A monte carlo-dbscan-padm medium and long term provincial source and load dynamic probability balance calculation method, system, device and medium

Abstract

The present application relates to the technical field of power analysis, and discloses a Monte Carlo-DBSCAN-PADM long-term provincial source-load dynamic probability balance calculation method, system, equipment and medium, which comprises the following steps: collecting power grid data and performing data cleaning, and setting power grid parameter constraint conditions; generating a source-load full amount uncertainty scenario through Monte Carlo simulation, extracting key features, and clustering the standardized scenario feature vectors; setting a multi-objective function, adjusting the target priority through a weight coefficient, and constructing a source-load dynamic probability balance optimization model; solving the model, and generating a source-load dynamic probability balance result. The present application fully covers the random fluctuations of the source-load double sides, overcomes the defects of the scene missing of the traditional deterministic method, retains the key fluctuation characteristics, greatly reduces the calculation complexity, realizes the parallel solving of sub-problems, and solves the difficulty in solving the multi-constraint and multi-year continuous hour-level problem of the provincial power grid.

Description

Technical Field

[0001] This invention relates to the field of power analysis technology, and in particular to a method, system, equipment, and medium for calculating the dynamic probabilistic balance of provincial source and load in the medium and long term using Monte Carlo-DBSCAN-PADM. Background Technology

[0002] The proportion of new energy sources such as wind power and photovoltaics in the provincial power structure has continued to soar, gradually becoming one of the core forces of power supply. However, the output of new energy sources is highly constrained by natural conditions. Wind power follows a Weibull distribution, and photovoltaics follows a Beta distribution, exhibiting significant randomness and volatility in output. At the same time, the load side is affected by seasonality, industrial activities, and electric vehicle charging, resulting in increasing fluctuations and prediction errors. Furthermore, the superposition of source and load fluctuations can create even more volatile net load curves. This strong uncertainty on both the source and load sides has completely disrupted the relatively stable supply and demand pattern of the traditional power system, posing a huge challenge to the province's medium- and long-term power supply and the consumption of new energy sources.

[0003] Previous methods for calculating provincial power source-load balance have primarily employed deterministic approaches such as tabular methods, estimating power source-load data only for representative months. This approach is completely inadequate to address the uncertainties arising from a high proportion of renewable energy. Even studies incorporating scenario analysis exhibit significant drawbacks: firstly, relying solely on Monte Carlo simulations to generate massive amounts of scenarios easily leads to the "curse of dimensionality," resulting in excessively large and inefficient subsequent calculations; secondly, rudimentary scenario selection or individual scenario simulation methods either miss key fluctuation characteristics or yield overly conservative scheduling results, failing to guarantee the economic efficiency of power system operation and failing to meet the dual requirements of supply guarantee and power consumption. Furthermore, traditional centralized solution methods suffer from low solution efficiency and difficulty in achieving global optimality when facing multiple physical constraints and multi-objective optimization needs within provincial power grids. Therefore, a highly accurate, fast, and reliable method is urgently needed to efficiently support the calculation of dynamic probabilistic power source-load balance across all scenarios and the entire power supply chain. Summary of the Invention

[0004] In view of the aforementioned existing problems, this invention is proposed. Therefore, this invention provides a Monte Carlo-DBSCAN-PADM method for calculating the dynamic probabilistic balance of provincial source loads in the medium and long term, addressing the current urgent need for a method with high accuracy, fast calculation speed, and strong reliability, capable of providing efficient support for the dynamic probabilistic balance calculation of provincial source loads across all scenarios and links.

[0005] To solve the above-mentioned technical problems, the present invention provides the following technical solution: In a first aspect, the present invention provides a method for calculating the dynamic probabilistic balance of provincial source loads in the Monte Carlo-DBSCAN-PADM medium- and long-term regions, including: Collect and clean power grid data, and set power grid parameter constraints. Based on the power grid data, a full uncertainty scenario of source and load is generated through Monte Carlo simulation. Key features of each effective scenario are extracted, high-dimensional feature vectors are generated and standardized. The standardized scenario feature vectors are then clustered to complete the reduction of typical scenarios. Based on the aforementioned feature vectors, a multi-objective function is defined, and the priority of the objectives is adjusted by weight coefficients to construct a source-load dynamic probabilistic balance optimization model, and multiple balance constraints are introduced. The source-load dynamic probability balance optimization model is solved by decomposing the global optimization problem into several sub-problems, solving each sub-problem independently, and then merging the sub-problems through global coordination variables to generate the source-load dynamic probability balance result.

[0006] As a preferred embodiment of the long-term provincial source-load dynamic probabilistic balance calculation method in Monte Carlo-DBSCAN-PADM described in this invention, the method includes: collecting and cleaning power grid data, and setting power grid parameter constraints, including: Collect power grid data, which includes power grid parameters, source-load time series data, and planning boundary conditions; The power grid data is cleaned, and time and spatial scales are set. Power grid topology connectivity constraints, power flow security constraints, source-load operation constraints, and source-load operation constraints are also set.

[0007] As a preferred embodiment of the long-term provincial source-load dynamic probability balance calculation method in Monte Carlo-DBSCAN-PADM described in this invention, the method includes: generating a scenario with full uncertainty of source-load through Monte Carlo simulation, extracting key features of each effective scenario, generating a high-dimensional feature vector and performing standardization processing, and clustering the standardized scenario feature vectors, including: Based on the source-side probability distribution model and the load-side probability distribution model, random sampling is performed through Monte Carlo simulation to generate a set of uncertain source-load full-scale scenarios for several consecutive years at the hour level. Invalid scenarios that violate the basic constraints are eliminated, and valid full-scale scenarios are retained. Extract the temporal and statistical features of each valid full scene, generate a high-dimensional scene feature vector, and perform standardization processing; Cluster the standardized scene feature vectors, automatically identify core points, boundary points and noise points, group scenes with high similarity into the same cluster, select the central scene of each cluster as a typical scene, and calculate the weight of each typical scene. The advantages of this preferred solution are that it can comprehensively simulate the random fluctuations of wind power and photovoltaic output as well as the variable characteristics of load, so that the differentiated information of the scenario can be quantitatively preserved. The DBSCAN clustering can automatically identify typical scenario clusters based on data density and clearly remove noise points, which greatly reduces the number of redundant scenarios and preserves the representative central scenario within the cluster.

[0008] As a preferred embodiment of the long-term provincial source-load dynamic probabilistic balance calculation method in Monte Carlo-DBSCAN-PADM as described in this invention, the construction of the source-load dynamic probabilistic balance optimization model includes: With the objectives of minimizing the probability of curtailment of renewable energy, the probability of load shedding, system operating costs, and maximizing carbon emission reduction, a dynamic probabilistic balance optimization model for source and load is constructed by combining the objective weighting coefficients. Define multiple equilibrium constraints for the model, including physical security constraints, source load balance constraints, and system operation constraints; The beneficial effect of this preferred scheme is that it simultaneously considers four mutually restrictive objectives: power supply reliability, renewable energy consumption, economy, and environmental protection. By flexibly adjusting the priority of each objective through weighting coefficients, it avoids the imbalance between supply and consumption caused by traditional single-objective optimization and achieves overall system coordination.

[0009] As a preferred embodiment of the long-term provincial source-load dynamic probability balance calculation method in Monte Carlo-DBSCAN-PADM as described in this invention, the solution of the source-load dynamic probability balance optimization model includes: The source-load dynamic probabilistic balance optimization model is solved using the PADM method. The global optimization problem is decomposed into spatial partitions and temporal segments. Spatial decomposition divides the power grid into K independent sub-regions according to power supply partitions. Each sub-region is a sub-problem. The sub-regions are coupled by tie-line power constraints. Temporal decomposition divides the hourly task over several years into T sub-periods according to preset time periods. The global coupling constraints are extracted separately, and the remaining constraints are embedded in each sub-problem. The optimal solutions of each sub-region or sub-period are fused through global coordination variables to generate the dynamic probability balance results of provincial source load at the hourly level for many consecutive years, and output core indicators, including probability balance indicators, spatiotemporal distribution indicators, economic and environmental protection indicators, and risk assessment indicators.

[0010] As a preferred embodiment of the long-term provincial source-load dynamic probability balance calculation method in Monte Carlo-DBSCAN-PADM described in this invention, the model solution further includes: Set adaptive penalty parameters and dynamically adjust them based on iterative residuals, initialize dual variables, and set convergence thresholds; The subproblems are solved in parallel. The current dual variable and penalty parameter are fixed. Each sub-region and time-series subproblem solves the local optimum independently. The objective function of the subproblem is the original objective plus the deviation term between the dual variable and the global coordination variable, plus the penalty term. Update the global coordination variable, collect the optimal solutions to all subproblems, and update the global coordination variable to the mean or weighted mean of the shared variables of the subproblems; Update the dual variable and penalty parameter, calculate the original residual and dual residual, update the dual variable, and adaptively adjust the penalty parameter according to the residual change. Determine whether the original residual and dual residual are both less than the set threshold. If so, terminate the iteration and repeat the above steps until the convergence condition is met or the maximum number of iterations is reached. The optimal solutions for each sub-region and sub-time period are fused through global coordination variables to reconstruct the global optimal solution, generating the dynamic probability balance results of provincial source load at the hourly level for consecutive years.

[0011] As a preferred embodiment of the long-term provincial source-load dynamic probability balance calculation method in Monte Carlo-DBSCAN-PADM described in this invention, the extraction of temporal and statistical features for each effective full-scale scenario includes: The time-series characteristics include peak and valley values ​​of renewable energy output, average load during peak load periods, and the correlation coefficient between renewable energy output and load. The statistical characteristics include the volatility of new energy output, the standard deviation of load, the mean curtailment rate within the scenario, and the mean load shedding rate.

[0012] Secondly, this invention provides a long-term provincial source-load dynamic probability balance calculation system in Monte Carlo-DBSCAN-PADM, comprising: The data acquisition module is used to collect power grid data, perform data cleaning, and set power grid parameter constraints. The extraction module is used to generate a full uncertainty scenario of source and load through Monte Carlo simulation based on the power grid data, extract the key features of each effective scenario, generate a high-dimensional feature vector and perform standardization processing, and cluster the standardized scenario feature vectors to complete the reduction of typical scenarios. The model building module is used to set multiple objective functions based on the feature vectors, adjust the priority of the objectives through weight coefficients, build a source-load dynamic probability balance optimization model, and introduce multiple balance constraints. The solution output module is used to solve the source-load dynamic probability balance optimization model. It decomposes the global optimization problem into several sub-problems, solves each sub-problem independently, and merges the sub-problems through global coordination variables to generate the source-load dynamic probability balance result.

[0013] Thirdly, the present invention provides a computer device, comprising: Memory and processor; The memory is used to store computer-executable instructions, and the processor is used to execute the computer-executable instructions. When the computer-executable instructions are executed by the processor, they implement the steps of the long-term provincial source-load dynamic probability balance calculation method in Monte Carlo-DBSCAN-PADM.

[0014] Fourthly, the present invention provides a computer-readable storage medium storing computer-executable instructions, which, when executed by a processor, implement the steps of the long-term provincial source-load dynamic probability balance calculation method in Monte Carlo-DBSCAN-PADM.

[0015] Compared with existing technologies, the beneficial effects of this invention are as follows: This invention collects core data of the provincial power grid, defines basic parameters and constraints, uses Monte Carlo simulation to generate uncertain scenarios of source and load, eliminates invalid scenarios that violate basic constraints, extracts the core features of each valid scenario to form a high-dimensional feature vector, applies the DBSCAN clustering algorithm to the standardized scenario feature vector to reduce typical scenarios, and takes minimizing the probability of new energy curtailment, minimizing the probability of load shedding, minimizing system operating costs, and maximizing carbon emission reduction as the core objectives. It coordinates the priority of objectives through weight coefficients, constructs a source-load dynamic probabilistic balance optimization model, uses PADM to decompose the global optimization problem by spatial partitioning and temporal segmentation, and finally fuses the optimal solutions of each sub-region / sub-period through global coordination variables to generate the hourly dynamic probabilistic balance results of provincial source and load for several consecutive years. Attached Figure Description

[0016] To more clearly illustrate the technical solutions of the embodiments of the present invention, the drawings used in the following description of the embodiments will be briefly introduced. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0017] Figure 1 This is a schematic diagram of the overall process of a long-term provincial source-load dynamic probability balance calculation method according to an embodiment of the present invention, specifically a Monte Carlo-DBSCAN-PADM method. Detailed Implementation

[0018] To make the above-mentioned objects, features, and advantages of the present invention more apparent and understandable, specific embodiments of the present invention will be described in detail below with reference to the accompanying drawings. Obviously, the described embodiments are only a part of the embodiments of the present invention, and not all of them. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort should fall within the protection scope of the present invention.

[0019] Reference Figure 1 As an embodiment of the present invention, a method for calculating the dynamic probabilistic balance of provincial source loads in the Monte Carlo-DBSCAN-PADM model is provided, comprising: S101: Collect power grid data and perform data cleaning; set power grid parameter constraints. S102, based on power grid data, generates a full uncertainty scenario of source and load through Monte Carlo simulation, extracts key features of each effective scenario, generates high-dimensional feature vectors and performs standardization processing, and clusters the standardized scenario feature vectors to complete the reduction of typical scenarios; S103, based on feature vectors, sets multiple objective functions, adjusts the priority of objectives through weight coefficients, constructs a source-load dynamic probability balance optimization model, and introduces multiple balance constraints; S104 solves the source-load dynamic probability balance optimization model by decomposing the global optimization problem into several sub-problems, solving each sub-problem independently, and then merging the sub-problems through global coordination variables to generate the source-load dynamic probability balance result.

[0020] In a preferred embodiment, the process includes collecting and cleaning power grid data, and setting power grid parameter constraints, including: Collect power grid data, which includes power grid parameters, source-load time series data, and planning boundary conditions; Data cleaning is performed on the power grid data, and time and spatial scales are set. Power grid topology connectivity constraints, power flow security constraints, source-load operation constraints, and source-load operation constraints are also set.

[0021] Specifically, core data of the provincial power grid is collected. Grid parameters include topology, equipment parameters, and unit parameters. Topology includes nodes, branches, and tie lines. Equipment parameters include line impedance, power flow limits, and node voltage deviation thresholds. Unit parameters include upper and lower limits of conventional unit output / ramp rates and installed capacity of renewable energy units. Source-load time-series data includes 10 consecutive years and hourly historical data of wind / solar power output, and load demand data, including industrial electricity consumption, residential electricity consumption, and extreme weather loads. Planning boundary conditions include renewable energy capacity expansion plans, unit maintenance cycles, inter-regional mutual assistance agreements, and carbon emission reduction targets. The collected data undergoes data cleaning, using the 3σ criterion to remove outliers, such as sudden changes in wind power output to 0 or exceeding rated capacity. Missing data is supplemented using spatiotemporal interpolation, and high-frequency noise is removed using smoothing filtering to ensure data integrity and reliability.

[0022] The time and space scales include a time dimension of several consecutive years at the hour level, such as 87,600 moments over 10 years, and a spatial dimension covering the entire province, power supply zones, and key nodes at three levels. The constraint boundaries clearly define the grid topology connectivity constraints, i.e., a radial structure without loops; power flow safety constraints, i.e., line power flow ≤ 1.1 times rated capacity; source and load operation constraints, i.e., unit ramp rate ≤ 5% of rated capacity / hour; and source and load operation constraints, i.e., a supply guarantee confidence probability ≥ 99.9% and a new energy consumption confidence probability ≥ 90%.

[0023] In a preferred embodiment, a Monte Carlo simulation is used to generate a scenario with full source-load uncertainty. Key features of each valid scenario are extracted, a high-dimensional feature vector is generated and standardized, and the standardized scenario feature vectors are clustered, including: Based on the source-side probability distribution model and the load-side probability distribution model, random sampling is performed through Monte Carlo simulation to generate a set of uncertain source-load full-scale scenarios for several consecutive years at the hour level. Invalid scenarios that violate the basic constraints are eliminated, and valid full-scale scenarios are retained. Extract the temporal and statistical features of each valid full scene, generate a high-dimensional scene feature vector, and perform standardization processing; The standardized scene feature vectors are clustered to automatically identify core points, boundary points, and noise points. Scenes with high similarity are grouped into the same cluster, and the central scene of each cluster is selected as a typical scene. The weight of each typical scene is calculated.

[0024] In a preferred embodiment, extracting the temporal and statistical features of each valid full-scale scene includes: The time-series characteristics include peak and valley values ​​of renewable energy output, average load during peak load periods, and the correlation coefficient between renewable energy output and load. Statistical characteristics include the volatility of new energy output, load standard deviation, mean curtailment rate within the scenario, and mean load shedding rate.

[0025] Specifically, Monte Carlo simulation is used to generate uncertain scenarios with full source and load capacity. Invalid scenarios that violate basic constraints are eliminated, and 800-1200 valid full scenarios are retained. Among them, the source-side modeling uses a Weibull distribution for wind power output, with the shape parameter k and scale parameter c obtained by fitting historical data. The photovoltaic output uses a Beta distribution, with parameters α and β calibrated using historical irradiance data. Conventional unit maintenance uses a binomial distribution, and the maintenance probability is set based on the maintenance cycle.

[0026] The Weibull distribution of wind power output has the following probability density function: In the formula, v is the wind speed (m / s); k is the shape parameter, reflecting the degree of concentration of wind speed distribution; and c is the scale parameter (m / s), reflecting the average wind speed level.

[0027] The photovoltaic output has a Beta distribution, and the probability density function is: In the formula, I is the normalized value of solar irradiance, α and β are shape parameters, which are calibrated from irradiance data. The parameter calibration method includes collecting and normalizing hourly irradiance data over many years, obtaining α and β using moment estimation / maximum likelihood estimation, and obtaining the photovoltaic output distribution by linear conversion between irradiance and photovoltaic output.

[0028] When modeling the load side, the normal load is adopted as the normal distribution, the mean is the historical average load for the same period, and the variance is the standard deviation of the load fluctuation. The extreme weather load is adopted as the log-normal distribution, and the fitting is based on historical data of extreme scenarios such as cold waves and high temperatures. The flexible load response error is adopted as the uniform distribution, with a range of ±10%.

[0029] The normal distribution of the conventional load has the following probability density function: In the formula, μ is the historical average load for the same period, σ is the standard deviation of load fluctuation, and P L This refers to the active power of a normal load.

[0030] Extreme weather loads follow a log-normal distribution with the following probability density function: Flexible load response error Δ P The uniform distribution has the following probability density function: Based on the above probability distribution, random sampling is performed using Monte Carlo simulation to generate a set of hourly-level full-volume uncertain scenarios for multiple consecutive years. Specifically, 100-150 scenarios are generated each year, totaling 1000-1500 initial scenarios over 10 years. Each scenario includes wind power output sequence, photovoltaic power output sequence, load demand sequence, and scenario occurrence probability, all initially equal. Invalid scenarios violating basic constraints, such as power flow exceeding limits or voltage deviation exceeding standards, are removed, ultimately retaining 800-1200 valid full-volume scenarios. When generating initial scenarios using Monte Carlo simulation, the initial probability of each scenario is equal, being the reciprocal of the total number of initial scenarios. After removing invalid scenarios through constraint verification, the probabilities of valid scenarios are renormalized so that the sum of valid scenario probabilities is 1. After obtaining typical scenarios through density-based spatial clustering of applications with noise (DBSCAN), the occurrence probability of typical scenarios is determined by the proportion of scenarios within a cluster to the total number of valid scenarios, serving as the scenario weight for subsequent optimization models.

[0031] Furthermore, after obtaining the full set of valid scenes using Monte Carlo simulation, the core features of each valid scene are extracted using the DBSCAN clustering method to form a high-dimensional feature vector, thereby reducing the number of typical scenes, as follows: Construct scene feature vectors, extract the core features of each effective scene, and form a high-dimensional feature vector. That is, combine multiple different types of key features of each effective scene to form a mathematical vector containing multiple dimensions of values. There are many of these features, and each feature corresponds to one dimension of the vector. The scenario feature vector includes time-series features, namely, peak / valley values ​​of renewable energy output, average load during peak load periods (19:00-22:00), and the correlation coefficient between renewable energy output and load. The correlation coefficient between renewable energy output and load refers to the linear correlation between the total hourly renewable energy output and the hourly load demand within a single scenario. It is calculated using the Pearson correlation coefficient, with a value range of [-1,1], to quantify the time-series matching characteristics of source and load. Using the hourly renewable energy output sequence and load sequence within the scenario as samples, the Pearson correlation coefficient between the two is calculated to reflect the synchronicity, complementarity, or reverse fluctuation characteristics of renewable energy generation and electricity load, serving as the core time-series feature for scenario clustering. Statistical features include renewable energy output volatility, load standard deviation, mean curtailment rate within the scenario, and mean load shedding rate. Min-Max standardization is used to map the feature vector to the [0,1] interval to avoid the impact of dimensional differences on the clustering effect.

[0032] DBSCAN clustering parameters are optimized and executed. The neighborhood radius ε is determined using the k-distance curve method, typically taken as 1.2-1.5 times the standard deviation of the feature vector. The minimum number of points (MinPts) is set based on the load / new energy fluctuation intensity: MinPts = 8-12 in load-intensive areas and 5-7 in remote areas. DBSCAN clustering is performed on the standardized scene feature vectors to automatically identify core points, boundary points, and noise points. Scenes with high similarity are grouped into the same cluster, while noise points, i.e., isolated scenes, are directly removed. The center scene of each cluster is selected as the typical scene, i.e., the scene corresponding to the mean vector of all scene feature vectors in the cluster. The weight of each typical scene is calculated as: weight = number of scenes in the cluster / total number of all valid scenes. Finally, 30-80 typical scenes are retained, which can cover more than 95% of the uncertainty features.

[0033] In a preferred embodiment, constructing the source-load dynamic probabilistic balance optimization model includes: With the objectives of minimizing the probability of curtailment of renewable energy, the probability of load shedding, system operating costs, and maximizing carbon emission reduction, a dynamic probabilistic balance optimization model for source and load is constructed by combining the objective weighting coefficients. Define multiple balancing constraints for the model, including physical security constraints, source-load balance constraints, and system operation constraints.

[0034] Specifically, with the core objectives of minimizing the probability of renewable energy curtailment, minimizing the probability of load shedding, minimizing system operating costs, and maximizing carbon emission reduction, a dynamic probabilistic balance optimization model for source and load is constructed by coordinating objective priorities through weighting coefficients. The objective function expression is as follows: In the formula, P cur For the probability of curtailment of renewable energy, P shed For the load shedding probability, C op The system operating cost includes conventional unit fuel costs, energy storage charging and discharging costs, and cross-regional mutual aid costs. E CO2 Carbon emission reductions are based on the amount of electricity generated by replacing thermal power with new energy sources. w 1- w 4 is the target weight, satisfying w 1+ w 2+ w 3+ w 4=1, with default values ​​of 0.4, 0.3, 0.2, and 0.1.

[0035] The constraints include physical safety constraints, source-load balance constraints, and operational constraints. Physical safety constraints include topology constraints, power flow constraints, and voltage constraints, which are expressed as follows: in, x ij The branch circuit is in an on / off state. N The total number of nodes. P ij For the branch line ij tidal current, P ij,max To maximize the availability of the trend. U i For nodes i Voltage, U i,min , U i,max These are the minimum and maximum voltage limits.

[0036] The source-load balance constraint is expressed as: In the formula, s For the scene, t For a moment, P exchange For cross-regional power exchange, P g For the active power output of a conventional unit g, P ren To contribute to the development of new energy power units P load For load power, P cur For wind / solar power curtailment, P shed This refers to the load shedding power.

[0037] The operational constraints are expressed as follows: In the formula, P g,min , P g,max These represent the minimum and maximum active power outputs.

[0038] In a preferred embodiment, solving the source-load dynamic probability balance optimization model includes: The source-load dynamic probabilistic balance optimization model is solved using the Parallel Alternating Direction Method of Multipliers (PADM). The global optimization problem is decomposed into spatial partitions and temporal segments. Spatial decomposition divides the power grid into K independent sub-regions according to power supply zones, with each sub-region serving as a sub-problem. The sub-regions are coupled by tie-line power constraints. Temporal decomposition divides the continuous multi-year hourly task into T sub-periods according to preset time periods. Global coupling constraints are extracted separately, while other constraints are embedded in each sub-problem. The optimal solutions of each sub-region or sub-period are fused through global coordination variables to generate the dynamic probability balance results of provincial source load at the hourly level for many consecutive years, and output core indicators, including probability balance indicators, spatiotemporal distribution indicators, economic and environmental protection indicators, and risk assessment indicators.

[0039] It should be noted that the global optimization problem is decomposed by spatial partitioning and temporal segmentation using PADM. Finally, the optimal solutions of each sub-region / sub-period are fused through global coordination variables to generate the dynamic probability balance results of provincial source load at the hourly level for many consecutive years.

[0040] In a preferred embodiment, model solving further includes: Set adaptive penalty parameters and dynamically adjust them based on iterative residuals, initialize dual variables, and set convergence thresholds; The subproblems are solved in parallel. The current dual variable and penalty parameter are fixed. Each sub-region and time-series subproblem solves the local optimum independently. The objective function of the subproblem is the original objective plus the deviation term between the dual variable and the global coordination variable, plus the penalty term. Update the global coordination variable, collect the optimal solutions to all subproblems, and update the global coordination variable to the mean or weighted mean of the shared variables of the subproblems; Update the dual variable and penalty parameter, calculate the original residual and dual residual, update the dual variable, and adaptively adjust the penalty parameter according to the residual change. Determine whether the original residual and dual residual are both less than the set threshold. If so, terminate the iteration and repeat the above steps until the convergence condition is met or the maximum number of iterations is reached. The optimal solutions for each sub-region and sub-time period are fused through global coordination variables to reconstruct the global optimal solution, generating the dynamic probability balance results of provincial source load at the hourly level for consecutive years.

[0041] Specifically, spatial decomposition divides the provincial power grid into K independent sub-regions according to power supply zones, with each sub-region serving as a sub-problem. Sub-regions are coupled through tie-line power constraints. Temporal decomposition divides the continuous multi-year hourly tasks into T sub-periods according to preset time periods, which are annual-quarter-period. Cross-period coupling constraints, such as the cumulative operating hours of generating units, are handled by the global coordination layer. Decoupling coupling constraints extracts global coupling constraints such as cross-regional tie-line power constraints and cross-period unit maintenance constraints separately. Other constraints, such as regional power flow constraints and intra-period source-load balance constraints, are embedded in sub-problems. Among them, cross-regional tie-line power constraints include tie-line transmission power upper and lower limit constraints and ramp rate constraints, ensuring that the tie-line exchange power does not exceed the safe transmission limit and that the power change rate meets the system stability requirements. This constraint is related to multiple regional sub-problems and needs to be coordinated uniformly in the upper-level main problem. The constraints on unit maintenance across time periods include the total annual maintenance time constraint and the continuity constraint of maintenance periods. These constraints ensure that the unit maintenance plan meets the maintenance cycle and operation and maintenance requirements. Since these constraints are coupled across multiple time periods, they cannot be handled separately in a single time period sub-problem and need to be optimized in a coordinated manner in the upper-level main problem.

[0042] Further, PADM parameters are configured and initialized, including penalty parameters, dual variables, and convergence thresholds. An adaptive penalty factor is set in the penalty parameters. p t initial value p 0=10, dynamically adjusted based on the iterative residual; when the residual increases... p t+1 =1.2 p t When the residual decreases p t+1 =0.9 p t； Initialize the dual variables, corresponding to the Lagrange multipliers of the coupling constraints, with initial values ​​set to 0; set the residual threshold, with the original residual €pri=10. -4 Dual residual €dual=10 -4 The maximum number of iterations, Tmax, is 50.

[0043] Furthermore, PADM iterative solutions are performed, where the dual variable λ is fixed during the parallel solution of subproblems. 0 and penalty parameters p k Each sub-region or sub-time period independently solves for the optimal solution of its sub-problem. For example, in region i, the unit output and renewable energy consumption are calculated, and the objective function of the sub-problem is: In the formula, z k These are global coordination variables, corresponding to the shared variables of coupling constraints. xi Let i be the local decision variable for subproblem i. f i ( x i Let be the original objective function of subproblem i, and let λ represent the optimization objective of the subproblem itself. k For the k-th iteration, the Lagrange multiplier is the dual variable, used to measure the impact of the deviation between the subproblem variable and the global coordination variable.

[0044] Update the global coordination variable and collect the optimal solutions to all subproblems. x i k Update global coordination variables z k+1 Take the mean or weighted mean of the shared variables in the subproblem.

[0045] Update the dual variable and penalty parameter by updating the dual variable based on the original residual and the dual residual, while adaptively adjusting the penalty parameter. p k+1 Repeat the iteration until the residual satisfies the threshold or reaches the upper limit of iteration, where the original residual, dual residual, and dual variable are respectively: in, This represents the subproblem in the k-th iteration. i The local exchange power variable, i.e., the local variable related to the inter-regional tie line, s k For the k-th iteration, the local exchange power dual variable (Lagrange multiplier) of the cross-regional connection line.

[0046] Furthermore, after decomposing and solving the global optimization problem using PADM, the optimal solutions for each sub-region / sub-period are fused through global coordination variables to generate multi-year, hourly-level provincial source-load dynamic probability balance results. Integrating the sub-problem results, the optimal solutions for each sub-region / sub-period are fused through global coordination variables to generate multi-year, hourly-level provincial source-load dynamic probability balance results, outputting core indicators.

[0047] The probability balance indicators include the confidence probability of ensuring supply, the confidence probability of new energy consumption, and the probability of exceeding the supply-demand imbalance limit, as detailed below: Confidence probability of supply R sup The percentage of scenarios in which the system can meet load demand across all scenarios, i.e., the percentage where no load shedding occurs or the load shedding amount is below a threshold, reflects power supply reliability and is expressed as: In the formula, N The total number of scenes, P shed,s For the load shedding power in scenario s, P shed,max This is the maximum allowable load shedding threshold, typically 0 or a very low value.

[0048] Confidence probability of new energy consumption R ren To determine the percentage of scenarios where the renewable energy absorption rate is not lower than the set target value across all scenarios, reflecting the reliability of renewable energy absorption, the formula is: In the formula, For the renewable energy consumption rate under scenario s, The set target for renewable energy consumption rate, such as 95%, P cur,s,t Let be the power of wind / solar curtailment at scene s and time t.

[0049] Probability of supply and demand imbalance exceeding the standard R imb The formula for the percentage of scenarios where the system's supply-demand imbalance (including load shedding, power curtailment, and inter-regional power transfer) exceeds the safety threshold across all scenarios reflects the overall operational risk of the system. In the formula, △ P imb,s Let △ be the total supply-demand imbalance in scenario s. P max The maximum allowable imbalance threshold.

[0050] Spatiotemporal distribution indicators include zonal balance coordination degree, load guarantee rate of key nodes, and distribution of curtailment rate in renewable energy-rich areas, as detailed below: Partition balance and coordination C reg To quantify the degree of difference in supply and demand balance between regions, a lower value indicates better coordination between regions, while a higher value indicates a more severe imbalance in supply and demand. The formula is as follows: In the formula, M represents the total number of regions, and △ P reg,m For the region m The supply and demand imbalance rate The average imbalance rate across all regions. P reg,m,t For the regionm , t The rate of supply and demand imbalance at any given moment. P ren,m,t For the region m , t Contributing effort at all times, P load,m,t For the region m , t Load power at any given time.

[0051] critical node load guarantee rate R node For critical load nodes, such as important users and hub substations, the probability of no power outage or voltage exceedance in all scenarios reflects the power supply reliability of key loads. The formula is: In the formula, U n,s,t Let n be the voltage of node n in scene s at time t. P shed,n,s,t U is the load shedding power at node n. min U max This represents the allowable range of node voltage.

[0052] Distribution of curtailment rates in renewable energy-rich areas This refers to the statistical distribution of curtailment rates in renewable energy-rich areas, such as regions with a high proportion of wind / solar power, under various scenarios. This includes the mean, maximum value, and quantiles, such as the 95th percentile curtailment rate. The formula is: In the formula, P cur,m,s,t Let m be the region, s be the scene, and t be the time when wind / solar curtailment power. P ren,m,s,t The active power output for region m, scene s, and time t.

[0053] The 95th percentile curtailment rate can be further calculated. , represented as: In the formula, m is the number of the new energy rich area. P cur,m,s,t Let be the power wasted in region m at scene s and time t.

[0054] Economic and environmental indicators include system operating costs, unit balance costs, and carbon emission reductions, as detailed below: System operating costs C sys This refers to the expected value of system operating costs across all scenarios, including conventional unit fuel costs, start-up and shutdown costs, load shedding costs, and curtailment penalty costs. The formula is: In the formula, T is the probability of scenario s occurring, and C g (P g,s,t Let C be the fuel cost function of unit g. su,g For the single start-up and shutdown cost of unit g, u g,s,t This is the unit start / stop status variable; 1 indicates start / stop and 0 indicates otherwise. (C) shed C cur The unit penalty cost for load shedding and power abandonment.

[0055] Unit Balance Cost C unit This refers to the operating cost corresponding to a unit of supply and demand equilibrium, reflecting the economic efficiency of the plan. The formula is: In the formula, the denominator is the sum of the total system load and the expected value of the renewable energy output, representing the total power scale that needs to be balanced.

[0056] Carbon emission reduction refers to the reduction in carbon dioxide emissions relative to a baseline scenario, i.e., without new energy sources or stationary operation methods, reflecting environmental benefits. The formula is: In the formula, E g (P) is the carbon emission intensity function of unit g under output P. For the output of unit g in the baseline scenario, π s Let be the probability of the occurrence of the s-th random scene.

[0057] Risk assessment indicators include the maximum supply-demand gap in extreme scenarios and the probability of trend exceeding limits, as detailed below: Maximum supply-demand gap in extreme scenarios △ P gap,max This refers to the maximum value of the system supply-demand gap across all scenarios, including load, power curtailment, and power output. It reflects the risk of supply disruption under extreme conditions. The formula is: In the formula, P load,s,t Let P be the load power at scene s and time t. g,s,t For scenario s and time t, the output power of conventional unit g is P. ren,s,t For scenario s, time t, the output power of the new energy unit ren, P cur,s,t Let be the power of wind / solar curtailment at scene s and time t.

[0058] Trend exceeding limits risk probability R flow This refers to the percentage of scenarios in which the power flow of a line / transformer exceeds the safety limit across all scenarios, reflecting the overload risk of power grid equipment. The formula is: In the formula, P l,s,t For the line l In scenario s and time t, the power flow P l,lim For the line l The trend of safety limits.

[0059] It should be noted that this invention collects core data of the provincial power grid, defines basic parameters and constraints, uses Monte Carlo simulation to generate uncertain scenarios of source and load, eliminates invalid scenarios that violate basic constraints, extracts the core features of each valid scenario to form a high-dimensional feature vector, applies the DBSCAN clustering algorithm to the standardized scenario feature vector to reduce typical scenarios, and then uses minimizing the probability of new energy curtailment, minimizing the probability of load shedding, minimizing system operating costs, and maximizing carbon emission reduction as the core objectives. By coordinating the priority of objectives through weight coefficients, a dynamic probabilistic balance optimization model of source and load is constructed. The global optimization problem is decomposed by spatial partitioning and temporal segmentation using PADM. Finally, the optimal solutions of each sub-region / sub-period are fused through global coordination variables to generate hourly dynamic probabilistic balance results of provincial source and load over several consecutive years.

[0060] This invention fully covers the random fluctuations on both the source and load sides through Monte Carlo simulation, overcoming the shortcomings of traditional deterministic methods in terms of scenario omissions; it utilizes DBSCAN clustering to automatically identify typical scenarios and eliminate redundant and noisy scenarios, retaining key fluctuation characteristics and significantly reducing computational complexity; it takes power curtailment, load shedding, operating costs, and carbon emission reduction as joint objectives, and flexibly balances multiple demands through adjustable weights; it employs PADM to decompose the global optimization into spatial and temporal components, enabling parallel solution of sub-problems and solving the difficulties of solving multi-constraint, multi-year, continuous hourly problems in provincial power grids.

[0061] The above is an illustrative scheme of a medium- and long-term provincial source-load dynamic probability balance calculation method according to this embodiment of the Monte Carlo-DBSCAN-PADM method. It should be noted that the technical solution of this Monte Carlo-DBSCAN-PADM medium- and long-term provincial source-load dynamic probability balance calculation system belongs to the same concept as the technical solution of the aforementioned Monte Carlo-DBSCAN-PADM medium- and long-term provincial source-load dynamic probability balance calculation method. Details not described in detail in this embodiment of the Monte Carlo-DBSCAN-PADM medium- and long-term provincial source-load dynamic probability balance calculation system can be found in the description of the aforementioned Monte Carlo-DBSCAN-PADM medium- and long-term provincial source-load dynamic probability balance calculation method.

[0062] This embodiment provides a long-term provincial source-load dynamic probability balance calculation system in Monte Carlo-DBSCAN-PADM, including: The data acquisition module is used to collect power grid data, perform data cleaning, and set power grid parameter constraints. The extraction module is used to generate uncertain source-load scenarios based on power grid data through Monte Carlo simulation, extract key features of each effective scenario, generate high-dimensional feature vectors and perform standardization processing, and cluster the standardized scenario feature vectors to complete the reduction of typical scenarios. The model building module is used to set multiple objective functions based on feature vectors, adjust the priority of objectives through weight coefficients, build a source-load dynamic probability balance optimization model, and introduce multiple balance constraints. The solution output module is used to solve the source-load dynamic probability balance optimization model. It decomposes the global optimization problem into several sub-problems, solves each sub-problem independently, and merges the sub-problems through global coordination variables to generate the source-load dynamic probability balance result.

[0063] This embodiment also provides a computer device suitable for long-term provincial source-load dynamic probability balance calculation in Monte Carlo-DBSCAN-PADM, including: The system includes a memory and a processor. The memory stores computer-executable instructions, and the processor executes these instructions to implement a long-term provincial source-load dynamic probability balance calculation method as proposed in the Monte Carlo-DBSCAN-PADM embodiment above.

[0064] This embodiment also provides a storage medium storing a computer program that, when executed by a processor, implements a long-term provincial source-load dynamic probability balance calculation method as proposed in the Monte Carlo-DBSCAN-PADM embodiment above.

[0065] The storage medium proposed in this embodiment belongs to the same inventive concept as the method for calculating the dynamic probability balance of provincial source load in Monte Carlo-DBSCAN-PADM proposed in the above embodiment. Technical details not described in detail in this embodiment can be found in the above embodiments, and this embodiment has the same beneficial effects as the above embodiments.

[0066] Based on the above description of the implementation methods, those skilled in the art can clearly understand that the present invention can be implemented using software and necessary general-purpose hardware, and of course, it can also be implemented using hardware, but in many cases the former is a better implementation method. Based on this understanding, the technical solution of the present invention, or the part that contributes to the prior art, can be embodied in the form of a software product. This computer software product can be stored in a computer-readable storage medium, such as a computer floppy disk, read-only memory (ROM), random access memory (RAM), flash memory, hard disk, or optical disk, etc., including several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute the methods of the various embodiments of the present invention.

[0067] It should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and not to limit it. Although the present invention has been described in detail with reference to preferred embodiments, those skilled in the art should understand that modifications or equivalent substitutions can be made to the technical solutions of the present invention without departing from the spirit and scope of the technical solutions of the present invention, and all such modifications or substitutions should be covered within the scope of the claims of the present invention.

Claims

1. A Monte Carlo-DBSCAN-PADM medium and long term provincial load dynamic probability balance calculation method, characterized in that, include: Collect and clean power grid data, and set power grid parameter constraints. Based on the power grid data, a full uncertainty scenario of source and load is generated through Monte Carlo simulation. Key features of each effective scenario are extracted, high-dimensional feature vectors are generated and standardized. The standardized scenario feature vectors are then clustered to complete the reduction of typical scenarios. Based on the aforementioned feature vectors, a multi-objective function is defined, and the priority of the objectives is adjusted by weight coefficients to construct a source-load dynamic probabilistic balance optimization model, and multiple balance constraints are introduced. The source-load dynamic probability balance optimization model is solved by decomposing the global optimization problem into several sub-problems, solving each sub-problem independently, and then merging the sub-problems through global coordination variables to generate the source-load dynamic probability balance result.

2. The Monte Carlo-DBSCAN-PADM mid-long term provincial source-load dynamic probability balance calculation method according to claim 1, characterized in that, Collect and clean power grid data, and set power grid parameter constraints, including: Collect power grid data, which includes power grid parameters, source-load time series data, and planning boundary conditions; The power grid data is cleaned, and time and spatial scales are set. Power grid topology connectivity constraints, power flow security constraints, source-load operation constraints, and source-load operation constraints are also set.

3. The Monte Carlo-DBSCAN-PADM mid-long term provincial source-load dynamic probability balancing calculation method according to claim 1, characterized in that, Monte Carlo simulations were used to generate scenarios with full source-load uncertainty. Key features of each valid scenario were extracted, high-dimensional feature vectors were generated and standardized, and the standardized scenario feature vectors were clustered, including: Based on the source-side probability distribution model and the load-side probability distribution model, random sampling is performed through Monte Carlo simulation to generate a set of uncertain source-load full-scale scenarios for several consecutive years at the hour level. Invalid scenarios that violate the basic constraints are eliminated, and valid full-scale scenarios are retained. Extract the temporal and statistical features of each valid full scene, generate a high-dimensional scene feature vector, and perform standardization processing; The standardized scene feature vectors are clustered to automatically identify core points, boundary points, and noise points. Scenes with high similarity are grouped into the same cluster, and the central scene of each cluster is selected as a typical scene. The weight of each typical scene is calculated.

4. The Monte Carlo-DBSCAN-PADM mid-long term provincial source-load dynamic probability balancing calculation method of claim 3, wherein, The construction of the source-load dynamic probabilistic balance optimization model includes: With the objectives of minimizing the probability of curtailment of renewable energy, the probability of load shedding, system operating costs, and maximizing carbon emission reduction, a dynamic probabilistic balance optimization model for source and load is constructed by combining the objective weighting coefficients. Define multiple balancing constraints for the model, including physical security constraints, source-load balance constraints, and system operation constraints.

5. The Monte Carlo-DBSCAN-PADM mid-long term provincial source-load dynamic probability balancing calculation method according to claim 4, characterized in that, Solving the source load dynamic probability balance optimization model includes: The source-load dynamic probabilistic balance optimization model is solved using the PADM method. The global optimization problem is decomposed into spatial partitions and temporal segments. Spatial decomposition divides the power grid into K independent sub-regions according to power supply partitions. Each sub-region is a sub-problem. The sub-regions are coupled by tie-line power constraints. Temporal decomposition divides the hourly task over several years into T sub-periods according to preset time periods. The global coupling constraints are extracted separately, and the remaining constraints are embedded in each sub-problem. The optimal solutions of each sub-region or sub-period are fused through global coordination variables to generate the dynamic probability balance results of provincial source load at the hourly level for many consecutive years, and output core indicators, including probability balance indicators, spatiotemporal distribution indicators, economic and environmental protection indicators, and risk assessment indicators.

6. The Monte Carlo-DBSCAN-PADM mid-long term provincial source-load dynamic probability balancing calculation method of claim 5, wherein, The model solution also includes: Set adaptive penalty parameters and dynamically adjust them based on iterative residuals, initialize dual variables, and set convergence thresholds; The subproblems are solved in parallel. The current dual variable and penalty parameter are fixed. Each sub-region and time-series subproblem solves the local optimum independently. The objective function of the subproblem is the original objective plus the deviation term between the dual variable and the global coordination variable, plus the penalty term. Update the global coordination variable, collect the optimal solutions to all subproblems, and update the global coordination variable to the mean or weighted mean of the shared variables of the subproblems; Update the dual variable and penalty parameter, calculate the original residual and dual residual, update the dual variable, and adaptively adjust the penalty parameter according to the residual change. Determine whether the original residual and dual residual are both less than the set threshold. If so, terminate the iteration and repeat the above steps until the convergence condition is met or the maximum number of iterations is reached. The optimal solutions for each sub-region and sub-time period are fused through global coordination variables to reconstruct the global optimal solution, generating the dynamic probability balance results of provincial source load at the hourly level for consecutive years.

7. The method for calculating the dynamic probabilistic balance of provincial source loads in the long term using Monte Carlo-DBSCAN-PADM as described in claim 3, characterized in that, The temporal and statistical features extracted for each valid full-data scenario include: The time-series characteristics include peak and valley values ​​of renewable energy output, average load during peak load periods, and the correlation coefficient between renewable energy output and load. The statistical characteristics include the volatility of new energy output, the standard deviation of load, the mean curtailment rate within the scenario, and the mean load shedding rate.

8. A Monte Carlo-DBSCAN-PADM medium-to-long-term provincial source-load dynamic probabilistic balance calculation system, employing the Monte Carlo-DBSCAN-PADM medium-to-long-term provincial source-load dynamic probabilistic balance calculation method as described in any one of claims 1 to 7, characterized in that, include: The data acquisition module is used to collect power grid data, perform data cleaning, and set power grid parameter constraints. The extraction module is used to generate a full uncertainty scenario of source and load through Monte Carlo simulation based on the power grid data, extract the key features of each effective scenario, generate a high-dimensional feature vector and perform standardization processing, and cluster the standardized scenario feature vectors to complete the reduction of typical scenarios. The model building module is used to set multiple objective functions based on the feature vectors, adjust the priority of the objectives through weight coefficients, build a source-load dynamic probability balance optimization model, and introduce multiple balance constraints. The solution output module is used to solve the source-load dynamic probability balance optimization model. It decomposes the global optimization problem into several sub-problems, solves each sub-problem independently, and merges the sub-problems through global coordination variables to generate the source-load dynamic probability balance result.

9. A computer device, comprising: include: Memory and processor; The memory is used to store computer-executable instructions, and the processor is used to execute the computer-executable instructions. When the computer-executable instructions are executed by the processor, they implement the steps of the long-term provincial source-load dynamic probability balance calculation method according to any one of claims 1 to 7.

10. A computer-readable storage medium, characterized in that, It stores computer-executable instructions, which, when executed by a processor, implement the steps of the long-term provincial source-load dynamic probability balance calculation method according to any one of claims 1 to 7 in Monte Carlo-DBSCAN-PADM.

Images