] The target uncertainty influencing factor screening module is used to screen candidate uncertainty influencing factors based on the dynamic simulation results of power demand through uncertainty assessment and sensitivity correlation analysis, so as to obtain the target uncertainty influencing factors that meet the preset assessment conditions.

[0041] The uncertainty assessment module is used to construct multi-scenario power demand uncertainty assessment reports based on the results of dynamic simulation of power demand and the factors affecting the uncertainty of the target.

[0042] A device for analyzing power demand uncertainty based on multi-scenario simulation, the device including a processor and a memory;

[0043] The memory is used to store computer program code and to transmit the computer program code to the processor;

[0044] The processor is used to execute the above-described method for analyzing power demand uncertainty based on multi-scenario simulation according to the instructions in the computer program code.

[0045] A computer-readable storage medium storing computer-executable instructions, which, when executed on a computer, implement the aforementioned method for analyzing power demand uncertainty based on multi-scenario simulation.

[0046] Compared with the prior art, the beneficial effects of the present invention are as follows:

[0047] 1. This invention discloses a method and system for analyzing power demand uncertainty based on multi-scenario simulation. The method first collects multi-source heterogeneous operating data of a novel power system to construct a power demand analysis dataset. From this dataset, three categories of uncertainty influencing factors—climate, energy, and load—are extracted and categorized. A quantitative model is used to mine three types of features: distribution, deviation, and correlation. A multi-scenario system of baseline, extreme, and coupled scenarios is established. Then, a dynamic scenario simulation model is used to obtain power demand simulation results under different scenarios. Combining uncertainty assessment and sensitivity analysis, key influencing factors are screened. Finally, based on the simulation data and target factors, a multi-scenario power demand uncertainty assessment report is obtained. In application, this design integrates multi-source heterogeneous data, classifies and quantifies various uncertainty influencing factors, and conducts multi-scenario simulations to comprehensively characterize the uncertainty characteristics of power demand. It achieves accurate identification of core influencing factors, effectively improves the problem of large deviations between traditional analysis and actual operating conditions, enhances the accuracy of power demand analysis results, and ensures the overall safe operation of the power system.

[0048] 2. This invention provides a method and system for analyzing power demand uncertainty based on multi-scenario simulation. It utilizes the analytic hierarchy process (AHP) combined with the entropy weight method, integrating subjective and objective weights to construct a comprehensive weighting system. This system assigns weights to categorized candidate uncertainty influencing factors and then constructs an uncertainty quantification model for each different type of candidate uncertainty influencing factor. In application, this design constructs comprehensive weights to differentiate and quantify uncertainty influencing factors of different categories, such as climate, energy, and load. This unifies various indicators to the same evaluation scale, enabling horizontal data comparability and achieving differentiated quantitative expression and comparable characterization of different types of uncertainty influencing factors under a unified scale.

[0049] 3. The present invention provides a method and system for analyzing power demand uncertainty based on multi-scenario simulation. It quantifies the classified candidate uncertainty influencing factors using an uncertainty quantification model and constructs a multi-scenario system based on the quantification results. In application, this design couples multiple extreme scenarios to form a synergistic scenario, sets scenario parameters for each scenario, and constructs a multi-scenario system. This achieves systematic coverage and unified modeling of the changing characteristics of power demand under normal operation, single extreme disturbance, and multi-source coupling conditions, improving the comprehensiveness, accuracy, and scenario adaptability of power demand analysis. Attached Figure Description

[0050] Figure 1 This is a flowchart of the method of the present invention.

[0051] Figure 2 This is a system structure diagram of the present invention.

[0052] Figure 3 This is a structural diagram of the device of the present invention.

[0053] In the diagram: 1. Data acquisition module; 2. Uncertainty influencing factor classification module; 3. Multi-scenario system construction module; 4. Dynamic simulation module; 5. Target uncertainty influencing factor screening module; 6. Uncertainty assessment module; 7. Processor; 8. Memory; 9. Computer program code. Detailed Implementation

[0054] The present invention will be further described in detail below with reference to the accompanying drawings and specific embodiments.

[0055] Example 1:

[0056] See Figure 1 A method for analyzing electricity demand uncertainty based on multi-scenario simulation, comprising:

[0057] S101. Obtain multi-source heterogeneous operation data to characterize the changes in power demand of the target new power system, and construct a power demand analysis dataset.

[0058] Furthermore, the acquisition of multi-source heterogeneous operational data to characterize changes in the power demand of the target new power system, and the construction of a power demand analysis dataset, specifically includes:

[0059] Acquire electricity demand data and data on uncertain influencing factors as multi-source heterogeneous operation data; the electricity demand data includes historical electricity demand data and real-time electricity demand data; the data on uncertain influencing factors includes climate factor data, energy factor data, load structure data, and power grid operation data;

[0060] Data quality optimization processing is performed on multi-source heterogeneous operation data, and correlation analysis is used to remove redundant data in the multi-source heterogeneous operation data to obtain a power demand analysis data set; the data quality optimization processing includes outlier removal processing, missing value completion processing, and data denoising processing.

[0061] Specifically, regarding the construction and processing of multi-source heterogeneous operation data, the first step is to acquire electricity demand data and data on uncertain influencing factors, and then unify these two data into multi-source heterogeneous operation data. The electricity demand data includes historical electricity demand data and real-time electricity demand data. Historical electricity demand data is used to characterize the changing patterns of electricity load at different time scales, while real-time electricity demand data is used to reflect the dynamic changes in load under the current operating conditions. The electricity demand data includes parameters such as load power, load time distribution, and load fluctuation coefficient to form a multi-dimensional representation of electricity demand changes.

[0062] The uncertainties and influencing factors data include climate factor data, energy factor data, load structure data, and power grid operation data. Climate factor data includes historical and real-time meteorological data such as sunlight, wind speed, temperature, precipitation, and extreme weather conditions like high temperatures, cold waves, and heavy rain, used to characterize the impact of climate disturbances on electricity demand and energy output. Energy factor data includes clean energy output data, conventional power output data, and electricity price fluctuation data, used to reflect the instability of the energy supply side. Load structure data includes the proportion, growth trend, and regulation characteristics of various load types, used to describe the load composition and its evolution. Power grid operation data includes operating parameters such as voltage, current, power, and frequency at each node of the power grid, as well as power grid supply and demand balance data, used to reflect system operating constraints and supply and demand relationships.

[0063] After obtaining multi-source heterogeneous operational data, data quality optimization processing is performed on the multi-source heterogeneous operational data to eliminate noise, missing values, and anomalies in the original data. Data quality optimization processing includes outlier removal, missing value completion, and data denoising. In this embodiment, outlier removal is performed by identifying and removing data points that deviate from the normal fluctuation range through statistical distribution characteristics. Missing value completion combines time series continuity to interpolate and reconstruct missing data and compensates for it through sequence prediction. Data denoising suppresses high-frequency noise components through multi-scale signal decomposition and reconstruction, thereby ensuring the stability and usability of the data.

[0064] After completing the data quality optimization process, further correlation analysis is performed on the multi-source heterogeneous operating data. By calculating the correlation strength between each data variable, data items with low contribution to changes in electricity demand or with redundant expressions are identified and removed to reduce data dimensionality and improve subsequent calculation efficiency. After the redundant data is removed, a set of electricity demand analysis data with consistent structure and effective information is obtained to support the subsequent extraction of factors affecting demand uncertainty, scenario construction, and dynamic simulation analysis.

[0065] S102. Extract and classify candidate uncertainty influencing factors based on the electricity demand analysis data set; the candidate uncertainty influencing factors include climate uncertainty factors, energy uncertainty factors, and load uncertainty factors;

[0066] In this embodiment, the multidimensional operational data that has been cleaned, completed, denoised, and redundantly removed is further transformed into a set of factors that can characterize the sources of power demand fluctuations, thereby providing basic support for subsequent quantitative analysis and scenario construction. During processing, the disturbance sources that are significantly related to changes in power demand are first identified based on the time correlation characteristics, fluctuation response characteristics, and change synchronization characteristics in the power demand analysis data set, and the identified disturbance sources are extracted as candidate uncertainty influencing factors. The extraction here is not just a simple enumeration of the original fields, but rather a combination of the fluctuation range, abrupt change location, duration, and response lag relationship in the process of power demand change, merging and mapping data from different sources so that the candidate factors can accurately reflect the main driving sources that cause unstable changes in power demand.

[0067] After extracting candidate uncertainty factors, these factors are classified into three categories: climate uncertainty factors, energy uncertainty factors, and load uncertainty factors. Climate uncertainty factors characterize the disturbances caused by changes in the external meteorological environment to electricity demand and energy supply. Specifically, they can be extracted from temperature sequences, precipitation sequences, wind speed sequences, sunshine sequences, and extreme weather records, and further refined into factors such as extreme high temperatures, extreme cold waves, heavy rain, gusts, and cloud cover.

[0068] The extreme high temperatures and extreme cold waves mainly reflect the sudden increase in cooling or heating loads caused by significant temperature deviations from the normal range; heavy rain mainly reflects the disturbances brought about by short-term heavy precipitation processes to local load activities, power grid operation status, and user energy consumption behavior; gusts mainly reflect the impact of rapid changes in short-period wind speed on the stability of clean energy output and the energy environment; cloud cover mainly reflects the fluctuations caused by sudden changes in sunlight conditions on photovoltaic output and related load response relationships; through the above refinement, climate disturbances can be separated from coarse-grained environmental variables to form candidate factors that can be expressed for subsequent quantification.

[0069] The extraction of energy uncertainty factors mainly revolves around the energy supply status and electricity price change characteristics. Data related to power output status, energy fluctuation processes, and price signal changes in the electricity demand analysis dataset are correlated, identified, and categorized to form candidate sources of energy-side disturbances. These energy uncertainty factors specifically include wind power output fluctuations, photovoltaic power output fluctuations, conventional power source failures, and electricity price fluctuations. Specifically, wind power output fluctuations and photovoltaic power output fluctuations reflect the unstable output characteristics of clean energy under meteorological disturbances and changes in resource conditions; conventional power source failures reflect supply-side uncertainties caused by equipment malfunctions, shutdowns, or restricted operation of traditional energy supply units; and electricity price fluctuations reflect the impact of price mechanism changes on user-side electricity consumption behavior, load shifting behavior, and energy structure adjustments. Although some of these factors originate from the supply side, they ultimately affect the electricity demand change process through supply-demand balance, price transmission, and dispatch response relationships; therefore, they need to be extracted separately and managed as an independent category.

[0070] The extraction of load uncertainty factors mainly focuses on changes in user-side load composition, load behavior fluctuations, and differences in load regulation response. It identifies relevant information from the electricity demand analysis dataset that can lead to randomness, abrupt changes, and adjustable uncertainty in demand, and categorizes these into candidate load factors. These load uncertainty factors include the randomness of emerging loads, load abrupt changes, and uncertainties in load regulation behavior. Specifically, the randomness of emerging loads mainly corresponds to the unstable changes in the access scale, start-up and shutdown times, and durations of new energy vehicle charging loads, data center loads, and other new electricity consumption units. Load abrupt changes mainly correspond to rapid increases or decreases in demand within a short period. Uncertainties in load regulation behavior mainly correspond to differences in response and execution deviations among different load entities in demand response, peak-shifting, and flexible regulation. This factor extraction effectively distinguishes the structural evolution characteristics and behavioral fluctuation characteristics within the demand side, avoiding the misrepresentation of specific disturbance sources by total load changes in subsequent analysis, thereby improving the accuracy of factor identification.

[0071] After classifying climate uncertainty factors, energy uncertainty factors, and load uncertainty factors, it is also necessary to establish a correspondence between each candidate factor and changes in electricity demand, so that each candidate factor is associated with its source data, range of influence, and change characteristics. This will not only form a set of candidate uncertainty factors with clear hierarchy and well-defined boundaries, but also ensure that appropriate quantification methods and scenario mapping methods can be directly applied to different categories of factors when performing classification and quantification processing, thereby supporting subsequent uncertainty quantification, scenario construction, and dynamic simulation analysis.

[0072] S103. Based on the uncertainty quantification model, perform quantification processing on the classified candidate uncertainty influencing factors, and construct a multi-scenario system based on the quantification results; the quantification results include normal distribution characteristics, abnormal deviation characteristics, and correlation structure characteristics; the multi-scenario system includes a baseline scenario, an extreme scenario, and a coupled scenario;

[0073] Furthermore, the quantification process performed on the classified candidate uncertainty influencing factors based on the uncertainty quantification model specifically includes:

[0074] Based on the analytic hierarchy process combined with the entropy weight method, and by integrating subjective and objective weights, a comprehensive weight is constructed and weights are assigned to the candidate uncertainty influencing factors after classification.

[0075] An uncertainty quantification model is constructed for each different type of candidate uncertainty influencing factor; wherein, the types of the candidate uncertainty influencing factors include random factors, fuzzy factors, and nonlinear correlation factors.

[0076] Specifically, based on the uncertainty quantification model, numerical mapping and influence intensity quantification are performed on the classified candidate uncertainty influencing factors, transforming each candidate uncertainty influencing factor into a comparable quantitative representation, and outputting the quantification results under a unified scale;

[0077] In this embodiment, regarding the construction and application of the uncertainty quantification model, the analytic hierarchy process combined with the entropy weight method is first used to perform weight assignment on the classified candidate uncertainty influencing factors. By introducing a collaborative fusion mechanism of subjective weights and objective weights, a quantitative weight system for the degree of influence of each candidate uncertainty influencing factor on the uncertainty of electricity demand is established.

[0078] Subjective weights are determined based on expert scoring results and are used to reflect the importance of different candidate uncertainty factors in engineering experience and actual operation cognition. Objective weights are determined based on the statistical correlation between each candidate uncertainty factor and changes in electricity demand in the electricity demand analysis dataset and are used to reflect the actual impact intensity under data-driven conditions. By combining and integrating subjective and objective weights, a comprehensive weight is obtained, thereby avoiding the bias problem caused by a single weight assignment method and making the weight allocation both empirically reasonable and data objective.

[0079] Furthermore, to avoid the problems of static fusion, insufficient response to extreme samples, and weak comparability of different factors in the traditional analytic hierarchy process and entropy weight method in the scenario of uncertain electricity demand, this embodiment adopts a scenario-adaptive subjective and objective weight fusion strategy.

[0080] Specifically, firstly, based on the category, data reliability, fluctuation intensity, response lag time, and scenario type of the candidate uncertainty influencing factors, a factor state vector is constructed. ,in Indicates the first The completeness and reliability of the data for each factor. Indicates the degree of fluctuation and variation. Indicates the first The degree of lag in the response of each factor to changes in electricity demand. This indicates the intensity of the disturbance in the current scenario.

[0081] In the objective weight calculation process, the original entropy weight result is not directly used. Instead, the information entropy dispersion, fluctuation contribution, and lagged correlation contribution of the factors are jointly corrected to obtain the corrected objective weight. In the subjective and objective weight fusion process, a fixed fusion ratio is not used. Instead, fusion is based on the scene-adaptive fusion coefficient, and the fusion expression is as follows:

[0082] ;

[0083] in: These are the weighting coefficients after fusion; These are the subjective weights obtained using the analytic hierarchy process. , All are adjustments to objective weights; , All are scene-adaptive fusion coefficients, used to characterize the first... , The factor in the first The degree of dependence on the weight of subjective experience in each scenario;

[0084] When a factor has few historical samples, a high missing rate, or is in an extreme scenario The weights are increased accordingly to enhance the binding effect of expert experience and operational constraints on the weighting results; when the data sample for a certain factor is sufficient, the statistical distribution is stable, and it has a significant lagged correlation with changes in electricity demand, The weights are reduced accordingly to enhance the driving effect of objective data on the weighting results. Through this scenario-adaptive fusion rule, climate, energy and load uncertainties can obtain differentiated weights under different scenarios, avoiding the problem of insufficient identification of extreme disturbances and coupled disturbances in traditional fixed-weight fusion methods.

[0085] After assigning weights, differentiated uncertainty quantification models are constructed for different types of candidate uncertainty influencing factors. These factors are categorized into random factors, fuzzy factors, and nonlinear correlation factors based on their attributes, enabling targeted modeling. For random factors, their historical fluctuation sequences and statistical distribution characteristics in the electricity demand analysis dataset are analyzed, and a probability distribution model is used to characterize their uncertainty. By fitting their probability distribution form, the fluctuation range and probability of occurrence are described, thus achieving a quantitative expression of random fluctuation characteristics. For fuzzy factors, since they are difficult to characterize directly with precise numerical values, an uncertainty quantification model is constructed based on the fuzzy comprehensive evaluation method. By constructing a set of evaluation indicators and membership functions, the fuzzy description is transformed into calculable membership values, thereby achieving a quantitative expression of the degree of uncertainty. For nonlinear correlation factors, based on their complex coupling relationship with electricity demand, a nonlinear correlation model is constructed using machine learning algorithms. By training on the electricity demand analysis dataset, the model learns the mapping relationship between candidate uncertainty influencing factors and electricity demand fluctuations, thus outputting the impact of each factor on changes in electricity demand, achieving uncertainty quantification under nonlinear relationships.

[0086] Furthermore, to enhance the adaptability of the aforementioned uncertainty quantification model to electricity demand business scenarios, this solution does not simply adopt probability distribution models, fuzzy comprehensive evaluation methods, and machine learning algorithms in parallel. Instead, it first determines the model adaptation label based on the data completeness, random volatility, boundary fuzziness, and nonlinear coupling strength of the factors, and selects or combines the corresponding quantification models accordingly. For random factors, a mixed probability distribution model with quantile constraints is adopted. Kernel density estimation or Gaussian mixture distribution fitting is used for the normal interval, and tail enhancement parameters are used to correct the abnormal tail interval, so as to improve the ability to characterize low-frequency, high-impact events such as extreme high temperatures, cold waves, sudden drops in wind power, and sudden drops in photovoltaic power.

[0087] For fuzzy factors, a dynamic membership function is used instead of a fixed membership function for quantification. The membership boundary is determined by historical quantiles, grid operation safety boundary, and demand response execution deviation. This allows factors that are difficult to quantify precisely, such as the uncertainty of load regulation behavior and the fuzziness of electricity price response, to be dynamically updated according to the actual operating status. For nonlinear correlation factors, a nonlinear mapping model is constructed that includes time lag terms, interaction terms, and scenario identification terms. This enables the model to identify the chain transmission relationship between "meteorological disturbance - change in renewable energy output - price or dispatch response - change in load demand".

[0088] After outputting from various model types, the data is uniformly converted into a quantified feature vector containing the normal fluctuation range, abnormal deviation magnitude, tail occurrence probability, response lag, and influence intensity. This vector is then fused with a comprehensive weight to form a unified uncertainty quantification result. Thus, stochastic factors are primarily used to determine the probability distribution and fluctuation range, fuzzy factors are primarily used to determine the uncertainty level and membership strength, and nonlinear correlation factors are primarily used to determine the coupling influence path and transmission strength. These three factors are combined and applied through a unified feature vector, thereby improving the completeness and interpretability of uncertainty quantification.

[0089] After the above-mentioned different types of uncertainty quantification models are constructed, the quantification results of each candidate uncertainty influencing factor are linked and integrated with the corresponding comprehensive weights to form a set of quantification results in a unified expression form. This enables uncertainty factors from different sources and with different attributes to be compared and integrated on a unified scale, thereby providing consistent parameterized input for the subsequent construction of coupled uncertainty scenario system and dynamic simulation of power demand.

[0090] Furthermore, the construction of a multi-scenario system based on the quantification results specifically includes:

[0091] A baseline scenario is constructed based on the normal distribution characteristics in the quantification results; the baseline scenario is a scenario in which the uncertain influencing factors are in the normal fluctuation range and the target new power system maintains a stable power operation state;

[0092] Extreme scenarios are constructed based on the abnormal deviation characteristics in the quantification results; the extreme scenarios are those in which a single uncertain influencing factor is in an abnormal fluctuation range, causing the target new power system to be unable to maintain a stable operating state.

[0093] Coupled scenarios are constructed based on the correlation structure features in the quantification results; the coupled scenario is a synergistic scenario formed by superimposing or combining multiple extreme scenarios according to preset correlation constraints.

[0094] Scenario parameters were set for the baseline scenario, extreme scenario, and coupled scenario, and a multi-scenario system was constructed.

[0095] In this embodiment, regarding the construction of the multi-scenario system, a baseline scenario is first constructed based on the normal distribution characteristics in the quantification results. Specifically, by extracting the normal fluctuation range and statistical stability characteristics of each uncertainty influencing factor in historical data, normal constraints are set for each factor to ensure that each uncertainty influencing factor is within the normal operating range. This forms a scenario in which the uncertainty influencing factors are within the normal fluctuation range and the target new power system maintains a stable power operation state, serving as a reference benchmark for subsequent uncertainty analysis. This baseline scenario is used to reflect the basic evolution trend of power demand under conditions without significant disturbances and to provide a comparison baseline for extreme scenarios and coupled scenarios.

[0096] After the baseline scenario is constructed, an extreme scenario is constructed based on the abnormal deviation characteristics in the quantification results. By identifying the abnormal deviation ranges of each uncertain influencing factor relative to the normal distribution, and performing extreme reinforcement settings on a single uncertain influencing factor to make it enter the abnormal fluctuation range, while keeping other factors in the normal range, an extreme disturbance state dominated by a single factor is formed. This results in a scenario where the target new power system cannot maintain a stable operating state due to the abnormal fluctuation range of a single uncertain influencing factor. This extreme scenario is used to characterize the amplification effect of a single disturbance source on power demand and its impact on system stability, thereby achieving independent analysis of the extreme impact path of a single factor.

[0097] Based on extreme scenarios, coupled scenarios are constructed based on the correlation structure characteristics in the quantification results. By analyzing the correlation and coupling strength between different uncertainty influencing factors, multiple extreme scenarios are superimposed or combined under correlation constraints. This allows multiple uncertainty influencing factors in the abnormal fluctuation range to act together according to preset correlation constraints, thereby forming a composite disturbance state of synergistic amplification or mutual superposition. The resulting coupled scenario is used to characterize the complex fluctuation characteristics and nonlinear superposition effects of power demand under the combined action of multiple uncertain factors.

[0098] The preset correlation constraints include time synchronization constraints, spatial region constraints, physical transmission constraints, statistical dependency constraints, and operational safety constraints; the time synchronization constraints are used to limit the occurrence time or lag time of multiple extreme factors to meet the following conditions. ,in , The first , The actual moment when each extreme factor occurs; For the first , Theoretically inherent time lag between extreme factors; The upper limit of allowable error is defined by the following constraints: spatial region constraints are used to limit the factors involved in coupling to be located in the same power supply area, the same meteorological influence area, or adjacent nodes with electrical connections; physical conduction constraints are used to limit the direction of interaction between factors to be consistent with the power system operation mechanism, such as the conduction direction between high temperature load increase and photovoltaic output decrease, and wind power fluctuation and reserve demand increase; statistical dependency constraints are used to limit the correlation coefficient, tail correlation coefficient, or joint occurrence probability between factors to reach a preset threshold; and operational safety constraints are used to limit the coupling scenario generation results to meet the basic boundaries of power grid flow, safety margin, and supply and demand balance.

[0099] In the specific construction process, the correlation constraint matrix is ​​first constructed based on the correlation coefficient, lagged correlation coefficient, spatial distance weight, physical causal identifier, and joint tail probability of each candidate uncertainty influencing factor. and threshold ;when When, then determine the first The factor and the first Several factors can enter the same coupled scenario; when If the operational safety constraints are not met, the corresponding factors will not be combined into a coupled scenario.

[0100] The expression for the disturbance strength of the coupled scenario is as follows:

[0101] ;

[0102] in: , The first , Whether a factor is selected as a state variable in a coupled scenario; , The first , Standardized outlier deviations of each factor; The correlation strength.

[0103] Through the aforementioned constraint matrix, multiple extreme scenarios are not simply superimposed, but combined under the condition that constraints in time, space, physical transmission, and statistical dependence are satisfied. The combined coupled scenario parameters are further input into the dynamic scenario simulation model, enabling the model to express the synergistic amplification, mutual cancellation, or chain transmission effects of multiple factors within the constraint framework, thereby achieving the computability and traceability of the combined effects of multiple factors.

[0104] After constructing the baseline scenario, extreme scenario, and coupled scenario, scenario parameter setting processing is further performed on each scenario. By combining the fluctuation range, occurrence probability, and influence intensity of each uncertainty influencing factor in the quantification results, corresponding parameter value ranges and constraints are set for each scenario, so that each scenario has a clear computable boundary. Scenario parameter setting includes limiting the value range, duration, and influence intensity of each uncertainty influencing factor to ensure the rationality and simulability of scenario construction. After completing the parameter setting, the baseline scenario, extreme scenario, and coupled scenario are uniformly organized and structurally integrated to form a multi-scenario system covering normal operation, single extreme disturbance, and multi-source coupling effects, thereby achieving a comprehensive characterization of the uncertainty characteristics of power demand.

[0105] Furthermore, the range of scenario parameters is determined jointly based on the historical quantiles, mean standard deviation, outlier thresholds, operational safety boundaries, and impact intensity of each uncertainty factor. For the baseline scenario, the factor values ​​are limited to the normal distribution range, for example, by using... quantile intervals or using The normal fluctuation range; for extreme scenarios, limit individual factors to the abnormal deviation range, for example, by using a range higher than the normal fluctuation range. or below The tail section, or using more than The abnormal interval, in which The factors are determined based on their historical extreme volatility and the power grid safety margin; for coupled scenarios, the factors satisfy the correlation constraint matrix. Based on this, corresponding abnormal intervals are adopted for each factor, and the value intervals are corrected according to the joint occurrence probability and coupling strength.

[0106] The scenario parameters include factor values. Duration Intensity of action Probability of occurrence and constraint identifier Among them, the intensity of action The expression is as follows:

[0107] ;

[0108] in: For the first The central values ​​of each factor under the baseline scenario; and These represent the upper and lower bounds of the normal fluctuation range; To prevent the minimum value where the denominator is zero; duration The probability of occurrence is determined by a combination of factor autocorrelation decay time, duration of historical extreme events, and real-time detection window. Calculated based on the probability distribution model or joint probability distribution model.

[0109] The parameter adaptation rules for different scenarios are as follows: the baseline scenario focuses on maintaining a high probability of occurrence and a low intensity of impact, the extreme scenario focuses on highlighting the tail deviation and continuous impact of a single factor, and the coupled scenario focuses on satisfying the common constraints of joint probability, correlation strength and operational safety boundary. Through this rule, the fluctuation range, probability of occurrence and intensity of impact of a factor can be mapped to the value range, duration and intensity parameters, respectively, so that the parameters of each scenario have clear delineation basis and calculable boundaries.

[0110] S104. Input the quantification results into the dynamic scenario simulation model to obtain the dynamic simulation results of the power demand of the target new power system under multiple scenario systems.

[0111] Furthermore, the step of inputting the quantification results into the dynamic scenario simulation model to obtain the dynamic simulation results of the power demand of the target new power system under multiple scenario frameworks specifically includes:

[0112] A dynamic scenario simulation model is constructed based on an improved long short-term memory network and combined with quantification results;

[0113] The quantitative results are input into the dynamic scenario simulation model, and the output is the dynamic simulation results of the power demand of the target new power system under multiple scenarios. The dynamic simulation results of power demand include the time-series variation curve of power demand, fluctuation range, peak load and valley load.

[0114] Specifically, taking the quantitative results as input, the dynamic scenario simulation model is used to perform time-series evolution simulation processing on each scenario under the multi-scenario system. The quantitative results of the uncertain influencing factors and the scenario parameters are mapped to the power demand change process, and the dynamic simulation results of the power demand of the target new power system under different scenario conditions are output.

[0115] In this embodiment, regarding the construction and application of the dynamic scenario simulation model, an improved long short-term memory network is first used, and a dynamic scenario simulation model is constructed by combining the quantification results. By fusing and encoding the quantification results of uncertain influencing factors with the scenario parameters in the multi-scenario system, a unified time-series input expression is formed, enabling the model to simultaneously characterize the historical electricity demand change characteristics and the process of uncertain disturbances. An attention mechanism is introduced into the model structure to adaptively allocate the weights of different uncertain influencing factors, so that high-weight factors receive higher attention during the model calculation process, thereby improving the response capability to key disturbance factors.

[0116] Furthermore, the improved Long Short-Term Memory (LSTM) network adopts a scene-aware multi-branch temporal fusion structure, rather than simply superimposing a common attention mechanism on the traditional LSM network. This structure includes a feature selection layer, a multi-source temporal encoding layer, a scene embedding layer, a dual attention layer, and a residual correction layer. The feature selection layer selects input features that are highly relevant to the current scenario from multi-source heterogeneous operating data based on correlation, maximum information coefficient, Granger lag test, or model perturbation contribution, so as to reduce the interference of redundant features on model training. The multi-source temporal encoding layer encodes the historical electricity demand sequence, climate perturbation sequence, energy output sequence, load structure sequence, and grid operation constraint sequence respectively, and then fuses the encoding results of each branch into a unified temporal expression.

[0117] The scenario embedding layer encodes the scenario type, intensity, duration, and associated constraint matrix of the baseline scenario, extreme scenario, and coupled scenario into scenario vectors, and introduces these scenario vectors into the calculation process of the input gate, forget gate, and output gate, so that the proportion of historical information retention, the proportion of current disturbance input, and the proportion of output response under different scenarios can dynamically change with the scenario intensity.

[0118] The dual attention layer includes factor attention and time attention. Factor attention is used to identify the importance of different uncertainty factors in the current scenario, while time attention is used to identify the impact of key disturbance periods on changes in electricity demand, thereby avoiding the problem of ignoring the duration of disturbances by only assigning weights from the factor dimension.

[0119] The residual correction layer constructs an error sequence based on the deviation between the actual and simulated electricity demand values ​​over a recent period, and feeds this error sequence back to the model output to correct the prediction results for the next time period or the next scenario cycle. Simultaneously, the model output is not limited to a single-point prediction value, but rather outputs interval-based simulation results including upper quantile demand, lower quantile demand, and expected demand, directly supporting subsequent uncertainty assessment and risk level classification. Through the combination of feature selection, multi-source branch coding, scenario embedding, dual attention, and residual correction, the temporal transmission and coupling amplification effects of multi-source uncertainties under different scenarios can be more accurately characterized.

[0120] The expression for the improved Long Short-Term Memory network is as follows:

[0121] ;

[0122] in: The output of the forget gate controls the degree to which historical information is retained, and its value ranges from 0 to 1. This is the Sigmoid function, used to compress the result to the range of 0 to 1; This is the weight matrix corresponding to the forget gate, and its values ​​are obtained through training; This is the concatenated vector of the hidden state from the previous time step and the current input features. This refers to the state information from the previous moment. The relevant characteristics and uncertainties of the current electricity demand input are quantified. This is a bias term used to adjust the overall offset of the calculation results; this expression is used to determine how much historical information to retain based on the current input and historical state.

[0123] ;

[0124] in: The output of the input gate is used to control the proportion of the current input information entering the memory unit; , These are the weight matrix and bias term of the input gate, respectively, and their meanings and values ​​are similar to those described above.

[0125] ;

[0126] in: These are candidate memory states, used to generate new candidate information; This is the hyperbolic tangent function, used to limit the output to between -1 and 1; , These are the corresponding weight matrix and bias term, respectively;

[0127] ;

[0128] in: The current memory state; This represents the memory state of the previous moment; the expression achieves the fusion and update of historical and current information through the synergistic effect of the forget gate and the input gate.

[0129] ;

[0130] in: This is an output gate used to control the proportion of the memory state output. , These are the weight matrix and bias term of the output gate, respectively;

[0131] ;

[0132] in: The output represents the hidden state at the current moment, which is used as an intermediate representation for electricity demand forecasting.

[0133] After introducing the attention mechanism, the weighted expression for each uncertainty factor is as follows:

[0134] ;

[0135] in: For the first The attention weights of the uncertain influencing factors range from 0 to 1 and the sum of all weights is 1; , The first , The importance score of each uncertainty factor is calculated by the model based on its degree of impact on electricity demand; This represents the total number of uncertain influencing factors.

[0136] ;

[0137] in: The input features are after attention weighting; For the first One uncertain influencing factor at time... The quantized input value; this expression, by weighting and fusing different factors, makes the key factors have a more significant impact on the model output.

[0138] After the model is built, dynamic simulation of power demand is performed for the baseline scenario, extreme scenario and coupled scenario respectively. By inputting the parameter constraints of the corresponding scenario into the dynamic scenario simulation model, the time-series variation curve, fluctuation range, peak load and valley load of power demand under each scenario are obtained, thus forming the dynamic simulation results of power demand.

[0139] During the dynamic adaptive adjustment process, the system continuously acquires real-time updated data on uncertain influencing factors and maps them to the input of the dynamic scenario simulation model. The system then dynamically corrects the scenario parameters and model weights. When abnormal fluctuations are detected to exceed a preset threshold, the system automatically switches or triggers the simulation process of the corresponding scenario, thereby achieving real-time updates of the simulation results and improving the model's ability to respond to sudden disturbances.

[0140] In the simulation accuracy verification process, the root mean square error and mean absolute error are used to evaluate the simulation results, and their expressions are as follows:

[0141] ;

[0142] in: Root mean square error is used to measure the overall deviation between the predicted value and the actual value. The number of samples; For a moment The actual electricity demand value; This represents the electricity demand value predicted by the model; the expression uses the average back square root of the squared differences to highlight the impact of larger errors.

[0143] ;

[0144] in: The mean absolute error is used to measure the average deviation between the predicted and actual values; the meanings of the parameters are consistent with those above; this expression reflects the overall error level by averaging the absolute values.

[0145] Through the and A threshold is determined. If the model accuracy meets the preset requirements, the dynamic simulation result of power demand is output. Otherwise, the parameters of the dynamic scenario simulation model are iteratively optimized until the accuracy requirements are met, thereby ensuring that the dynamic scenario simulation model has high prediction accuracy and stability under multiple scenario conditions.

[0146] S105. Based on the dynamic simulation results of electricity demand, candidate uncertainty influencing factors are screened through uncertainty assessment and sensitivity correlation analysis to obtain target uncertainty influencing factors that meet the preset assessment conditions.

[0147] Furthermore, based on the dynamic simulation results of electricity demand, candidate uncertainty influencing factors are screened through uncertainty assessment and sensitivity correlation analysis to obtain target uncertainty influencing factors that meet preset assessment conditions, specifically including:

[0148] Calculate the fluctuation range of electricity demand under various scenarios; the fluctuation range includes the maximum fluctuation range and the average fluctuation range.

[0149] Based on the dynamic simulation results of electricity demand, a probability distribution curve of electricity demand is constructed;

[0150] Based on the fluctuation amplitude value and probability distribution curve, the candidate uncertainty influencing factors of each scenario are classified into risk levels; the risk levels include high risk level, medium risk level and low risk level.

[0151] Based on the risk level, sensitivity quantification is performed on the candidate uncertainty influencing factors to obtain the sensitivity coefficient corresponding to each candidate uncertainty influencing factor.

[0152] Candidate uncertainty influencing factors are ranked and screened based on sensitivity coefficients, and those that meet the preset sensitivity threshold conditions are selected as target uncertainty influencing factors.

[0153] Specifically, based on the dynamic simulation results of electricity demand, the contribution assessment and importance screening of candidate uncertainty influencing factors are performed through uncertainty assessment and sensitivity correlation analysis. From this, key factors that have a significant impact on electricity demand fluctuations are identified, and target uncertainty influencing factors that meet the preset assessment conditions are output.

[0154] In this embodiment, focusing on uncertainty assessment and sensitivity correlation analysis, the uncertainty of electricity demand under each scenario is first quantitatively assessed based on the dynamic simulation results of electricity demand. By constructing a joint assessment mechanism of "fluctuation amplitude - probability distribution - risk level", the scope of uncertainty impact and probability of occurrence are synergistically characterized.

[0155] Furthermore, to avoid insufficient evaluation dimensions due to relying solely on maximum fluctuation amplitude, average fluctuation amplitude, and probability distribution curves, this embodiment adds evaluation dimensions such as tail risk probability, over-limit duration, ramp change rate, prediction interval width, and risk recovery time to the joint evaluation mechanism. Specifically, tail risk probability characterizes the probability of electricity demand falling into a high-risk range; over-limit duration characterizes the continuous duration of demand exceeding the safety boundary; ramp change rate characterizes the impact of rapid load increases or decreases on dispatch response; prediction interval width characterizes the range of uncertainty in the model output; and risk recovery time characterizes the time required for the system to recover from a high-risk state to a normal state.

[0156] Risk level classification employs a combination of fixed and dynamic thresholds. The fixed thresholds are determined by historical operational statistics and power grid safety standards, while the dynamic thresholds are adjusted based on current reserve capacity, load factor, adjustable resource scale, and confidence level of renewable energy output. When the comprehensive risk assessment value exceeds the high-risk threshold, the probability of tail risks exceeds a preset probability threshold, or the duration of exceeding limits exceeds a preset duration threshold, the corresponding scenario or factor is classified as high-risk. This approach ensures that risk levels reflect not only the magnitude of fluctuations but also the probability of fluctuations, their sustained impact, and the system's carrying capacity.

[0157] In sensitivity analysis, in addition to calculating the local sensitivity coefficient of a single factor, global sensitivity, interaction sensitivity, and risk-weighted sensitivity are also calculated; the expression for the risk-weighted sensitivity is as follows:

[0158] ;

[0159] in: It is a local sensitivity; For global sensitivity; For the first , The sensitivity of interactions between factors; The correlation strength in the correlation constraint matrix; For the first The risk assessment value corresponding to each factor , , , All are weighting coefficients; It is used to screen factors affecting target uncertainty, and the screening results are fed back to scenario parameters and model attention weights to achieve a closed-loop linkage between uncertainty assessment and sensitivity analysis.

[0160] For the assessment of fluctuation amplitude, the maximum and average fluctuation amplitude values ​​of electricity demand under each scenario are calculated;

[0161] The expression for the maximum fluctuation amplitude value is as follows:

[0162] ;

[0163] in: This is the maximum fluctuation range value, used to reflect the maximum fluctuation range of electricity demand within a scenario period; It is the maximum value in the electricity demand sequence; This is the minimum value in the electricity demand sequence; The average electricity demand value is obtained by averaging the electricity demand values ​​over all times; this expression characterizes the overall fluctuation intensity by the ratio of the difference between the maximum and minimum values ​​to the average level.

[0164] The expression for the average fluctuation amplitude value is as follows:

[0165] ;

[0166] in: This is the average fluctuation range value, used to reflect the overall degree of fluctuation in electricity demand relative to the average level; This represents the total number of sampling points in the time series; the expression obtains the overall fluctuation level by averaging the absolute values ​​of the relative deviations at each time point.

[0167] In probability distribution assessment, the probability distribution curve of electricity demand is constructed based on the dynamic simulation results of electricity demand. The expression of its probability density function is as follows:

[0168] ;

[0169] in: The value of electricity demand is... The probability density at that time; This is the kernel function, used to describe the contribution of a sample point to the probability density; The bandwidth parameter controls the smoothness, and its value is determined empirically or through cross-validation. This expression forms a continuous probability distribution curve by weighting and superimposing sample points, thereby characterizing the probability of electricity demand occurring in different intervals.

[0170] Based on the fluctuation amplitude and probability distribution curves, the risk levels of candidate uncertainty influencing factors for each scenario are classified, and the risk level function can be expressed as follows:

[0171] ;

[0172] in: This is a comprehensive risk assessment value; , , The weighting coefficients for different evaluation indicators are set according to actual needs; This is a high-risk electricity demand range; the expression obtains a unified risk assessment result by integrating fluctuation amplitude and probability distribution information, and classifies the risk into high-risk, medium-risk and low-risk levels accordingly.

[0173] After completing the risk level classification, sensitivity quantification is performed on the candidate uncertainty influencing factors based on the risk level. By calculating the response strength of each factor to changes in electricity demand, the sensitivity coefficient is obtained, and its expression is as follows:

[0174] ;

[0175] in: For the first The sensitivity coefficients of each candidate uncertainty influencing factor are used to characterize the degree of impact of changes in influencing factors on changes in electricity demand. The rate of change of electricity demand with respect to influencing factors can be obtained through model perturbation or regression analysis; For the first The expression takes values ​​for each candidate uncertainty influencing factor; it is normalized to make different factors comparable.

[0176] After obtaining the sensitivity coefficients corresponding to each candidate uncertainty influencing factor, the candidate uncertainty influencing factors are sorted and screened, and the screening expression is as follows:

[0177] ;

[0178] in: The set of factors influencing the uncertainty of the objective; The preset sensitivity threshold is set according to the analysis requirements; This represents the sensitivity coefficient of the corresponding factor; this expression filters out key factors that significantly affect the uncertainty of electricity demand by setting a threshold.

[0179] Through the above processing, scenario simulation and uncertainty assessment are linked, enabling the uncertainty of electricity demand under different scenarios to be quantitatively expressed, and key influencing factors to be screened from candidate uncertainty influencing factors, thereby providing targeted decision-making basis for power grid planning, dispatch optimization and risk prevention and control.

[0180] S106. Based on the dynamic simulation results of electricity demand and the factors affecting the uncertainty of the target, construct a multi-scenario electricity demand uncertainty assessment report.

[0181] Furthermore, after constructing a multi-scenario electricity demand uncertainty assessment report based on the dynamic simulation results of electricity demand and the influencing factors of target uncertainty, it also includes:

[0182] The report on uncertainty assessment of multi-scenario electricity demand is compared and analyzed with actual electricity demand operation data, and the deviation analysis results are output.

[0183] Based on the results of the deviation analysis, the model parameters of the uncertainty quantification model, the scenario parameters of the multi-scenario system, and the simulation parameters of the dynamic scenario simulation model are subjected to collaborative optimization.

[0184] In this embodiment, based on the dynamic simulation results of electricity demand and the factors affecting the uncertainty of the target, the uncertainty characteristics of electricity demand under multiple scenarios are comprehensively integrated and structurally expressed to construct a multi-scenario electricity demand uncertainty assessment report.

[0185] After constructing a multi-scenario power demand uncertainty assessment report, the report is further compared and analyzed with actual power demand operation data. By comparing the dynamic simulation results of power demand under different scenarios with the actual power demand changes in the corresponding time periods, the deviation between the simulation results and actual operation is calculated. Combined with the performance of the target uncertainty influencing factors in actual operation, the sources of deviation are analyzed, and the deviation analysis results are output. The deviation analysis results are used to characterize the distribution characteristics of simulation error and assessment error, and identify the key reasons for the deviation, including scenario parameter setting deviation, uncertainty quantification model parameter deviation, and dynamic scenario simulation model parameter deviation.

[0186] After obtaining the deviation analysis results, the model parameters of the uncertainty quantification model, the scenario parameters of the multi-scenario system, and the simulation parameters of the dynamic scenario simulation model are jointly optimized based on the deviation analysis results. By adjusting the weight allocation of the uncertainty influencing factors, correcting the parameter value range in each scenario, and retraining or updating the internal parameters of the dynamic scenario simulation model, the model output results are made to gradually approach the actual operating state, thereby improving the accuracy and consistency of the dynamic simulation results of power demand and the uncertainty assessment results.

[0187] While performing parameter optimization, the multi-scenario system is updated in light of changes in the operating environment of the target new power system. By introducing new uncertain influencing factors or adjusting the relationships between existing factors, the construction methods of the baseline scenario, extreme scenario, and coupled scenario are expanded or modified, so that the multi-scenario system can cover new operating conditions and disturbance modes, thereby ensuring the integrity and adaptability of the scenario system.

[0188] After completing the multi-scenario system update, based on continuously acquired historical and real-time operational data, the uncertainty quantification model, multi-scenario system, and dynamic scenario simulation model are iteratively updated. By periodically repeating the deviation analysis and parameter optimization process, a closed-loop operation mechanism consisting of data acquisition, simulation analysis, uncertainty assessment, and feedback optimization is formed, enabling the power demand uncertainty analysis process to continuously and adaptively adjust, thereby continuously improving the analysis accuracy and practical application effect.

[0189] This embodiment adds processing steps such as scenario adaptation and weighting, categorized uncertainty quantification, coupled scenario generation driven by correlation constraints, scenario-aware dynamic simulation, and risk-sensitivity closed-loop feedback to the original multi-source heterogeneous data processing, multi-scenario construction, and dynamic simulation process. This enables various uncertain influencing factors to no longer participate in the analysis only in the form of static indicators, but to be modeled differently according to their fluctuation source, duration, spatial region, transmission direction, response lag, and joint occurrence probability.

[0190] Through the above supplementary processing, this solution can establish a calculable, verifiable, and dynamically updatable technical link between the baseline scenario, the single extreme scenario, and the multi-factor coupled scenario. It can also reverse the weight parameters, scenario parameters, and simulation model parameters based on the actual operational deviations, thereby forming a closed-loop analysis mechanism of "factor identification - uncertainty quantification - scenario generation - dynamic simulation - risk screening - deviation feedback" to improve the accuracy, robustness, and scenario adaptability of the results of power demand uncertainty analysis.

[0191] Example 2:

[0192] See Figure 2 A power demand uncertainty analysis system based on multi-scenario simulation, which is applied to the method described in Example 1, the system comprising:

[0193] Data acquisition module 1 is used to acquire multi-source heterogeneous operation data to characterize the changes in power demand of the target new power system, and preprocess the data to obtain a power demand analysis dataset.

[0194] Uncertainty influencing factor classification module 2 is used to extract and classify candidate uncertainty influencing factors based on the power demand analysis data set; the classified candidate uncertainty influencing factors include climate uncertainty factors, energy uncertainty factors and load uncertainty factors;

[0195] The multi-scenario system construction module 3 is used to perform quantification processing on the classified candidate uncertainty influencing factors based on the uncertainty quantification model, and construct a multi-scenario system based on the quantification results; the quantification results include normal distribution characteristics, abnormal deviation characteristics, and correlation structure characteristics; the multi-scenario system includes a baseline scenario, an extreme scenario, and coupled scenarios;

[0196] Dynamic simulation module 4 is used to input the quantification results into the dynamic scenario simulation model to obtain the dynamic simulation results of the power demand of the target new power system under multiple scenario systems.

[0197] The target uncertainty influencing factor screening module 5 is used to screen candidate uncertainty influencing factors based on the dynamic simulation results of power demand through uncertainty assessment and sensitivity correlation analysis, so as to obtain the target uncertainty influencing factors that meet the preset assessment conditions.

[0198] Uncertainty assessment module 6 is used to construct a multi-scenario power demand uncertainty assessment report based on the results of dynamic simulation of power demand and the factors affecting the uncertainty of the target.

[0199] Furthermore, for the specific implementation steps of the above modules, please refer to the corresponding description in Example 1, which will not be repeated here.

[0200] Example 3:

[0201] See Figure 3 A device for analyzing power demand uncertainty based on multi-scenario simulation, the device including a processor 7 and a memory 8;

[0202] The memory 8 is used to store computer program code 81 and to transmit the computer program code 81 to the processor 7;

[0203] The processor 7 is used to execute the power demand uncertainty analysis method based on multi-scenario simulation as described in Embodiment 1 according to the instructions in the computer program code 81.

[0204] This embodiment also includes a computer-readable storage medium storing computer-executable instructions. When the computer-executable instructions are executed on a computer, the power demand uncertainty analysis method based on multi-scenario simulation described in Embodiment 1 is implemented.

[0205] Generally, computer instructions for implementing the method of the present invention can be carried on any combination of one or more computer-readable storage media. Non-transitory computer-readable storage media can include any computer-readable medium except for the signal itself which is temporarily propagating.

[0206] Computer-readable storage media can be, for example, but not limited to, electrical, magnetic, optical, electromagnetic, infrared, or semiconductor systems, apparatuses, or any combination thereof. More specific examples (a non-exhaustive list) of computer-readable storage media include: electrical connections having one or more wires, portable computer disks, hard disks, random access memory (RAM), read-only memory (ROM), erasable programmable read-only memory (EPROM or flash memory), optical fibers, portable compact disk read-only memory (CD-ROM), optical storage devices, magnetic storage devices, or any suitable combination thereof. In this invention, a computer-readable storage medium can be any tangible medium containing or storing a program that can be used by or in conjunction with an instruction execution system, apparatus, or device.

[0207] Computer program code for performing the operations of this invention can be written in one or more programming languages ​​or a combination thereof. These programming languages ​​include object-oriented programming languages—such as Java, Smalltalk, and C++—as well as conventional procedural programming languages—such as the "C" language or similar programming languages. In particular, Python, suitable for neural network computation, and platform frameworks based on TensorFlow, PyTorch, etc., can be used. The program code can be executed entirely on the user's computer, partially on the user's computer, as a standalone software package, partially on the user's computer and partially on a remote computer, or entirely on a remote computer or server. In cases involving remote computers, the remote computer can be connected to the user's computer via any type of network, including a local area network (LAN) or a wide area network (WAN), or connected to an external computer (e.g., via the Internet using an Internet service provider).

[0208] The aforementioned equipment and non-transitory computer-readable storage media can be found in the detailed description of a method for analyzing power demand uncertainty based on multi-scenario simulation and its beneficial effects, which will not be repeated here.

[0209] Although embodiments of the present invention have been shown and described above, it should be understood that the above embodiments are exemplary and should not be construed as limiting the present invention. Those skilled in the art can make changes, modifications, substitutions and variations to the above embodiments within the scope of the present invention.

Claims

1. A method for analyzing the uncertainty of electricity demand based on multi-scenario simulation, characterized in that, include: Acquire multi-source heterogeneous operational data to characterize changes in electricity demand in the target new power system, and construct a power demand analysis dataset; Candidate uncertainty factors were extracted and classified based on the electricity demand analysis dataset; the classified candidate uncertainty factors include climate uncertainty factors, energy uncertainty factors, and load uncertainty factors. The candidate uncertainty influencing factors are quantified based on the uncertainty quantification model, and a multi-scenario system is constructed based on the quantification results. The quantification results include normal distribution characteristics, abnormal deviation characteristics, and correlation structure characteristics. The multi-scenario system includes a baseline scenario, an extreme scenario, and a coupled scenario. The quantitative results are input into the dynamic scenario simulation model to obtain the dynamic simulation results of the power demand of the target new power system under multiple scenario systems. Based on the dynamic simulation results of electricity demand, candidate uncertainty influencing factors are screened through uncertainty assessment and sensitivity correlation analysis to obtain target uncertainty influencing factors that meet the preset assessment conditions. Based on the dynamic simulation results of electricity demand and the factors affecting the uncertainty of the target, a multi-scenario electricity demand uncertainty assessment report is constructed.

2. The method for analyzing power demand uncertainty based on multi-scenario simulation according to claim 1, characterized in that: The acquisition of multi-source heterogeneous operational data to characterize changes in the power demand of the target new power system, and the construction of a power demand analysis dataset, specifically includes: Acquire electricity demand data and data on uncertain influencing factors as multi-source heterogeneous operation data; the electricity demand data includes historical electricity demand data and real-time electricity demand data; the data on uncertain influencing factors includes climate factor data, energy factor data, load structure data, and power grid operation data; Data quality optimization processing was performed on the multi-source heterogeneous operation data, and correlation analysis was used to remove redundant data in the multi-source heterogeneous operation data to obtain a set of power demand analysis data.

3. The method for analyzing power demand uncertainty based on multi-scenario simulation according to claim 1, characterized in that: The process of quantifying the classified candidate uncertainty influencing factors based on the uncertainty quantification model specifically includes: Based on the analytic hierarchy process combined with the entropy weight method, and by integrating subjective and objective weights, a comprehensive weight is constructed and weights are assigned to the candidate uncertainty influencing factors after classification. An uncertainty quantification model is constructed for each different type of candidate uncertainty influencing factor; wherein, the types of the candidate uncertainty influencing factors include random factors, fuzzy factors, and nonlinear correlation factors.

4. The method for analyzing power demand uncertainty based on multi-scenario simulation according to claim 1, characterized in that: The construction of a multi-scenario system based on quantification results specifically includes: A baseline scenario is constructed based on the normal distribution characteristics in the quantification results; the baseline scenario is a scenario in which the uncertain influencing factors are in the normal fluctuation range and the target new power system maintains a stable power operation state; Extreme scenarios are constructed based on the abnormal deviation characteristics in the quantification results; the extreme scenarios are those in which a single uncertain influencing factor is in an abnormal fluctuation range, causing the target new power system to be unable to maintain a stable operating state. Coupled scenarios are constructed based on the correlation structure features in the quantification results; the coupled scenario is a synergistic scenario formed by superimposing or combining multiple extreme scenarios according to preset correlation constraints. Scenario parameters were set for the baseline scenario, extreme scenario, and coupled scenario, and a multi-scenario system was constructed.

5. The method for analyzing power demand uncertainty based on multi-scenario simulation according to claim 1, characterized in that: The step of inputting the quantification results into the dynamic scenario simulation model to obtain the dynamic simulation results of the power demand of the target new power system under multiple scenario frameworks specifically includes: A dynamic scenario simulation model is constructed based on an improved long short-term memory network and combined with quantification results; The quantitative results are input into the dynamic scenario simulation model, and the output is the dynamic simulation results of the power demand of the target new power system under multiple scenarios. The dynamic simulation results of power demand include the time-series variation curve of power demand, fluctuation range, peak load and valley load.

6. The method for analyzing power demand uncertainty based on multi-scenario simulation according to claim 5, characterized in that: Based on the dynamic simulation results of electricity demand, candidate uncertainty influencing factors are screened through uncertainty assessment and sensitivity correlation analysis to obtain target uncertainty influencing factors that meet preset assessment conditions, specifically including: Calculate the fluctuation range of electricity demand under various scenarios; the fluctuation range includes the maximum fluctuation range and the average fluctuation range. Based on the dynamic simulation results of electricity demand, a probability distribution curve of electricity demand is constructed; Based on the fluctuation amplitude value and probability distribution curve, the candidate uncertainty influencing factors of each scenario are classified into risk levels; the risk levels include high risk level, medium risk level and low risk level. Based on the risk level, sensitivity quantification is performed on the candidate uncertainty influencing factors to obtain the sensitivity coefficient corresponding to each candidate uncertainty influencing factor. Candidate uncertainty influencing factors are ranked and screened based on sensitivity coefficients, and those that meet the preset sensitivity threshold conditions are selected as target uncertainty influencing factors.

7. The method for analyzing power demand uncertainty based on multi-scenario simulation according to claim 1, characterized in that: After constructing a multi-scenario electricity demand uncertainty assessment report based on the dynamic simulation results of electricity demand and the influencing factors of target uncertainty, it also includes: The report on uncertainty assessment of multi-scenario electricity demand is compared and analyzed with actual electricity demand operation data, and the deviation analysis results are output. Based on the results of the deviation analysis, the model parameters of the uncertainty quantification model, the scenario parameters of the multi-scenario system, and the simulation parameters of the dynamic scenario simulation model are subjected to collaborative optimization.

8. A power demand uncertainty analysis system based on multi-scenario simulation, characterized in that, The system is applied to the method according to any one of claims 1-7, the system comprising: The data acquisition module (1) is used to acquire multi-source heterogeneous operation data to characterize the changes in power demand of the target new power system and to construct a power demand analysis data set; The uncertainty influencing factor classification module (2) is used to extract candidate uncertainty influencing factors based on the power demand analysis data set and classify them; the classified candidate uncertainty influencing factors include climate uncertainty factors, energy uncertainty factors and load uncertainty factors; The multi-scenario system construction module (3) is used to perform quantification processing on the classified candidate uncertainty influencing factors based on the uncertainty quantification model, and to construct a multi-scenario system based on the quantification results; the quantification results include normal distribution characteristics, abnormal deviation characteristics and correlation structure characteristics; the multi-scenario system includes a baseline scenario, an extreme scenario and a coupled scenario; The dynamic simulation module (4) is used to input the quantification results into the dynamic scenario simulation model to obtain the dynamic simulation results of the power demand of the target new power system under the multi-scenario system. The target uncertainty influencing factor screening module (5) is used to screen candidate uncertainty influencing factors based on the dynamic simulation results of power demand through uncertainty assessment and sensitivity correlation analysis, so as to obtain the target uncertainty influencing factors that meet the preset assessment conditions. The uncertainty assessment module (6) is used to construct a multi-scenario power demand uncertainty assessment report based on the dynamic simulation results of power demand and the influencing factors of target uncertainty.

9. A device for analyzing power demand uncertainty based on multi-scenario simulation, characterized in that: The device includes a processor (7) and a memory (8); The memory (8) is used to store computer program code (81) and to transmit the computer program code (81) to the processor (7). The processor (7) is used to execute the power demand uncertainty analysis method based on multi-scenario simulation according to the instructions in the computer program code (81) as described in any one of claims 1-7.

10. A computer-readable storage medium, characterized in that: The computer-readable storage medium stores computer-executable instructions, which, when executed on a computer, implement the power demand uncertainty analysis method based on multi-scenario simulation as described in any one of claims 1-7.

Images