Power grid investment planning scheme rationality evaluation method considering low-carbon target

Abstract

The invention relates to a power grid investment planning scheme rationality evaluation method considering a low-carbon target, and the method comprises the steps: collecting related data and indexes related to power grid investment planning, and carrying out the region division of a power grid load prediction region based on the related data and indexes, the method comprises the following steps: decomposing load data of a power system into three component parts by using a seasonal decomposition method of a time sequence, constructing a prediction model for different components to obtain a load prediction result, clustering annual daily wind and light output and load data by using a k-means algorithm, and selecting a typical day of each quarter to perform short-term operation simulation to obtain a load prediction result. A planning scheme evaluation model with the minimum operation cost as the target is constructed, the matching relation between output and load curves of various power supply types is simulated, and the peak regulation capacity and economical efficiency of the system and the investment and planning scheme of the power system are optimized; and solving the multi-target model through a marine predator algorithm, and if the planning schemes all meet the peak regulation requirements of the system, considering that the planning result is reasonable.

Description

Technical Field

[0001] The present invention belongs to the technical field of power grid planning and evaluation, and in particular to a method for evaluating the rationality of a power grid investment planning scheme taking into account a low-carbon target. Background Art

[0002] With the severe challenges of global climate change and the increasing urgency of sustainable development goals, power grid investment planning has shifted from simply pursuing economic benefits and power supply security to more comprehensive considerations, among which low-carbon goals have become one of the key orientations. Traditional power grid planning often focuses on meeting the growing demand for electricity, optimizing the power grid structure, and reducing costs, but insufficient consideration is given to environmental impacts, especially carbon emissions. With the rapid development of renewable energy, the transformation of energy structure, and the fulfillment of international emission reduction commitments, power grid investment must take into account both economic and environmental benefits to ensure that while meeting future load growth, it promotes the efficient use of low-carbon energy and significantly reduces carbon emissions.

[0003] Problems with existing technology

[0004] Static planning methods: Many existing grid planning methods fail to fully consider the dynamics of energy transformation and carbon emission constraints, resulting in planning results that may not be able to adapt to future adjustments to low-carbon policies and rapid changes in renewable energy.

[0005] Incomplete carbon emission assessment: When evaluating power grid investment projects, carbon emission analysis is often limited to direct emissions, ignoring indirect emissions throughout the life cycle, such as equipment manufacturing, transportation, construction and decommissioning.

[0006] Data and model limitations: Existing models have difficulty in obtaining data when dealing with complex systems, such as grid stability under a high proportion of renewable energy access, cross-regional energy flow and scheduling, and other issues. Model assumptions are overly simplified, making it difficult to accurately predict carbon emissions and costs.

[0007] Efficiency and accuracy of optimization algorithms: Traditional optimization algorithms have problems such as slow convergence and low solution quality when dealing with large-scale, multi-objective, and strongly constrained power grid planning problems, making it difficult to find the optimal investment plan that is both economical and environmentally friendly.

[0008] Application (patent) number: CN201510078491.6 "Distribution network planning investment optimization method based on coordination of reliability and economy" mentions determining the unit load reliability improvement cost and unit load increase cost, then determining the investment benefits of the corresponding partitions, and finally establishing a distribution network investment optimization allocation model with the goal of maximizing investment benefits. Compared with this patent, the advantages of this patent are:

[0009] Comprehensive evaluation model: This plan proposes a rationality evaluation model for power grid investment planning that comprehensively considers low-carbon goals. By integrating multiple factors such as load forecasting, energy supply and demand balance, carbon emission measurement, and economic cost analysis, a comprehensive and dynamic evaluation system is formed.

[0010] Dynamic optimization framework: Adopt a method that combines rolling planning with scenario analysis to adapt to the rapid changes in energy policies and market demand, dynamically adjust investment direction and scale, and ensure the foresight and flexibility of planning schemes.

[0011] Application of advanced algorithms: Combined with advanced optimization algorithms, such as genetic algorithms, particle swarm optimization, deep reinforcement learning, etc., to improve the computational efficiency and solution accuracy of the model when processing large-scale data and multi-objective optimization problems, and quickly find the optimal or approximately optimal low-carbon investment strategy.

[0012] Integration of policies and market mechanisms: Fully consider external factors such as carbon trading markets, green financial instruments, and subsidy policies, optimize investment cost and benefit analysis, and promote the economic feasibility and market competitiveness of clean energy projects. Summary of the invention

[0013] In order to solve the above problems existing in the prior art, the present invention proposes a method for evaluating the rationality of a power grid investment planning scheme taking into account a low-carbon target.

[0014] The technical solution of the present invention is as follows:

[0015] On the one hand, the present invention provides a method for evaluating the rationality of a power grid investment planning scheme considering a low-carbon target, comprising the following steps:

[0016] Collect relevant data and indicators related to power grid investment planning, divide the power grid load forecast area based on the relevant data and indicators, use the seasonal decomposition method of time series to decompose the power system load data into three parts: real-time variation, trend stability component and irregular fluctuation component, preprocess the power system load data to obtain a data set, divide the preprocessed data set into a training data set and a test data set, build a power grid load forecast model for different components, train and learn the load forecast model with the divided training data set, use the trained load forecast model to predict the samples in the test data set to obtain the load forecast result, perform denormalization on the load forecast result to obtain the final load forecast result; compare the actual load value and the final load forecast result to obtain the error between the two, evaluate the forecast model through the error, and finally obtain the load evaluation result;

[0017] The k-means algorithm is used to cluster the daily wind and solar power output and load data throughout the year, and typical days in each quarter are selected for short-term operation simulation;

[0018] Construct a planning scheme evaluation model with the goal of minimizing operating costs, simulate the matching relationship between the output and load curves of various power sources, optimize the peak-shaving capacity and economy of the power system based on the matching relationship, and optimize the investment and planning scheme of the power system;

[0019] The complex constraints in the planning scheme evaluation model are solved by the marine predator algorithm. If all planning schemes meet the system peak-shaving requirements, the planning results are considered reasonable.

[0020] Preferably, the relevant data include: local statistical yearbooks, national economic and social development plans, urban and rural master plans, land use master plans, power grid operation and basic statistical data, and superior power grid planning results;

[0021] The indicators include: regional GDP, population, grid asset size, geographical area, and load density;

[0022] Cluster analysis method is used to divide the grid load forecast area into regions by combining relevant data and indicators. The specific steps include:

[0023] Assume c is the number of clusters, is a vector set in the feature space, where v i (i=1, 2...) represents the cluster prototype vector of the i-th class; μ ik ∈U, U is a c×n fuzzy matrix that satisfies the following conditions:

[0024]

[0025] The objective function of the improved fuzzy clustering algorithm is defined as:

[0026] max{J(U,V)}

[0027]

[0028] in μ ik For data point x k The membership degree of cluster i; μ ik For data point x k The membership degree V belonging to cluster j is the clustering result;

[0029] Given the number of clustering categories c (2≤c≤n, n is the number of data); set the iteration stop threshold ε and the fuzzy index m; initialize the clustering prototype model v (0) ; Set iteration counter b = 0;

[0030] For any i, k, if there exists d ik b >0, then

[0031]

[0032] If there exists i, r such that d ik b = 0, then μ ir ( b )=1, and for j≠i, μ ir ( b )=0;

[0033]

[0034] If ||V (b+1) -V (b) ||<ε, the improved fuzzy clustering algorithm stops; and outputs the partition matrix U and cluster prototype V, otherwise b=b+1, and iterates to execute the improved fuzzy clustering algorithm;

[0035] Determine the best classification tree C m ; Calculate the value of J(C). When the condition is met: J(Cm) = max{J(C)}, Cm is considered to be the optimal number of classifications; obtain the corresponding output partition matrix U and cluster prototype to form the final clustering result.

[0036] Preferably, the steps of decomposing the power system load data into three parts: real-time variation, trend stability component and irregular fluctuation component include:

[0037] Decompose the original load sequence in the power system load data into subsequences; t It is expressed as the actual time series value of the t period. The time series seasonal trend adjustment method converts Y t Decomposed into three components: seasonal component, trend component, and irregular component; these three components are combined by an additive or multiplicative model as follows:

[0038] The additive decomposition model is defined as:

[0039] Y t =Trend t +Seasonal t +Irregular t

[0040] The multiplicative factorization model is defined as:

[0041] Y t =Trend t ×Seasonal t ×ιrregular t

[0042] In the above two formulas, Trendt is the trend component in time period t, Seasonal t is the seasonal component in time period t, Irregular t is the irregular component in time period t;

[0043] When the seasonal fluctuations are proportional to the level of the time series, a multiplicative model is used, and when the seasonal fluctuations do not depend on the level of the time series, an additive decomposition model is used;

[0044] The feedforward neural network method is used to build prediction models for the above decomposition models respectively, and finally superimpose them to obtain the final load prediction result.

[0045] Preferably, the k-means algorithm is used to cluster the daily wind and solar power output and load data throughout the year, and a typical day in each quarter is selected for short-term operation simulation. The specific steps are:

[0046] Data collection and preprocessing: First, collect wind and solar power output data and grid load data for every day of the year. The wind and solar power output data includes wind speed corresponding to wind power generation, and sunshine duration and intensity corresponding to photovoltaic power generation. Clean the data, remove outliers, and standardize the data.

[0047] Determine the number of clusters: Select the K value, which is predetermined by the elbow rule and silhouette coefficient. For power system data with obvious seasonality, the K value is set to 4, corresponding to the typical day of each quarter of the year;

[0048] Initialization and iteration: The K-means algorithm starts by randomly selecting K center points and iterating the following steps until convergence:

[0049] Assign each data point to the category of the nearest center point; recalculate the center of each category until the change in the center point is less than a preset threshold or the maximum number of iterations is reached;

[0050] Typical day selection: The center point of each cluster represents a typical day in a quarter, which contains the typical characteristics of wind power, solar power generation and grid load in that quarter; typical day data is used for simulation and analysis to reduce the amount of data processing during actual simulation operation.

[0051] Preferably, the planning scheme evaluation model is constructed with the goal of minimizing the operating cost, including an objective function and constraints;

[0052] The operation simulation layer takes the lowest operation cost as the objective function to verify the system peak load regulation capability; the operation cost F 2 It includes three aspects: fuel cost and coal-fired power unit start-up and shutdown cost F 1t , Energy storage device charging, discharging and maintenance costs F 2tand interruptible load compensation and wind and solar curtailment penalty costs F 3t ; Specifically as shown in the following formula:

[0053]

[0054] Where: t is the number of hours, T is the total number of hours for short-term operation simulation; X RA , X CA , X KA are the number of fuel units, energy storage units, and renewable energy units participating in the short-term operation simulation; are the fuel cost, startup cost, and shutdown cost of the i-th fuel unit respectively; They are the actual output, startup state variable, and shutdown state variable of the i-th fuel unit at the t-th hour; are the unit electricity discharging cost, charging cost, and maintenance cost of the j-th energy storage device respectively; are the real-time discharge power, charging power, and actual output of the j-th energy storage device at the t-th hour; is the interruptible load; ρ IL , ka They are interruptible load compensation cost coefficient and wind and solar power abandonment factor respectively; are the available power and actual output of the kth energy storage device at the tth hour respectively;

[0055] Mode The fuel cost can be expressed as a quadratic function:

[0056]

[0057] Where: a i , b i 、c i are the consumption coefficients of the i-th fuel unit respectively.

[0058] Preferably, the constraints specifically include: power balance constraints, fuel unit output constraints, unit start and stop time constraints, renewable energy output constraints, various power supply ramp constraints, system hot standby constraints and energy storage constraints;

[0059] The power balance constraint is that in the power system, the output of all power sources and the consumption of all loads need to be balanced, as shown in the following formula:

[0060]

[0061] Where: are the active outputs of the fuel unit, energy storage unit, and renewable energy unit at the tth hour respectively; is the interruptible load of the system in the short-term operation simulation at the tth hour; is the load demand of the system at hour t;

[0062] The fuel unit output constraint is that during the operation of the power system, the real-time output of the fuel unit needs to meet the upper and lower limit constraints, as shown in the following formula:

[0063]

[0064] Where: U i，t The start and stop status of the unit; are the minimum output and maximum output of the fuel unit respectively; is the actual output of the fuel unit at hour t;

[0065] The start and stop time constraint of the unit is specifically shown in the following formula:

[0066]

[0067] Where: T A，i (t), T B，i (t) are the time that the i-th unit needs to keep running or shut down at the t-th hour, which is calculated as:

[0068]

[0069] Where: I A，i ,I B，i They are the minimum operating interval between two shutdowns and the minimum shutdown interval between two starts of the unit respectively;

[0070] The specific step of the renewable energy output constraint is to set the maximum output limit of renewable energy, as shown in the following formula:

[0071]

[0072] in, The amount of available resources for renewable energy;

[0073] The various power supply ramp constraints include the following aspects, as shown in the following formula:

[0074]

[0075] Wherein, R is the unit climbing rate;

[0076] The system hot standby constraint is specifically shown in the following formula:

[0077]

[0078] Among them, XW The number of hydropower units participating in the short-term operation simulation; are the available resources and actual output of hydropower generation respectively; ρ is the hot standby coefficient;

[0079] The energy storage constraint is to constrain the working state of the energy storage device in the short-term operation simulation to meet the system requirements and adjust the balance between the system load and the power supply. The specific upper and lower limits of the charge and discharge power are as follows:

[0080]

[0081] in, are the charging and discharging power of the energy storage device respectively; are the rated power of the energy storage device and the input power to the grid respectively; are the charging and discharging states of the energy storage device respectively; k is the margin coefficient;

[0082] Working state constraints are as follows:

[0083]

[0084] The energy storage constraints for charging and discharging are as follows:

[0085]

[0086] Among them, E j，t is the actual storage capacity of the energy storage device at the tth hour; j is the energy self-consumption coefficient of the energy storage device; It is the energy conversion efficiency of the energy storage device during the charging and discharging process; are the maximum and minimum storage capacity of the energy storage device respectively; the energy storage device has an equal capacity constraint at the beginning and end of the short-term operation cycle, as shown in the following formula:

[0087]

[0088] Preferably, the method of solving the complex constraints in the planning scheme evaluation model by the marine predator algorithm includes a high speed ratio stage, a medium speed ratio stage and a low speed ratio stage, specifically the following steps:

[0089] Initialize the positions of predators and preys, construct an elite matrix with predators with the best fitness, and construct a prey matrix with evenly distributed preys;

[0090] The high speed ratio stage is the initial stage of algorithm optimization iteration, and the global position information is surveyed. The mathematical model of this stage is expressed as follows:

[0091]

[0092] in, is the moving step length; is a random vector based on Brownian walk normal distribution; is a term-by-term multiplication operation; P is a constant, equal to 0.5; is a uniform random vector in [0, 1]; Iter is the current iteration number; Max_Iter is the maximum iteration number; N is the population size;

[0093] The medium speed ratio stage is the middle stage of the algorithm iteration. The population is divided into two parts, in which the prey performs Lévy flight and is responsible for the development of the space, and the predator performs Brownian motion and is responsible for the exploration of the space. The mathematical model expression of this stage is as follows:

[0094]

[0095] in, represents the random vector of Levy motion, CF represents the adaptive parameter controlling the moving step length of the predator;

[0096] The medium speed ratio stage is the late stage of algorithm iteration, and the predator adopts the Lévy flight strategy and pays more attention to the development of local areas. The mathematical model expression of this stage is as follows:

[0097]

[0098] External environmental factors such as fish aggregation devices and eddy effects change the foraging strategy of predators to jump out of local extreme values, so that the algorithm avoids premature convergence. The mathematical model expression is as follows:

[0099]

[0100] Among them, FADs is the impact probability, which is generally taken as 0.2; is a binary vector; r is a random number in [0, 1]; r 1 、r 2 are random indices of the prey matrix, respectively;

[0101] The ocean predator algorithm is used to solve the objective function under complex constraints such as coal-fired power output constraints and power balance constraints, and optimize the operation of the power grid, including the optimization of peak-shaving capacity and economy. After optimization, the gap between the actual peak-shaving capacity and the expected peak load demand is calculated. If the gap is lower than the pre-set threshold, the power planning and operation strategy is effective and can meet the system peak-shaving needs, and the power grid operation is safe and stable. Otherwise, it is necessary to readjust the optimization model parameters or increase the peak-shaving resources, and continue to use the ocean predator algorithm for optimization.

[0102] On the other hand, the present invention also provides a system for evaluating the rationality of a power grid investment planning scheme taking into account a low-carbon target, comprising:

[0103] The load assessment module collects relevant data and indicators related to the power grid investment planning, divides the power grid load forecast area based on the relevant data and indicators, uses the seasonal decomposition method of time series to decompose the power system load data into three parts: real-time variation, trend stability component and irregular fluctuation component, preprocesses the power system load data to obtain a data set, divides the preprocessed data set into a training data set and a test data set, builds a power grid load forecasting model for different components, trains and learns the load forecasting model with the divided training data set, uses the trained load forecasting model to predict the samples in the test data set to obtain the load forecasting result, and performs denormalization on the load forecasting result to obtain the final load forecasting result; compares the actual load value and the final load forecasting result to obtain the error between the two, evaluates the forecasting model through the error, and finally obtains the load assessment result;

[0104] The simulation module uses the k-means algorithm to cluster the daily wind and solar power output and load data throughout the year, and selects typical days in each quarter for short-term operation simulation;

[0105] The optimization module builds a planning scheme evaluation model with the goal of minimizing operating costs, simulates the matching relationship between the output and load curves of various power sources, optimizes the peak-shaving capacity and economy of the power system based on the matching relationship, and optimizes the investment and planning scheme of the power system;

[0106] The optimization solution module solves the complex constraints in the planning scheme evaluation model through the marine predator algorithm. If all planning schemes meet the system peak-shaving requirements, the planning results are considered reasonable.

[0107] On the other hand, the present invention further provides an electronic device having a computer program stored thereon, wherein when the computer program is executed by a processor, the method for evaluating the rationality of a power grid investment planning scheme taking into account a low-carbon target as described in any embodiment of the present invention is implemented.

[0108] On the other hand, the present invention also provides a computer-readable medium for storing one or more programs, which, when executed by the one or more processors, enables the one or more processors to implement a method for evaluating the rationality of a power grid investment planning scheme that takes into account low-carbon goals as described in any embodiment of the present invention.

[0109] The present invention has the following beneficial effects:

[0110] 1. The present invention realizes high-precision prediction of power grid load through comprehensive data collection and analysis. The diversification of data sources, including but not limited to regional GDP, population size, power grid asset scale, geographical area and load density, provides a rich information basis for load forecasting and ensures the comprehensiveness and accuracy of the model. By accurately dividing the forecast area and combining the seasonal decomposition method of time series analysis, the load data is decomposed into three parts: trend, seasonality and random fluctuations, and prediction models are established for different components. This method not only captures the inherent laws of load changes, but also effectively improves the accuracy of the prediction. The combination of refined regional division and component prediction model makes the load forecast closer to the actual situation and provides a solid foundation for subsequent planning;

[0111] 2. The present invention uses the K-means clustering algorithm to intelligently cluster the wind and solar power output and load data throughout the year, effectively identifying the typical days in each quarter. This strategy greatly simplifies the complexity of the operation simulation while ensuring the representativeness of the simulation. Through in-depth analysis and simulation of typical days, the operation status of the power grid in different seasons and weather conditions can be more accurately evaluated, which provides an important basis for optimizing the peak-shaving strategy and improving the utilization rate of renewable energy such as wind and solar power, and promotes the flexibility and efficiency of the power system;

[0112] 3. The planning scheme evaluation model constructed by the present invention aims to minimize operating costs, while taking into account multiple factors such as peak-shaving capacity, economy, and carbon emissions, reflecting systematicity and comprehensiveness. The model optimizes the power supply configuration and scheduling strategy by simulating the matching relationship between the output and load curves of various power sources. It not only improves the peak-shaving capacity of the system, but also effectively reduces operating costs, achieving dual optimization of economic and environmental benefits. This multi-objective optimization from a global perspective is not available in traditional single-objective planning methods, and has opened up a new path for achieving sustainable development of power grids;

[0113] 4. The present invention uses the marine predator algorithm to solve the multi-objective model, which is a highlight of this patented technology. Compared with traditional optimization algorithms, the marine predator algorithm, with its unique biological inspiration mechanism, exhibits better global search capabilities and faster convergence speed, and can effectively deal with the complex and variable factors and constraints involved in power grid planning. Through the introduction of this algorithm, the system can quickly find the optimal or near-optimal solution that meets multiple constraints such as peak-shaving demand, economy, and low-carbon emissions, ensuring the rationality and feasibility of the planning results;

[0114] 5. The technical framework of the present invention emphasizes dynamic adaptability to market changes, policy adjustments and technological advances. Through rolling planning and real-time data updates, the model can timely adjust the forecast model parameters, optimization strategies and investment decisions to ensure the forward-looking and flexibility of the planning scheme. This dynamic optimization mechanism ensures the long-term effectiveness of power grid investment planning and its ability to resist uncertainty, and is the key to achieving stable operation of the power grid and low-carbon transformation. BRIEF DESCRIPTION OF THE DRAWINGS

[0115] Figure 1 It is a schematic diagram of the process of the present invention;

[0116] Figure 2 This is a schematic diagram of the neural network structure of an embodiment of the present invention;

[0117] Figure 3 This is a schematic diagram of the neural network concept of an embodiment of the present invention;

[0118] Figure 4 The following is a schematic diagram of optimizing an embodiment of the present invention. DETAILED DESCRIPTION

[0119] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0120] It should be understood that the step numbers used in this document are only for convenience of description and are not intended to limit the order in which the steps are executed.

[0121] It should be understood that the terms used in the present specification are only for the purpose of describing specific embodiments and are not intended to limit the present invention. As used in the present specification and the appended claims, unless the context clearly indicates otherwise, the singular forms "a", "an" and "the" are intended to include plural forms.

[0122] The terms “include” and “comprising” indicate the presence of described features, integers, steps, operations, elements and / or components, but do not exclude the presence or addition of one or more other features, integers, steps, operations, elements, components and / or combinations thereof.

[0123] The term "and / or" means and includes any and all possible combinations of one or more of the associated listed items.

[0124] Embodiment 1:

[0125] See also Figure 1-4, a method for evaluating the rationality of a power grid investment planning scheme considering a low-carbon target, comprising the following steps:

[0126] S10. Collect relevant data and indicators related to power grid investment planning, divide the power grid load forecast area into regions based on the relevant data and indicators, use the seasonal decomposition method of time series to decompose the power system load data into three parts: real-time variation, trend stability component and irregular fluctuation component, preprocess the power system load data to obtain a data set, divide the preprocessed data set into two parts: a training data set and a test data set, build a power grid load forecast model for different components, train and learn the load forecast model with the divided training data set, use the trained load forecast model to predict the samples in the test data set to obtain the load forecast result, perform denormalization on the load forecast result to obtain the final load forecast result; compare the actual load value and the final load forecast result to obtain the error between the two, evaluate the forecast model through the error, and finally obtain the load evaluation result;

[0127] S20. Use the k-means algorithm to cluster the daily wind and solar power output and load data throughout the year, and select typical days in each quarter for short-term operation simulation;

[0128] S30. Construct a planning scheme evaluation model with the goal of minimizing operating costs, simulate the matching relationship between the output and load curves of various power types, optimize the peak-shaving capacity and economy of the power system based on the matching relationship, and optimize the investment and planning scheme of the power system;

[0129] S40. The complex constraints in the planning scheme evaluation model are solved by the marine predator algorithm. If all planning schemes meet the system peak-shaving requirements, the planning results are considered reasonable.

[0130] As a preferred implementation mode of this embodiment, in steps S10-S40, relevant data are first collected, the grid load is predicted, and the load prediction area is accurately divided. The regional division is performed based on indicators such as regional GDP, population size, grid asset scale, geographical area, and load density to improve the accuracy of load prediction. Then, the seasonal decomposition method of time series is used to decompose the power system load data into three parts: real-time variation, trend stability component, and irregular fluctuation component. Then, prediction models are constructed for different components to finally form load prediction results.

[0131] The main documents and relevant data for planning are collected, including: the statistical yearbook of the region, the national economic and social development plan, the overall urban and rural plan, the overall land use plan, the operation of the power grid, basic statistical data, and the planning results of the superior power grid. Combined with data collection, regional division is carried out based on indicators such as regional GDP, population, power grid asset scale, geographical area, and load density to improve the accuracy of load forecasting.

[0132] Using cluster analysis methods and combining the above-mentioned related indicators, we divide regions and user types and carry out load forecasting work in a targeted manner.

[0133] Assume c is the number of clusters, is a vector set in the feature space, where v i (i=1, 2...) represents the cluster prototype vector of the i-th class. ik ∈U, U is a c×n fuzzy matrix,

[0134] This matrix satisfies the following conditions:

[0135]

[0136]

[0137] The objective function of the improved fuzzy clustering algorithm is defined as:

[0138] max{J(U,V)}

[0139]

[0140] in μ ik For data point x k The membership degree of cluster i; μ ik For data point x k The membership degree V belonging to cluster j is the clustering result;

[0141] Given the number of clustering categories c (2≤c≤n, n is the number of data); set the iteration stop threshold ε and the fuzzy index m. Initialize the clustering prototype pattern v(0); set the iteration counter b=0

[0142] For any i, k, if there exists d ik b >0, then

[0143]

[0144] If there exists i, r such that d ik b = 0, then μ ir (b) =1, and for j≠i, μir (b) =0

[0145]

[0146] If ||V (b+1) -V (b) ||<ε, the improved fuzzy clustering algorithm stops, and outputs the partition matrix U and cluster prototype V, otherwise b=b+1 and goes to the previous step.

[0147] Determine the best classification tree C m . Calculate the value of J(C). When the condition is met: J(Cm) = max{J(C)}, Cm is considered to be the optimal number of classifications. The corresponding output partition matrix U and cluster prototype are obtained, that is, the final clustering result.

[0148] For grid load forecasting, the seasonal decomposition method of time series is used to decompose the power system load data into three parts: real-time variation, trend stability component and irregular fluctuation component. Then, prediction models are constructed for different components to finally form load forecast results.

[0149] The general steps of load forecasting are as follows.

[0150] (1). Load data preprocessing.

[0151] (2) Divide the preprocessed dataset into two parts: training dataset and test dataset.

[0152] (3) Construct a forecasting model. According to the requirements of load forecasting, select and construct the corresponding load forecasting model.

[0153] (4) Sample training. Use the training samples divided in the second step to train the load forecasting model.

[0154] (5) Load prediction. Use the load prediction model trained in the fourth step to predict the samples in the test data set. The load prediction value is the model output value.

[0155] (6) Forecast result processing and evaluation. The final load forecast result can be obtained by denormalizing the load forecast result in the previous step. Then, the error between the actual load value and the predicted value is used to evaluate the forecast model.

[0156] The seasonal decomposition method of time series is used to decompose the power system load data into three parts: real-time variation, trend stability component and irregular fluctuation component. For time series regression, seasonal and trend decomposition is a process of improving parameter estimation properties. Seasonal and trend adjustments usually help to better understand time series data. For example, if the electricity consumption in June increased by 20% compared to May. By decomposing the data, it is easier to find that the increase in electricity consumption is mainly caused by seasonal effects related to weather. Therefore, if the original load series is decomposed into better explained subsequences, better load forecasting results can be obtained.

[0157] For time series regression, seasonal and trend decomposition is a process that improves the properties of parameter estimates. Seasonal and trend adjustments often help to better understand time series data. For example, if electricity consumption in June increased by 20% compared to May. By decomposing the data, it is easier to find that the increase in electricity consumption is mainly caused by seasonal effects related to weather. Therefore, if the original load series is decomposed into better explained subsequences, better load forecasting results can be obtained.

[0158] Y t It is expressed as the actual time series value of the t period. The time series seasonal trend adjustment method converts Y t It is decomposed into three components: trend component, seasonal component and irregular component. These three components are represented by an additive or multiplicative model.

[0159] Combination.

[0160] The additive decomposition model is defined as:

[0161] Y t =Trend t +Seasonal t +Irregular t

[0162] The multiplicative factorization model is defined as:

[0163] Y t =Trend t ×Seasonal t ×Irregular t

[0164] In the above two formulas, Trend t is the trend component in time period t, Seasonal t is the seasonal component in time period t,

[0165] Irregular t is the irregular component in time period t.

[0166] The multiplicative model is suitable when the seasonal fluctuations are proportional to the level of the time series, and the additive decomposition model is better when the seasonal fluctuations do not depend on the level of the time series.

[0167] In addition to the two basic decomposition models mentioned above, the decomposition model can also be defined as:

[0168] Y t =Seasonal t +Trend t and Irregular t

[0169] When the seasonal component changes slowly, the above model can be used to decompose the time series. Therefore, the non-seasonal forecast model can be applied to the combination of trend and irregular components. Similar models are also applicable to multiplicative models:

[0170] Y t =Seasonal t ×Trend t and Irregular t

[0171] The feedforward neural network method is used to build prediction models for the above three components respectively, and finally superimpose them to obtain the final load forecasting result.

[0172] The feedforward neural network method is used to build prediction models for the above three components respectively, and finally superimpose them to obtain the final load forecasting result.

[0173] Let X 1 , X 2 , …, X n is the BP neural network input vector, Y 1 , Y 2 , …, Y m is the output value, w ij and w jk is the weight, the typical BP neural network topology diagram is as follows Figure 1 shown.

[0174] When the number of input and output nodes of the BP neural network is n and m respectively, it reflects the mapping relationship between n independent variables and m dependent variables. BP neural network modeling and prediction includes three steps: network structure construction, training and prediction. The basic workflow is as follows: Figure 3 shown.

[0175] Assume that the activation function of each layer of nodes in the network is an S-type function, and the input of the first layer i node in the network is denoted as net i , the output is recorded as o i , the output of the kth node in the output layer is y k, then the input of the jth node in the middle layer is:

[0176]

[0177] o j =f(net j )

[0178]

[0179] Define the network error as the difference between the expected output and the actual output, then we have If the output layer has i neurons, the square error between the actual output and the expected output is defined as:

[0180]

[0181] Since the BP algorithm corrects the weights according to the negative gradient of the error E, the modification of the weights can be expressed as;

[0182] W m+1 =w m +Δw m =wm-λg m

[0183] Where m represents the number of iterations,

[0184] where λ is the learning step size.

[0185]

[0186] Because it is the output layer, is the actual output value, according to e k The definition of and square error can be obtained:

[0187]

[0188] According to e k The definition of can be obtained:

[0189]

[0190] According to the above formula We can get:

[0191]

[0192] According to the above formula k =∑ j w kj o j We can get:

[0193]

[0194] Finally we get:

[0195]

[0196] Now let the learning error of the output layer be:

[0197] σ k =e k f′(net k )

[0198] have to:

[0199]

[0200] (2) The weight modification value of the hidden layer neural unit Δw kj :

[0201]

[0202] According to the above formula j =∑ i w ji o i and j =f(net j ) can be obtained:

[0203]

[0204] Because we are looking for the change in the weight of the hidden layer, we should consider the effect of the previous layer on it, which is inherent:

[0205]

[0206] According to the net k =∑ j w kj o j Knowable

[0207]

[0208] According to We can get:

[0209]

[0210] Bundle Bring-in Derived:

[0211]

[0212] Let the learning error of the hidden layer be:

[0213]

[0214] The model is then optimized using a genetic algorithm.

[0215] The k-means algorithm is further used to cluster the daily wind and solar power output and load data throughout the year, and typical days in each quarter are selected for short-term operation simulation.

[0216] The K-means clustering algorithm is an iterative clustering analysis method that aims to split a data set into K clusters so that each data point belongs to the center (centroid) of the nearest cluster. Its main purpose is to minimize the sum of squared errors from all data points to the centroid of the cluster to which they belong. The following are the basic steps and related formulas of the K-means algorithm:

[0217] Initialization: Randomly select K data points as the initial centroids (μ1, μ2, …, μk).

[0218] Assignment: For each data point x in the dataset, calculate its distance to all centroids and assign it to the cluster with the nearest centroid.

[0219] Distance calculation usually uses Euclidean distance:

[0220]

[0221] Among them, xx is an n-dimensional data point, μi is the i-th centroid, and xj and μij are their values ​​in the j-th dimension respectively.

[0222] Update the centroid: For each cluster Ci, recalculate the mean of all its members as the new centroid:

[0223]

[0224] Here, |Ci| represents the number of data points in cluster Ci.

[0225] Repeat: Repeat steps 2 and 3 until the centroid no longer changes significantly or the preset maximum number of iterations is reached.

[0226] Result: Finally, K clusters are obtained, each cluster consists of its centroid and all data points belonging to the cluster.

[0227] Stopping criteria:

[0228] The change in the centroid position is less than the threshold: If the moving distance of all centroids is less than a small threshold ε, the iteration is stopped.

[0229] Reaching the maximum number of iterations: A maximum number of iterations is pre-set, and the algorithm stops when it is reached, even if the centroid is still changing. The center point of each cluster represents a "typical day" in a quarter, which contains the typical characteristics of wind power, solar power generation and grid load in that quarter. These typical day data can be used for simulation and analysis, reducing the amount of data processing when actually running the simulation while maintaining a high degree of representativeness.

[0230] Further construct a planning scheme evaluation model with the goal of minimizing operating costs, simulate the matching relationship between the output and load curves of various power types, so as to optimize the system's peak-shaving capacity and economy, and optimize the investment and planning schemes of the power system.

[0231] 1. Objective Function

[0232] The operation simulation layer takes the lowest operation cost as the objective function to verify the system peak load regulation capability. Considering the base load operation, stable operation and usually not participating in peak load regulation operation, the operation cost F of the operation simulation layer is 2 It includes three aspects: ① Fuel costs (mainly coal and natural gas) and coal-fired power unit start-up and shutdown costs F 1t ;②Charging, discharging and maintenance costs of energy storage device F 2t ; ③ Interruptible load compensation and wind and solar power abandonment penalty fees F 3t .

[0233]

[0234] Where: t is the number of hours, T is the total number of hours for short-term operation simulation; X RA , X CA , X KA are the number of fuel units, energy storage units, and renewable energy units participating in the short-term operation simulation; are the fuel cost, startup cost, and shutdown cost of the i-th fuel unit respectively; They are the actual output, startup state variable, and shutdown state variable of the i-th fuel unit at the t-th hour; are the unit electricity discharging cost, charging cost, and maintenance cost of the j-th energy storage device respectively; are the real-time discharge power, charging power, and actual output of the j-th energy storage device at the t-th hour; is the interruptible load; ρ IL , ka They are interruptible load compensation cost coefficient and wind and solar power abandonment factor respectively; are the available power and actual output of the kth energy storage device at the tth hour respectively.

[0235] Mode The fuel cost can be expressed as a quadratic function:

[0236]

[0237] Where: a i 、b i 、c i are the consumption coefficients of the i-th fuel unit respectively.

[0238] 2. Constraints

[0239] (1) Power balance constraints

[0240] The power balance constraint means that in the power system, the output of all power sources and the consumption of all loads need to be balanced. At this stage, the constraint that the sum of the output of all types of power sources is equal to the total social load is very important to ensure the stable operation of the power system. If the output power of the power source is lower than the power consumed by the load, it will cause an imbalance in the power system, which may cause grid failures and even power outages.

[0241]

[0242] Where: are the active outputs of the fuel unit, energy storage unit, and renewable energy unit at the tth hour respectively; is the interruptible load of the system in the short-term operation simulation at the tth hour; is the load demand of the system at the tth hour.

[0243] (2) Fuel unit output constraints

[0244] During the operation of the power system, the real-time output of the fuel units needs to meet the upper and lower limit constraints.

[0245]

[0246] Where: U i，t Indicates the start and stop status of the unit; are the minimum output and maximum output of the fuel unit respectively; It is the actual output of the fuel unit at the tth hour.

[0247] (3) Unit start and stop time constraints

[0248] The start-up time and shutdown time are related to the flexible adjustment capability of the unit, and play an important role in coping with fluctuations in electricity demand and improving the consumption of clean energy. Generally speaking, the shorter the start-up and shutdown time, the higher the reliability of the system. However, too frequent start-ups and shutdowns will also affect the life of the generator set.

[0249] Therefore, there should be a time interval between each start and stop of the unit.

[0250]

[0251] Where: T A，i (t), T B，i (t) are the time that the i-th unit needs to keep running or shut down at the t-th hour, which is calculated as:

[0252]

[0253] Where: I A，i ,I B，i They are respectively the minimum operating interval between two shutdowns of the unit and the minimum shutdown interval between two starts.

[0254] (4) Renewable energy output constraints

[0255] The output of renewable energy such as wind, solar, and hydropower is affected by weather and environment, so its maximum output is also volatile and uncertain. In short-term operation simulation, in order to ensure the stable operation and safety of the system, it is necessary to set the maximum output limit of renewable energy.

[0256]

[0257] Where: The amount of renewable energy available.

[0258] (5) Various power supply ramp constraints

[0259] In the short-term operation simulation of the power system, the ramp constraint of various power sources is an important constraint condition, which refers to the time required for the output of various generators in the power system to increase from zero to the rated power. The ramp constraints of various power sources usually include the following aspects:

[0260]

[0261] Where: R is the unit climbing rate.

[0262] (6) System hot standby constraints

[0263] System hot standby constraint is an important constraint in the short-term operation simulation of the power system. It is used to quickly dispatch the backup generator sets in the event of sudden load fluctuations or unit failures. These backup units can be quickly started in emergency situations such as load fluctuations or unit failures to ensure system reliability. Common hot standby units include coal-fired units and hydropower units.

[0264]

[0265] Where: XW is the number of hydropower units participating in the short-term operation simulation; P l W,max , are the available resources and actual output of hydropower respectively; ρ is the hot standby coefficient.

[0266] (7) Energy storage constraints

[0267] In short-term operation simulation, it is necessary to constrain the working state of the energy storage device to ensure that it can meet the needs of the system and be more efficient in adjusting the balance between system load and power supply.

[0268]

[0269] Where: are the charging and discharging power of the energy storage device respectively; are the rated power of the energy storage device and the input power to the grid respectively; are the charging and discharging states of the energy storage device respectively; k is the margin coefficient.

[0270] Working status constraints:

[0271]

[0272] Energy storage constraints for charging and discharging:

[0273]

[0274] Where: E j,t is the actual storage capacity of the energy storage device at the tth hour; j is the energy self-consumption coefficient of the energy storage device; It is the energy conversion efficiency of the energy storage device during the charging and discharging process; The maximum and minimum storage capacity of the energy storage device are respectively. The energy storage device has equal capacity constraints at the beginning and end of the short-term operation cycle:

[0275]

[0276] Finally, the multi-objective model is solved by the marine predator algorithm. If all planning schemes meet the system peak-shaving requirements, the planning results are considered reasonable.

[0277] The specific solution steps are as follows:

[0278] The marine predator algorithm is a new meta-heuristic optimization algorithm proposed by Afshin Faramarzi et al. in 2020. By simulating the natural law of survival of the fittest in the ocean, various marine organisms constantly switch their identities as predators and prey, and change their foraging strategies according to different situations, thereby performing an optimization process.

[0279] The marine predator algorithm believes that the foraging strategy of marine predators varies between Lévy flight and Brownian walk, and chooses between the two strategies according to different scenarios to obtain the optimal predation strategy.

[0280] First, the positions of predators and prey are initialized, and the elite matrix is ​​constructed with the predator with the best fitness, and the prey matrix is ​​constructed with the uniformly distributed prey, as shown in the following formulas.

[0281] According to different speed ratios, the optimization process of MPA is divided into three stages.

[0282] Phase 1: Survey Phase

[0283] This stage is also called the high-speed ratio stage. In this stage, the prey is much faster than the predator. The predator adopts a static strategy, while the prey performs Brownian motion. This stage often occurs in the early stages of algorithm optimization iterations and is a survey of global position information. The mathematical model of this stage is described as follows:

[0284]

[0285] Remember to change SPE and i=1, n

[0286] in is the moving step length; is a random vector based on Brownian walk normal distribution; is a term-by-term multiplication operation; P is a constant, equal to 0.5; is a uniform random vector in [0,1]; Iter is the current number of iterations; Max_Iter is the maximum number of iterations; N is the population size.

[0287] Stage 2: Medium speed ratio stage

[0288] This stage is also called the medium speed ratio stage. In this stage, the speeds of the predator and the prey are similar, and both are looking for their own prey. This stage generally occurs in the middle of the algorithm iteration. The population is divided into two parts, in which the prey performs Lévy flight and is responsible for the development of the space, and the predator performs Brownian motion and is responsible for the exploration of the space. The mathematical model expression of this stage is as follows:

[0289]

[0290]

[0291] in, represents the Lévy motion random vector, and CF represents the adaptive parameter that controls the predator's moving step size.

[0292] Phase 3: Development Phase

[0293] This stage is also called the low-vulgarity stage. In this stage, the speed of the predator is greater than the speed of the prey. It mainly occurs in the late stage of algorithm iteration. The predator adopts the Lévy flight strategy and pays more attention to the development of local areas. The mathematical model expression of this stage is as follows:

[0294]

[0295] In addition, the algorithm also takes into account external environmental factors such as fish aggregation devices and eddy effects, and changes the foraging strategy of predators to jump out of local extremes and avoid premature convergence. The mathematical model expression is as follows:

[0296]

[0297] Where FADs is the impact probability, which is generally taken as 0.2; is a binary vector; r is a random number in [0,1]; r 1 、r 2 are random indices of the prey matrix, respectively.

[0298] The Marine Predators Algorithm (MPA) is used to solve the objective function under complex constraints including coal power output constraints and power balance constraints, aiming to achieve the optimization of power grid operation, especially the optimization of peak load regulation capacity and economy. After obtaining the optimized solution, the gap between the actual peak load regulation capacity and the expected peak load demand (peak load shortage) is calculated. If the value is lower than the pre-set threshold, it indicates that the power supply planning and operation strategy is effective, can meet the system peak load demand, and the power grid operation is safe and stable; otherwise, it is necessary to readjust the optimization model parameters or increase the peak load regulation resources to continue the optimization.

[0299] Embodiment 2:

[0300] The load assessment module collects relevant data and indicators related to the power grid investment planning, divides the power grid load forecast area into regions based on the relevant data and indicators, uses the seasonal decomposition method of time series to decompose the power system load data into three parts: real-time variation, trend stability component and irregular fluctuation component, preprocesses the power system load data to obtain a data set, divides the preprocessed data set into two parts: a training data set and a test data set, constructs a power grid load forecasting model for different components, trains and learns the load forecasting model with the divided training data set, uses the trained load forecasting model to predict the samples in the test data set to obtain the load forecasting result, and performs denormalization on the load forecasting result to obtain the final load forecasting result; compares the actual load value and the final load forecasting result to obtain the error between the two, evaluates the forecasting model through the error, and finally obtains the load assessment result; this module is used to implement the function of step S10 in the above-mentioned implementation one, and will not be described in detail;

[0301] The simulation module uses the k-means algorithm to cluster the daily wind and solar power output and load data throughout the year, and selects typical days in each quarter for short-term operation simulation; this module is used to implement the function of step S20 in the above implementation, and the details are not repeated here;

[0302] An optimization module is used to construct a planning scheme evaluation model with the goal of minimizing operating costs, simulate the matching relationship between the output and load curves of various power types, optimize the peak-shaving capacity and economy of the power system based on the matching relationship, and optimize the investment and planning scheme of the power system; this module is used to implement the function of step S30 in the above-mentioned implementation 1, and the details are not repeated here;

[0303] The optimization solution module solves the complex constraints in the planning scheme evaluation model through the marine predator algorithm. If all planning schemes meet the system peak-shaving requirements, the planning results are considered reasonable. This module is used to implement the function of step S40 in the above-mentioned implementation one, and the details will not be repeated here.

[0304] Embodiment three:

[0305] This embodiment provides an electronic device having a computer program stored thereon, and when the computer program is executed by a processor, a method for evaluating the rationality of a power grid investment planning scheme taking into account a low-carbon target as described in any embodiment of the present invention is implemented.

[0306] Embodiment 4:

[0307] This embodiment provides a computer-readable medium for storing one or more programs. When the one or more programs are executed by the one or more processors, the one or more processors implement a method for evaluating the rationality of a power grid investment planning scheme that takes into account low-carbon goals as described in any embodiment of the present invention.

[0308] In the embodiments of the present application, "at least one" refers to one or more, and "more than one" refers to two or more. "And / or" describes the association relationship of associated objects, indicating that three relationships may exist. For example, A and / or B can represent the existence of A alone, the existence of A and B at the same time, and the existence of B alone. Among them, A and B can be singular or plural. The character " / " generally indicates that the previous and next associated objects are in an "or" relationship. "At least one of the following" and similar expressions refer to any combination of these items, including any combination of single or plural items. For example, at least one of a, b and c can be represented by: a, b, c, a and b, a and c, b and c, or a and b and c, where a, b, c can be single or multiple.

[0309] Those of ordinary skill in the art will appreciate that the various units and algorithm steps described in the embodiments disclosed herein can be implemented in a combination of electronic hardware, computer software, and electronic hardware. Whether these functions are performed in hardware or software depends on the specific application and design constraints of the technical solution. Professional and technical personnel can use different methods to implement the described functions for each specific application, but such implementation should not be considered to be beyond the scope of this application.

[0310] Those skilled in the art can clearly understand that, for the convenience and brevity of description, the specific working processes of the systems, devices and units described above can refer to the corresponding processes in the aforementioned method embodiments and will not be repeated here.

[0311] In several embodiments provided in the present application, if any function is implemented in the form of a software functional unit and sold or used as an independent product, it can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of the present application, or the part that contributes to the prior art or the part of the technical solution, can be embodied in the form of a software product, which is stored in a storage medium and includes several instructions for a computer device (which can be a personal computer, a server, or a network device, etc.) to perform all or part of the steps of the method described in each embodiment of the present application. The aforementioned storage medium includes: U disk, mobile hard disk, read-only memory (Read-Only Memory; hereinafter referred to as: ROM), random access memory (Random Access Memory; hereinafter referred to as: RAM), disk or optical disk, and other media that can store program codes.

[0312] The above descriptions are merely embodiments of the present invention and are not intended to limit the patent scope of the present invention. Any equivalent structure or equivalent process transformation made using the contents of the present invention specification and drawings, or directly or indirectly applied in other related technical fields, are also included in the patent protection scope of the present invention.

Claims

1. A method for evaluating the rationality of a power grid investment planning scheme considering a low-carbon goal, characterized in that: The following steps are involved: Collect relevant data and indicators related to power grid investment planning, divide the power grid load forecast area based on the relevant data and indicators, use the seasonal decomposition method of time series to decompose the power system load data into three parts: real-time variation, trend stability component and irregular fluctuation component, preprocess the power system load data to obtain a data set, divide the preprocessed data set into a training data set and a test data set, build a power grid load forecast model for different components, train and learn the load forecast model with the divided training data set, use the trained load forecast model to predict the samples in the test data set to obtain the load forecast result, and perform denormalization on the load forecast result to obtain the final load forecast result; Compare the actual load value and the final load forecast result to get the error between the two, evaluate the forecast model based on the error, and finally get the load assessment result; The k-means algorithm is used to cluster the daily wind and solar power output and load data throughout the year, and typical days in each quarter are selected for short-term operation simulation; Construct a planning scheme evaluation model with the goal of minimizing operating costs, simulate the matching relationship between the output and load curves of various power sources, optimize the peak-shaving capacity and economy of the power system based on the matching relationship, and optimize the investment and planning scheme of the power system; The complex constraints in the planning scheme evaluation model are solved by the marine predator algorithm. If all planning schemes meet the system peak-shaving requirements, the planning results are considered reasonable.

2. The method for evaluating the rationality of a power grid investment planning scheme considering a low-carbon goal according to claim 1 is characterized in that: The relevant data include: the statistical yearbook of the region, the national economic and social development plan, the overall urban and rural plan, the overall land use plan, the operation of the power grid, basic statistical data, and the planning results of the superior power grid; The indicators include: regional GDP, population, grid asset size, geographical area, and load density; Cluster analysis method is used to divide the grid load forecast area into regions by combining relevant data and indicators. The specific steps include: Assume c is the number of clusters, is a vector set in the feature space, where v i (i=1, 2...) represents the cluster prototype vector of the i-th class; μ ik ∈U, U is a c×n fuzzy matrix that satisfies the following conditions: The objective function of the improved fuzzy clustering algorithm is defined as: max{J(U,V)} in μ ik For data point x k The membership degree of cluster i; μ ik For data point x k The membership degree V belonging to cluster j is the clustering result; Given the number of clustering categories c (2≤c≤n, n is the number of data); set the iteration stop threshold ε and the fuzzy index m; initialize the clustering prototype model v (0) ; Set iteration counter b = 0; For any i, k, if there exists d ik b >0, then If there exists i, r such that d ik b = 0, then μ ir (b) =1, and for j≠i, μ ir (b) =0; If ||V (b+1) -V (b) ||<ε, the improved fuzzy clustering algorithm stops; and outputs the partition matrix U and cluster prototype V, otherwise b=b+1, and iterates to execute the improved fuzzy clustering algorithm; Determine the best classification tree C m ; Calculate the value of J(C). When the condition is met: J(Cm) = max{J(C)}, Cm is considered to be the optimal number of categories.

3. A method for evaluating the rationality of a power grid investment planning scheme considering low-carbon goals according to claim 1, characterized in that: The steps of decomposing the power system load data into three parts: real-time variation, trend stability component and irregular fluctuation component include: Decompose the original load sequence in the power system load data into subsequences; t It is expressed as the actual time series value of the t period. The time series seasonal trend adjustment method converts Y t Decomposed into three components: seasonal component, trend component, and irregular component; these three components are combined by an additive or multiplicative model as follows: The additive decomposition model is defined as: Y t =Trend t +Seasonal t +ιrregular t The multiplicative factorization model is defined as: AND t =Trend t ×Seasonal t ×Irregular t In the above two formulas, Trend t is the trend component in time period t, Seasonal t is the seasonal component in time period t, Irregular t is the irregular component in time period t; When the seasonal fluctuations are proportional to the level of the time series, a multiplicative model is used, and when the seasonal fluctuations do not depend on the level of the time series, an additive decomposition model is used; The feedforward neural network method is used to build prediction models for the above decomposition models respectively, and finally superimpose them to obtain the final load prediction result.

4. A method for evaluating the rationality of a power grid investment planning scheme considering low-carbon goals according to claim 1, characterized in that: The k-means algorithm is used to cluster the daily wind and solar power output and load data throughout the year, and a typical day in each quarter is selected for short-term operation simulation. The specific steps are as follows: Data collection and preprocessing: First, collect wind and solar power output data and grid load data for every day of the year. The wind and solar power output data includes wind speed corresponding to wind power generation, and sunshine duration and intensity corresponding to photovoltaic power generation. Clean the data, remove outliers, and standardize the data. Determine the number of clusters: Select the K value, which is predetermined by the elbow rule and silhouette coefficient. For power system data with obvious seasonality, the K value is set to 4, corresponding to the typical day of each quarter of the year; Initialization and iteration: The K-means algorithm starts by randomly selecting K center points and iterating the following steps until convergence: Assign each data point to the category of the nearest center point; recalculate the center of each category until the change in the center point is less than a preset threshold or the maximum number of iterations is reached; Typical day selection: The center point of each cluster represents a typical day in a quarter, which contains the typical characteristics of wind power, solar power generation and grid load in that quarter; typical day data is used for simulation and analysis to reduce the amount of data processing during actual simulation operation.

5. A method for evaluating the rationality of a power grid investment planning scheme considering low-carbon goals according to claim 1, characterized in that: The planning scheme evaluation model is constructed with the goal of minimizing the operating cost, including an objective function and constraints; The operation simulation layer takes the lowest operation cost as the objective function to verify the peak load regulation capability of the system; the operation cost F2 of the operation simulation layer includes three aspects: fuel cost and coal-fired power unit start-up and shutdown cost F 1t , Energy storage device charging, discharging and maintenance costs F 2t and interruptible load compensation and wind and solar curtailment penalty costs F 3t ; Specifically as shown in the following formula: Where: t is the number of hours, T is the total number of hours for short-term operation simulation; X RA , X CA , X KA are the number of fuel units, energy storage units, and renewable energy units participating in the short-term operation simulation; are the fuel cost, startup cost, and shutdown cost of the i-th fuel unit respectively; They are the actual output, startup state variable, and shutdown state variable of the i-th fuel unit at the t-th hour; are the unit electricity discharging cost, charging cost, and maintenance cost of the j-th energy storage device respectively; are the real-time discharge power, charging power, and actual output of the j-th energy storage device at the t-th hour; is the interruptible load; ρ IL , ka They are interruptible load compensation cost coefficient and wind and solar power abandonment factor respectively; are the available power and actual output of the kth energy storage device at the tth hour respectively; Mode The fuel cost is expressed as a quadratic function: Where: a i 、b i 、c i are the consumption coefficients of the i-th fuel unit respectively.

6. A method for evaluating the rationality of a power grid investment planning scheme considering low-carbon goals according to claim 5, characterized in that: The constraints specifically include: power balance constraints, fuel unit output constraints, unit start and stop time constraints, renewable energy output constraints, various power supply ramp constraints, system hot standby constraints and energy storage constraints; The power balance constraint is that in the power system, the output of all power sources and the consumption of all loads need to be balanced, as shown in the following formula: Where: are the active outputs of the fuel unit, energy storage unit, and renewable energy unit at the tth hour respectively; is the interruptible load of the system in the short-term operation simulation at the tth hour; is the load demand of the system at hour t; The fuel unit output constraint is that during the operation of the power system, the real-time output of the fuel unit needs to meet the upper and lower limit constraints, as shown in the following formula: Where: U i，t Indicates the start and stop status of the unit; are the minimum output and maximum output of the fuel unit respectively; is the actual output of the fuel unit at hour t; The start and stop time constraint of the unit is specifically shown in the following formula: Where: T A，i (t), T B，i (t) are the time that the i-th unit needs to keep running or shut down at the t-th hour, which is calculated as: Where: I A，i ,I B，i They are respectively the minimum operating interval between two shutdowns and the minimum shutdown interval between two starts of the unit; The specific step of the renewable energy output constraint is to set the maximum output limit of renewable energy, as shown in the following formula: in, The amount of available resources for renewable energy; The various power supply ramp constraints include the following aspects, as shown in the following formula: Wherein, R is the unit climbing rate; The system hot standby constraint is specifically shown in the following formula: Among them, X W The number of hydropower units participating in the short-term operation simulation; are the available resources and actual output of hydropower generation respectively; ρ is the hot standby coefficient; The energy storage constraint is to constrain the working state of the energy storage device in the short-term operation simulation to meet the system requirements and adjust the balance between the system load and the power supply. The specific upper and lower limits of the charge and discharge power are as follows: in, are the charging and discharging power of the energy storage device respectively; are the rated power of the energy storage device and the input power to the grid respectively; are the charging and discharging states of the energy storage device respectively; k is the margin coefficient; Working state constraints are as follows: The energy storage constraints for charging and discharging are as follows: Among them, E j，t is the actual storage capacity of the energy storage device at the tth hour; j is the energy self-consumption coefficient of the energy storage device; It is the energy conversion efficiency of the energy storage device during the charging and discharging process; are the maximum and minimum storage capacity of the energy storage device respectively; the energy storage device has an equal capacity constraint at the beginning and end of the short-term operation cycle, as shown in the following formula:

7. A method for evaluating the rationality of a power grid investment planning scheme considering low-carbon goals according to claim 1, characterized in that: The complex constraints in the planning scheme evaluation model are solved by the marine predator algorithm, including the high speed ratio stage, the medium speed ratio stage and the low speed ratio stage. The specific steps are as follows: Initialize the positions of predators and preys, construct an elite matrix with predators with the best fitness, and construct a prey matrix with evenly distributed preys; The high speed ratio stage is the initial stage of algorithm optimization iteration, and the global position information is surveyed. The mathematical model of this stage is expressed as follows: in, is the moving step length; is a random vector based on Brownian walk normal distribution; is a term-by-term multiplication operation; P is a constant, equal to 0.5; is a uniform random vector in [0, 1]; Iter is the current iteration number; Max_Iter is the maximum iteration number; N is the population size; The medium speed ratio stage is the middle stage of the algorithm iteration. The population is divided into two parts, in which the prey performs Lévy flight and is responsible for the development of the space, and the predator performs Brownian motion and is responsible for the exploration of the space. The mathematical model expression of this stage is as follows: in, represents the random vector of Levy motion, CF represents the adaptive parameter controlling the moving step length of the predator; The medium speed ratio stage is the late stage of algorithm iteration, and the predator adopts the Lévy flight strategy and pays more attention to the development of local areas. The mathematical model expression of this stage is as follows: External environmental factors such as fish aggregation devices and eddy effects change the foraging strategy of predators to jump out of local extreme values, so that the algorithm avoids premature convergence. The mathematical model expression is as follows: Among them, FADs is the impact probability, which is generally taken as 0.2; is a binary vector; r is a random number in [0, 1]; r1 and r2 are random indices of the prey matrix respectively; The ocean predator algorithm is used to solve the objective function under complex constraints such as coal-fired power output constraints and power balance constraints, and optimize the operation of the power grid, including the optimization of peak-shaving capacity and economy. After optimization, the gap between the actual peak-shaving capacity and the expected peak load demand is calculated. If the gap is lower than the pre-set threshold, the power planning and operation strategy is effective and can meet the system peak-shaving needs, and the power grid operation is safe and stable. Otherwise, it is necessary to readjust the optimization model parameters or increase the peak-shaving resources, and continue to use the ocean predator algorithm for optimization.

8. A rationality evaluation system for power grid investment planning scheme considering low-carbon goals, characterized in that: include: The load assessment module collects relevant data and indicators related to the power grid investment planning, divides the power grid load forecast area based on the relevant data and indicators, uses the seasonal decomposition method of time series to decompose the power system load data into three parts: real-time variation, trend stability component and irregular fluctuation component, preprocesses the power system load data to obtain a data set, divides the preprocessed data set into a training data set and a test data set, builds a power grid load forecast model for different components, trains and learns the load forecast model with the divided training data set, uses the trained load forecast model to predict the samples in the test data set, obtains the load forecast result, and performs denormalization on the load forecast result to obtain the final load forecast result; Compare the actual load value and the final load forecast result to get the error between the two, evaluate the forecast model based on the error, and finally get the load assessment result; The simulation module uses the k-means algorithm to cluster the daily wind and solar power output and load data throughout the year, and selects typical days in each quarter for short-term operation simulation; The optimization module builds a planning scheme evaluation model with the goal of minimizing operating costs, simulates the matching relationship between the output and load curves of various power sources, optimizes the peak-shaving capacity and economy of the power system based on the matching relationship, and optimizes the investment and planning scheme of the power system; The optimization solution module solves the complex constraints in the planning scheme evaluation model through the marine predator algorithm. If all planning schemes meet the system peak-shaving requirements, the planning results are considered reasonable.

9. An electronic device comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, characterized in that: When the processor executes the program, a method for evaluating the rationality of a power grid investment planning scheme taking into account low-carbon goals is implemented as described in any one of claims 1 to 7.

10. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by a processor, it implements a method for evaluating the rationality of a power grid investment planning scheme taking into account a low-carbon target as described in any one of claims 1 to 7.

Images