Novel electric power system power supply planning method based on reliability evaluation feasible region

Abstract

The invention relates to a novel power system power supply planning method based on a reliability evaluation feasible region. In order to solve the problem that time sequence reliability simulation and power supply switching decision modeling optimization are not compatible in mathematical structure in the existing novel power system power supply planning, the method comprises the following steps: selecting a characteristic power supply type for each level year, and generating a characteristic power supply potential operable quantity decision sample set with a reliability evaluation label; based on the decision sample set, utilizing a recursion strategy of an embedded L-BFGS algorithm to train an oblique decision tree model, utilizing a depth-first search strategy to extract a reliability evaluation feasible region, and expressing the reliability evaluation feasible region in a logic or constraint form; and a convex hull modeling method is utilized to convert logic or constraint into mixed integer linear constraint, the mixed integer linear constraint is integrated into the new and old power supply switching decision and unit combination integrated optimization model, a novel electric power system power supply planning model embedded with a reliability evaluation feasible region is formed, and then novel electric power system power supply planning is realized based on the model.

Description

Technical Field

[0001] The present invention belongs to the intersection of new power system planning and machine learning, and specifically relates to a new power system power supply planning method. Background Art

[0002] The power sources of the new power system generally include conventional power sources and wind and solar power sources. Conventional power sources mainly refer to generators that use fossil fuels as primary energy, and wind and solar power sources mainly refer to wind farms and photovoltaic power stations. In the evolution of the new power system, the installed capacity and proportion of wind and solar new energy will grow rapidly year by year. Due to the large uncertainty of wind and solar power generation capacity, wind and solar power sources will gradually show the characteristics of "large installed capacity and small power". In the context of continued growth in electricity demand, if the construction of conventional power sources is blindly and excessively slowed down and the retirement of old power sources is radically accelerated, the new power system will face great challenges in supporting the supply of loads at all levels and all time scales. Current operating experience shows that in some provincial power systems with a high proportion of new energy, the difficulty of reliable power supply has shown a significant increase, and the overall situation of power supply and demand is changing from "tense in local areas and local time periods" to "comprehensive tension in multiple time periods of the entire network." Moreover, the frequent forced outages of conventional power sources and wind and solar power sources, the frequent occurrence of extreme operating scenarios such as extremely hot weather with little or no wind (often with extremely high evening peak loads and photovoltaic output close to zero), cloudy and rainy weather with a lack of light or no light, will further increase the difficulty of the new power system in maintaining a balance between supply and demand.

[0003] Given the medium- and long-term forecast of load demand, how to fully consider the information of unexpected forced shutdown of power supply, the uncertainty of wind and solar power output, and the information of extreme weather scenarios such as extremely low wind and solar output and extremely high evening peak load, and find the new and old power investment and withdrawal schemes that save the comprehensive investment and operation costs as much as possible, while ensuring that the power generation capacity in each level year in the future meets the requirements of power supply reliability indicators (the power supply to be decided and built in the planning area in each level year is called new power supply, and the power supply that already existed before the beginning of the planning period is called old power supply) is a key problem that must be solved in the construction of a new power system. Solving this problem involves the coordination and balance between the economic efficiency of power investment and operation and the reliability of power supply, which belongs to the field of new power system planning.

[0004] Different from traditional power planning, the power planning of new power systems must fully consider the temporal and spatial variation characteristics of wind power and photovoltaic power output. Therefore, it is usually modeled as an integrated optimization model that combines the investment and withdrawal decision of new and old power sources and the combination of units, in which the unit combination process includes not only conventional power sources, but also wind and solar power sources (wind farms and photovoltaic power stations). The integrated optimization model of investment and withdrawal decision of new and old power sources and unit combination generally uses the reserve margin index to measure the adequacy and reliability level of power generation capacity. By setting a certain margin in the power generation capacity constraint, it is ensured that the sum of the installed capacity of conventional power sources and the reliable capacity of new energy in each level year exceeds the reserve margin of a certain amount of the system peak load of the current year. The method of using capacity constraints has the advantages of being simple, intuitive and easy to implement, but it does not utilize the random fault outage probability information of different types of power sources, including the intermittent and volatile wind and solar power generation, and the historical time series information of 8760 hours (or longer time span, finer time scale) of new energy power generation and load in different extreme operating scenarios such as extremely low output and extremely high peak load. It is not based on strict mathematical statistics and probability theory, resulting in that the power planning scheme that meets the reserve margin index may not necessarily meet the reliability index requirements strictly based on mathematical statistics and probability theory (called: probabilistic reliability index, such as: annual expected load loss time index, Loss of Load Hour, abbreviated as: LOLH). In order to ensure the reliable power supply of the new power system, the new and old power supply investment and withdrawal schemes obtained based on the reserve margin index usually have a lot of redundant capacity in some normal years, which will cause a considerable waste of valuable resources and push up the comprehensive cost of power investment and operation and maintenance during the entire planning period.

[0005] Compared with the reserve margin index, the probabilistic reliability index can make full use of the historical time series information of random power failure outages and wind and solar power generation and load of 8760 hours (or longer time spans and finer time scales), so it has more practical reference significance for measuring the power generation reliability of new power systems. If the probabilistic reliability index requirements can be made into constraints of the power planning model of the new power system, then focusing on solving the model can find out the new and old power supply investment and withdrawal schemes that can not only meet the index requirements but also save the comprehensive investment and operation and maintenance costs and installed resources as much as possible. However, for the new power system, probabilistic reliability indicators such as LOLH are generally obtained through daily and hourly power generation reliability assessment procedures for each horizontal year. This is essentially an ultra-long time series simulation process, and the power planning problem is usually modeled as an ultra-large-scale optimization model. The two are incompatible in mathematical structure. If the probabilistic reliability index constraints are simply and crudely embedded in the integrated optimization model of the new and old power supply investment and withdrawal decision and unit combination, it will inevitably destroy the original mathematical structure of the model, resulting in the inability to solve the model. This is the main challenge faced by the new power system power planning to seamlessly compatibility with the probabilistic reliability index requirements. Summary of the invention

[0006] The present invention aims to solve the problem of incompatibility between timing reliability simulation and power supply decision modeling optimization in mathematical structure in the existing new power system power supply planning.

[0007] 1. A new power system power planning method based on reliability assessment feasible region, characterized by comprising the following steps:

[0008] Step 1: Generate a sample data set for oblique decision tree training for each level year of the planning period. The specific process includes:

[0009] Step 1.1: For the integrated optimization model of new and old power supply investment and withdrawal decision and unit combination, different reserve margin levels are set, and the model is solved to obtain the optimal decision values ​​of the potential operable number of new and old power supplies that meet the LOLH index, including:

[0010] Step 1.1.1: In the integrated optimization model of new and old power supply investment and decommissioning decision and unit combination, the reserve margin of each level year is calculated. Set it to zero, use Benders decomposition optimization technology to solve the integrated optimization model of new and old power supply investment and withdrawal decision and unit combination, and obtain the decision value of the potential operable number of different types of new and old power supplies in each horizontal year;

[0011] New power sources refer to power sources that are to be decided and constructed in each year, and old power sources refer to power sources that existed before the start of the planning period. The names of all types of new power sources are grouped into a set. Group the names of all types of legacy power supplies into one set

[0012] The meaning of the potential operational quantities of new and old power supplies is as follows:

[0013] For a certain horizontal year, the sum of the cumulative number of new power sources of a certain type invested and constructed in that year and the previous years is called the potential operable number of such new power sources in that horizontal year; the sum of the cumulative number of old power sources of a certain type retired in that year and the previous years is called the cumulative number of old power sources retired in that horizontal year, and the difference between the initial number of old power sources and the cumulative number of retired power sources is called the potential operable number of such old power sources in that horizontal year;

[0014] Step 1.1.2: For each horizontal year in the planning period, based on the decision values ​​of the potential operable number of different types of new and old power sources obtained in step 1.1.1, the capacity information of conventional power sources, wind and solar power sources, the random failure and outage probability information of conventional power sources, wind and solar power sources, the wind and solar power generation capacity coefficient and the per unit load H in the planned area are used to calculate the decision value of the potential operable number of new and old power sources of different types. L Hourly historical time series data and annual peak load forecast data are used to evaluate the power generation reliability index LOLH corresponding to the horizontal year;

[0015] Step 1.1.3: Set the benchmark step lengths from small to large. satisfy For each benchmark step, the reserve margin of each level year is gradually increased from zero, and the model is re-solved using the Benders decomposition optimization technology to obtain the potential operable number of different types of power sources in each level year, and the LOLH index of the power system to be expanded in each level year is evaluated. This process is repeated until the LOLH index of all years meets the reliability requirements, and the decision value of the potential operable number of different types of power sources in each level year is saved;

[0016] Step 1.2: For each horizontal year, based on the decision values ​​of the potential operable quantities of different types of new and old power sources in each horizontal year obtained in step 1.1.3, the decision value matrices of the potential operable quantities of new and old power sources are constructed respectively, with the power source type information as rows and the benchmark step information as columns; based on the decision value matrices of the potential operable quantities of new and old power sources, the name sets corresponding to each type of new and old power sources whose decision values ​​of the potential operable quantities change with the change of the backup margin benchmark step are recorded as Will The corresponding new power sources will be used as the characteristic new power sources of the year. The corresponding types of old power sources are used as the characteristic old power sources of the year, and the potential operable quantity of characteristic new and old power sources is calculated. As a feature variable of oblique decision tree;

[0017] At the same time, the type belongs to But not The power supply of this year is regarded as a non-characteristic new power supply, and the type belongs to But not The power supply is regarded as the non-characteristic old power supply, and the names of the non-characteristic new power supplies are combined into a set and recorded as The names of non-characteristic old power supplies are combined into a set and recorded as

[0018] Step 1.3: For each level year t, generate a decision sample of the potential operable quantity of characteristic power sources;

[0019] Step 1.4: For each horizontal year t, based on the generated value data of the potential operable quantity of characteristic power sources and the decision value data of the potential operable quantity of non-characteristic power sources corresponding to each decision sample, the corresponding LOLH index is evaluated, and the label value of each decision sample is determined according to the relationship between the LOLH index and the threshold value, and together with the corresponding decision sample, a labeled decision sample set D of the potential operable quantity of characteristic power sources is formed. t ;

[0020] Step 2: For each horizontal year, use the labeled decision sample set of the potential operational quantity of the characteristic power source and the decision variable of the potential operational quantity of the characteristic power source as the characteristic variable to train the oblique decision tree model of the horizontal year;

[0021] Step 3: Based on the oblique decision tree model of each level year, recursively extract the feasible domain of reliability assessment for each year;

[0022] Step 4: Based on the reliability assessment feasible domain of each year in step 3, construct the embedded H L A new power system power planning model based on the feasible domain of hourly reliability assessment;

[0023] Step 5: Use Benders decomposition technique to solve the embedded H L An integrated optimization model of new and old power source investment and withdrawal decision-making and unit combination in the feasible domain of hourly reliability assessment is used to obtain new and old power source investment and withdrawal plans that meet the reliability requirements at each level of the planning period and have the lowest comprehensive investment and operation and maintenance costs.

[0024] The beneficial effects of the present invention are as follows:

[0025] (1) Use oblique decision trees to construct H L The feasible domain of reliability assessment of hours solves the problem of incompatibility between timing reliability simulation and new power system power planning modeling optimization in mathematical structure without destroying the solution structure of the integrated optimization model of new and old power supply investment and withdrawal decision-making and unit combination.

[0026] (2) The constructed reliability assessment feasible domain can fully consider the impact of different extreme operating scenarios on maintaining the supply and demand balance of the new power system, such as forced outage of power supply, intermittent and fluctuating changes in wind and solar power output, extremely hot and windless or no wind, cloudy and rainy weather with little or no light, and extremely hot and high evening peak load (photovoltaic output is close to zero).

[0027] (3) It has good scalability and is compatible with random time series production operation simulation processes with various time spans (from one year to multiple years, etc.) and different time resolutions (60 minutes, 30 minutes, 15 minutes, etc.). Therefore, it is convenient to consider the impact of forced shutdown failures of scheduling resources such as energy storage, power-to-hydrogen, and hydropower on power generation reliability.

[0028] (4) Under the condition of known medium- and long-term regional load forecasts and wind and solar resource output characteristics, it is possible to obtain new and old power supply commissioning and decommissioning plans that meet the power generation reliability index requirements, save costs as much as possible, and reduce power supply redundant capacity in the next few years to decades, providing important technical support for the transition of power systems in different regions to new power systems. BRIEF DESCRIPTION OF THE DRAWINGS

[0029] Figure 1The working flow chart of the novel power system power planning method based on reliability assessment feasible region;

[0030] Figure 2 Actual capacity margin diagrams of the feasible region method and the reserve margin method for reliability assessment of Case 1;

[0031] Figure 3 Investment cost diagrams for the feasible region method and the reserve margin method for reliability assessment of Case 1;

[0032] Figure 4 Operation and maintenance cost diagrams for the feasible region method and the reserve margin method for reliability assessment of Case 1;

[0033] Figure 5 This is the power installed capacity diagram for Case 1;

[0034] Figure 6 This is the power output diagram under the condition of N-3 on the peak load day of the 20th year in Case 1;

[0035] Figure 7 This is the change diagram of actual capacity margin as the characteristic power source type and quantity are expanded in Case 2;

[0036] Figure 8 This is the total investment cost of Case 2 as the type and quantity of characteristic power sources expand;

[0037] Fig. 9 This is the total operation and maintenance cost of Case 2 as the type and quantity of characteristic power sources are expanded;

[0038] Fig.10 This is a graph showing how the comprehensive investment and operation and maintenance costs for Case 2 change with the expansion of the type and number of characteristic units. DETAILED DESCRIPTION

[0039] The present invention is aimed at a new type of power system, and aims to propose a method for constructing H L A power planning method is proposed that evaluates the feasible domain of hourly power generation reliability, embeds it into the integrated optimization model of new and old power supply investment and withdrawal decision-making and unit combination, and then solves it through decomposition and coordination optimization technology.

[0040] Specific implementation method 1: Combination Figure 1 Description of this embodiment

[0041] The novel power system power supply planning method based on reliability assessment feasible region described in this embodiment includes the following steps:

[0042] Step 1: Generate a sample data set for oblique decision tree training for each level year of the planning period. The specific steps are as follows:

[0043] Step 1.1: For the integrated optimization model of new and old power supply investment and withdrawal decision and unit combination, different reserve margin levels are set, and the model is solved to obtain the optimal decision value of the potential operable number of new and old power supplies that meet the LOLH index. The specific process is as follows:

[0044] Step 1.1.1: In the integrated optimization model of new and old power supply decision-making and unit combination, the reserve margin in the t-th level year is It means that the reserve margin of each horizontal year is set to zero, and the model is solved by using Benders decomposition optimization technology to obtain the decision values ​​of the potential operable number of different types of new and old power sources in each horizontal year.

[0045] Note: The power sources to be decided for construction in each horizontal year are called new power sources, and the power sources that already existed before the start of the planning period are called old power sources. New (old) power sources with the same primary energy type, prime mover type and single-unit capacity are classified as the same type of new (old) power sources. For a certain horizontal year, the sum of the cumulative number of new power sources of a certain type invested and constructed in that year and the previous years is called the potential operable number of such new power sources in that horizontal year; the sum of the cumulative number of old power sources of a certain type retired in that year and the previous years is called the cumulative number of old power sources retired in that horizontal year, and the difference between the initial number of old power sources and the cumulative number of retired old power sources is called the potential operable number of such old power sources in that horizontal year. The potential operable number of new and old power sources in each horizontal year can be obtained by solving the integrated optimization model of new and old power source investment and retirement decision and unit combination. The specific value obtained by solving the model is called the decision value of the potential operable number of new and old power sources.

[0046] Horizontal Year Number is the set of horizontal years within the planning period, T is the total number of horizontal years in the planning period, and the reserve margin of each horizontal year is set to zero, that is, For the integrated optimization model of new and old power supply investment and withdrawal decision and unit combination, please refer to the literature [C.Li, AJConejo, P.Liu, BPomell, JDSiirola, and I.E.Grossmann, "Mixed-integer linear programming models and algorithms for generation and transmission expansion planning of power systems," European Journal of Operational Research, vol. 297, no. 3, pp. 1071-1082, 2022, doi: 10.1016 / j.ejor.2021.06.024.]. Considering that the actual regional power system is generally divided into several sub-regional systems, and safe and economic operation is achieved through coordinated dispatch, this model divides the area to be planned into several sub-regions. Due to resource endowment and other reasons, some new power sources (such as wind and solar power sources) may only be invested and built in some local sub-regions in the area to be planned, and some old power sources may only exist in some local sub-regions. In order to ensure the universality of the planning method, new and old power sources are usually decided by sub-region.

[0047] For the convenience of discussion, each type of new power supply is given a name, and the names of all types of new power supplies are grouped into a set, recorded as The number of names in is the number of new power types, recorded as Will Each new power supply is numbered starting from 1 until Finish; The name of the new power supply of type i is recorded as The names of each sub-region where the i-th type of new power source can be built are grouped into a set, denoted as The number of names in is the number of sub-regions, denoted by Will The sub-areas in the Finish; The name of the rth sub-region in The potential operable number of new power sources of type i in sub-region r in the t-th horizontal year is recorded as

[0048] Similarly, give each type of old power supply a name, and form a set of names of all types of old power supplies, recorded as The number of names in is the number of types of old power supplies, recorded as Will Each new power supply is numbered starting from 1 until Finish; The name of the i-th type of old power supply is recorded as The names of each sub-area where the i-th type of old power source is located are grouped into a set, denoted as The number of names in is the number of sub-regions, denoted by Will The sub-areas in the Finish; The name of the rth sub-region in The potential operable quantity of the i-th type of old power source in the r-th sub-region in the t-th horizontal year is recorded as and It is a part of the decision variables for the integrated optimization of new and old power supply investment and withdrawal decisions and unit combination.

[0049] Step 1.1.1 Solve the model to obtain and The decision value of and The decision value of the potential operable quantity of the i-th type of new power source in the t-th level year is Decision value of the potential operable quantity of the i-th type of old power source in the t-th level year

[0050] The main problem model solved by Benders decomposition optimization is the decision model for the investment and withdrawal of new and old power sources, which is solved by calling the mixed integer linear programming module of any commercial solver; the sub-problem model is the unit combination model in which integer variables are relaxed into continuous variables, which is solved by calling the linear programming module of any commercial solver; the main problem model, sub-problem model and Benders decomposition solution process can be found in the literature [P.Liu, L.Cheng, J.Zhang, and J.Yu,"Customized BendersDecomposition for Unit Commitment Integrated Generation Expansion Planning," presented at the 2023International Conference on Power System Technology(PowerCon),2023.].

[0051] Step 1.1.2: For each horizontal year in the planning period, based on the decision values ​​of the potential number of new and old power sources of different types obtained in step 1.1.1, the capacity information of conventional power sources, wind and solar power sources, the random failure and outage probability information of conventional power sources, wind and solar power sources, the wind and solar power generation capacity coefficient and the per unit load H in the planned area are used to determine the decision value of the potential number of new and old power sources of different types. L Hourly historical time series data and annual peak load forecast data are used, and the literature [L.Cheng, P.Liu, D.Cheng, J.Zhang, J.Yu, "TuningReserveMargins in Long Term Unit Commitment Integrated Generation ExpansionPlanning to EnhanceCapacityAdequacy," presented at the 2023 3 rd New Energy and Energy Storage System Control Summit Forum (NEESSC), 2023.] method is used to evaluate the power generation reliability index LOLH corresponding to this level year.

[0052] The specific steps are as follows:

[0053] (1): Based on the decision values ​​of the potential operable quantities of different types of new and old power sources obtained in step 1.1.1, the decision value information of the potential operable quantities of conventional power sources, wind power sources, and photovoltaic power sources are respectively counted.

[0054] (2) Using the decision value information of the potential operable quantity of conventional power sources and wind and solar power sources, the capacity information of each conventional power source (each generator set), each wind power source (each wind farm), and each photovoltaic power source (each photovoltaic power station), and the random fault outage probability information, the power outage capacity probability distribution tables of conventional power sources, wind power sources, and photovoltaic power sources in the year to be evaluated are generated respectively.

[0055] The steps for generating the power outage capacity probability distribution table of conventional power sources, wind power sources and photovoltaic power sources are the same. The present invention takes conventional power sources as an example, and the specific steps include:

[0056] 1) Set the capacity step Δ con (The smaller the value, the more detailed the power outage capacity probability table is, but the longer it takes to generate it), input the decision value information of the potential operable number of conventional power sources in the year to be evaluated, the capacity information of each conventional power source and the set of random forced power outage probability information Among them C n and r n is the capacity of the nth conventional power source and the probability of random forced power outage, N conIt is the sum of the decision values ​​of the potential operable quantity of all types of conventional power sources in the year to be evaluated;

[0057] 2) Generate the probability distribution table of the power outage capacity of the first power source, p(0) = 1-r1, p(C1) = r1, the power source number is set to n = 1, and the current maximum power outage capacity C max =C1, the power outage probability when the current power outage capacity is X is recorded as p(X);

[0058] 3) If the set G con If there is a power source that has not participated in generating the power outage capacity probability distribution table, let n=n+1 and execute step 4); otherwise, obtain the probability distribution table corresponding to different power outage capacities;

[0059] 4) From the set G con Extract the forced power outage probability and unit capacity information of the nth power source from max =C max +C n , with Δ con As the step length, set the power outage capacity X = m·Δ con , increase m from 0 to C max / Δ con , calculate the power outage capacity X = m·Δ con The probability p′(X)=(1-r n )·p(X)+r n ·p(XC n ), use p′(X) to update the current power outage capacity probability distribution table and execute step 3).

[0060] (3): Loop through all the outage capacities and their corresponding probabilities in the outage capacity probability distribution table of conventional power sources, wind power sources and photovoltaic power sources, and calculate the annual expected load loss time index LOLH for each level year according to formula (1).

[0061]

[0062] Where: h represents hours; Θ con , Θ wind , Θ solar It is a set of all possible power outage capacity data of conventional power sources, wind power sources and photovoltaic power sources, obtained from their respective power outage probability distribution tables; X con , X wind and X solar Represents Θ con , Θ wind and θ solar A specific outage capacity of conventional power sources, wind power sources and photovoltaic power sources in the collection; P(X con )、P(X wind ) and P(Xsolar ) is the power outage probability corresponding to the corresponding power outage capacity, obtained from the respective power outage probability distribution table; is the annual peak load forecast value in the t-th horizontal year; L h , and are the unit load, wind power and photovoltaic power capacity factors in the hth hour respectively; T is the total number of horizontal years in the planning period.

[0063] is the indicator function of whether to cut the load. If the inequality in the square brackets holds, it means that the load value of the hth hour in the year to be evaluated (i.e. the tth horizontal year) exceeds the power generation capacity of the system power supply to be expanded, and the excess load of the hour must be cut, so the function value is 1; otherwise, the function value is 0.

[0064] Step 1.1.3: Set the benchmark step lengths from small to large. satisfy For each benchmark step, the reserve margin of each level year is gradually increased from zero, and the model is re-solved using the Benders decomposition optimization technology to obtain the potential operable number of different types of power sources in each level year, and the LOLH index of the power system to be expanded in each level year is evaluated. This process is repeated until the LOLH index of all years meets the reliability requirements, and the decision value of the potential operable number of different types of power sources in each level year is saved. In the process of gradually increasing the reserve margin, the step size is based on the literature [L.Cheng, P.Liu, D.Cheng, J.Zhang, J.Yu, "Tuning ReserveMargins in Long Term Unit Commitment Integrated Generation Expansion Planningto Enhance Capacity Adequacy," presented at the 2023 3 rd New Energy and EnergyStorage System Control Summit Forum (NEESSC), 2023.] Formula (9) is adaptively adjusted.

[0065] Note: In step 1.1.3, multiple benchmark step values ​​are set. By solving the model, multiple groups of reserve margin cumulative values ​​for each level year and multiple groups of decision values ​​for the potential number of new and old power sources of different types that can be operated can be obtained. The purpose is to determine whether the decision to put different types of power sources into or out of service changes with the change of reserve margin level. In order to minimize the measurement and calculation of the implementation process, it is generally sufficient to generate 5-15 groups of power planning solutions.

[0066] The backup margin reference step size is Under the condition of , the decision value vector of the potential operable quantity of various new and old power sources in the t-th level year is recorded as: is a vector of decision values ​​of the potential number of new and old power sources that can be operated. is the decision value of the potential number of new power sources of type i that can be operated, is the decision value of the potential number of old power sources of type i that can be operated,

[0067] and It can be obtained by solving the integrated optimization model of new and old power supply investment and withdrawal decision and unit combination, specifically: is the backup margin reference step length Under the condition, by dynamically increasing the reserve margin level of each level year and finally making the LOLH index of each level year meet the requirements, the decision value of the potential operable quantity of the i-th type of new power source in the r-th sub-area is obtained by solving the integrated optimization model of the new and old power source investment and withdrawal decision and unit combination. is the decision value of the potential operable quantity of the i-th type of old power source in the r-th sub-area.

[0068] Step 1.2: For each horizontal year, determine the characteristic variables of the oblique decision tree based on the decision information of multiple groups of potential operable quantities of new and old power sources obtained in step 1.1.3. The specific steps are as follows:

[0069] Step 1.2.1: Step at different reserve margin benchmarks Under these conditions, the decision values ​​of the potential operable number of new and old power sources of different types obtained in step 1.1.3 for each horizontal year (t∈{1,…,T}, T is the total number of horizontal years in the planning period) are used to construct the decision matrix shown in equations (2)-(3) with power type information as rows and benchmark step information as columns:

[0070]

[0071]

[0072] Note: Both equations (2) and (3) contain T matrices, one for each horizontal year. In equation (2), the i-th row of the t-th matrix represents the decision value of the potential operable number of the i-th type of new power source in the t-th horizontal year under different backup margin reference step conditions. In equation (3), the i-th row of the t-th matrix represents the decision value of the potential operable number of the i-th type of old power source in the t-th horizontal year under different backup margin reference step conditions. Since the cumulative amount of backup margin in each horizontal year can be determined by the reference step in step 1.1.3, the change law of the decision value of the potential operable number of each type of new and old power source in equations (2) and (3) is jointly affected by the change of the cumulative amount of backup margin in the current and previous years under different reference step conditions.

[0073] Step 1.2.2: For each new power source Formula (2) is used to select the new power type whose decision value of the potential operable quantity in a certain or certain level year changes with the change of the backup margin benchmark step size; for each type of old power source Formula (3) is used to select the old power types whose decision values ​​of potential operable quantity in one or some horizontal years change with the change of the reserve margin benchmark step.

[0074] For the convenience of discussion, the names of each type of new power source whose decision value of the potential operable quantity changes with the change of the reserve margin benchmark step in a certain or certain level year are grouped into a set and recorded as The number of names in is equal to the number of power types, recorded as The names of various types of old power sources whose decision values ​​of the potential operable quantity in a certain or certain level year change with the change of the backup margin benchmark step are formed into a set, which is recorded as The number of names in is equal to the number of power types, recorded as

[0075] Note: After screening in step 1.2.2, the remaining power sources can still be divided into new power sources and old power sources. The reasons why their potential operable quantities do not change with the change of the reserve margin benchmark step are different. Since the comprehensive cost of investment and operation and maintenance of new power sources is very expensive, when the cumulative reserve margin of each level increases (as the reserve margin benchmark step increases, the reserve margin is gradually accumulated through step 1.1.3 to finally make the system power generation capacity meet the LOLH index, the cumulative reserve margin of each level shows a gradual increase as a whole), this part of the new power will not be included in the investment portfolio, and the decision value of its potential operable quantity is always zero. As for the old power source, because the life extension cost is low, when the cumulative reserve margin increases, careful maintenance to keep it in a potential operable state is conducive to reducing the overall investment and operation and maintenance costs of the entire planning period. It can be seen that the dominant factor in deciding whether these new and old power sources are put into operation or retired is their comprehensive investment and operation and maintenance costs. Considering the random forced shutdown factor for these units will not significantly change their investment and retirement decisions.

[0076] The potential operable number of different types of power sources obtained in step 1.2.2 is sensitive to the changes in the cumulative reserve margin of each level. The investment and withdrawal decisions of these sensitive power sources are not only related to their investment and operation costs, but also depend to a considerable extent on factors related to power generation reliability. Therefore, considering the random forced outage factor will generally cause a more obvious change in the decision of the potential operable number of these power sources, thereby creating a more considerable optimization space for further saving power installed capacity resources and investment and operation costs. To this end, the potential operable number of different types of power sources selected in step 1.2.2 is used as a feature variable participating in the oblique decision tree training process, as described in step 1.2.3.

[0077] Step 1.2.3: For each horizontal year of the planning period Will The corresponding new power sources will be used as the characteristic new power sources of the year. The corresponding types of old power supplies are regarded as the characteristic old power supplies of that year, and the types belonging to But not The power supply of this year is regarded as a non-characteristic new power supply, and the type belongs to But not The power supply is taken as the non-characteristic old power supply, and the potential operable number of characteristic new and old power supplies is taken as the characteristic variable of the oblique decision tree.

[0078] For the convenience of discussion, the names of non-characteristic new power sources are grouped into a set and recorded as The names of non-characteristic old power supplies are grouped into a set, denoted as The set obtained in step 1.2.2 and Represent the characteristic new power supply name set and the characteristic old power supply name set respectively, and Represent the number of characteristic new power types and the number of characteristic old power types respectively; the potential operable number of characteristic new and old power sources is composed into vectors, which are recorded as and is the potential number of new power sources of the i-th type (as one of the characteristic variables of the oblique decision tree), is the name of the new power source with the i-th characteristic, is the potential operable quantity of the old power source of the i-th type (as one of the characteristic variables of the oblique decision tree), is the name of the old power source with the i-th characteristic,

[0079] Step 1.3: For each horizontal year t∈{1,…,T} in the planning period, generate a large number of decision samples of the potential operable quantity of characteristic power sources. The specific steps are as follows:

[0080] Step 1.3.1: For each horizontal year t∈{1,…,T} of the planning period, traverse each row of the decision value matrix of the potential operable quantity of new and old power sources constructed in step 1.2.1 (Equations (2) and (3)), and calculate the upper and lower bounds of the potential operable quantity decision value of each type of characteristic new power source and characteristic old power source in the t-th horizontal year, as shown in Equations (4)-(7).

[0081]

[0082]

[0083] Where: are the upper and lower bounds of the decision value of the potential operable quantity of the i-th type of characteristic new power source, are the upper and lower bounds of the decision value of the potential operable quantity of old power sources with the i-th characteristic.

[0084] Step 1.3.2: Based on the upper and lower bounds of the decision values ​​of the potential operable quantities of various types of characteristic new and old power sources in step 1.3.1, the quantity of characteristic old power sources that already existed in the initial year of the planning period, and the remaining life of characteristic old power sources in each horizontal year of the planning period, set the decision interval for the potential operable quantity of characteristic power sources.

[0085] For the convenience of discussion, the decision interval corresponding to the i-th characteristic new power source in the t-th level year is recorded as The decision interval corresponding to the i-th type of characteristic old power supply is recorded as

[0086] In order to make the oblique decision tree training sample data fully reflect the changing law of power generation reliability indicators under the conditions of different characteristic power source potential operating quantities, the decision interval of the characteristic new and old power source potential operating quantities satisfies equations (8)-(11):

[0087]

[0088] In addition, the upper limit of the potential operable number of characteristic old power sources should not be greater than the number that already existed in the initial year, and the lower limit should not be less than the difference between the number that already existed in the initial year and the number that has reached its service life by the t-th horizontal year, that is, satisfying equations (12) and (13):

[0089]

[0090] Where: is the number of old power sources with the i-th characteristic existing in the initial year of the planning period, It is the number of old power sources with characteristics of type i that have reached the end of their service life in the kth horizontal year (k=1,…,t).

[0091] Note: The decision interval setting for the potential number of operating characteristic power sources in step 1.3.2 is flexible. Under the conditions of (8)-(13), the wider the decision interval setting, the more decision samples are generated, the stronger the generalization ability of the oblique decision tree trained with it, and the more accurate the feasible domain of reliability assessment extracted by the oblique decision tree.

[0092] Step 1.3.3: For each horizontal year, within the decision interval of the potential operable quantity of each type of characteristic power source, with a single unit number of 1 as a step, a large number of decision samples containing information on the potential operable quantity of each type of characteristic power source are generated by cyclically enumerating the potential operable quantity of each type of characteristic power source (called the generated value of the potential operable quantity of characteristic power sources), and a decision sample set of the potential operable quantity of characteristic power sources is constructed.

[0093] Note: This step targets the characteristic power source and obtains its potential operable quantity through cyclic enumeration, which is called the generated value of the potential operable quantity of the characteristic power source. Its meaning is different from the meaning of the decision value in step 1.1.3. In step 1.1.3, the decision value of the potential operable quantity of the new and old power sources is a specific value obtained by solving the integrated optimization model of the new and old power source investment and withdrawal decision and unit combination. After executing steps 1.2.2 and 1.2.3, the new (old) power source is divided into the characteristic new (old) power source and the non-characteristic new (old) power source. The decision value of the potential operable quantity of the new (old) power source can be specifically subdivided into the decision value of the potential operable quantity of the characteristic new (old) power source and the decision value of the potential operable quantity of the non-characteristic new (old) power source. Step 1.3.3 generates the potential operable quantity of the characteristic power source (the generated value of the potential operable quantity of the characteristic power source) for the purpose of constructing the decision samples for the training of the oblique decision tree, and based on step 1.1.3, multiple groups of decision information on the potential operable quantity of the new and old power sources are obtained for the purpose of determining the characteristic power sources and characteristic variables involved in the training of the oblique decision tree.

[0094] The decision sample set of the potential operable quantity of characteristic power sources in the t-th horizontal year is: t∈{1,…,T},M t is the total number of decision samples, and the mth sample is recorded as: is a vector of generated values ​​of the potential number of new power sources that can be run for all features in the mth sample, is the generated value of the potential number of new power sources with the i-th characteristic in the m-th sample that can be operated, Generate a vector of values ​​for the number of potentially operational old power sources for all features in the mth sample, is the generated value of the potential operable number of old power supplies with the i-th characteristic in the m-th sample.

[0095] Step 1.4: For each horizontal year t∈{1,…,T}, based on the potential operable quantity generation value data of the characteristic power source corresponding to each decision sample and the potential operable quantity decision value data of the non-characteristic power source in step 1.1.3 (although the non-characteristic power source is defined in step 1.2.3, the actual data used is the data in step 1.1.3, that is, the non-characteristic power source corresponds to the data in step 1.1.3), the corresponding LOLH indicator is evaluated using the process of step 1.1.2 to determine the label value of each decision sample, and together with the corresponding decision sample, it forms a labeled characteristic power source potential operable quantity decision sample set.

[0096] Note: The LOLH label of the decision sample for evaluating the potential operable quantity of characteristic power sources must take into account the impact of the potential operable quantity of non-characteristic power sources. Since the decision value of the potential operable quantity of non-characteristic power sources does not change with the cumulative amount of reserve margin, its specific value can be the decision value obtained based on any reserve margin step in step 1.1.3. In this implementation, the "1 / 10" criterion formulated by the North American Electric Reliability Corporation (NERC) is used as the benchmark for the reliability index of power generation of the new power system, that is, the expected number of power outage hours per year does not exceed 2.4 hours (0.1 day), and LOLH≤2.4. For the decision sample of the potential operable quantity of characteristic power sources with LOLH≤2.4, the label is assigned a value of 1, indicating that the sample meets the power generation reliability requirements, otherwise the label is assigned a value of 0, and the set of decision samples of the potential operable quantity of characteristic power sources with labels D for each level year is obtained. t , as shown in formula (14).

[0097]

[0098] In the formula, They respectively represent the LOLH label data corresponding to the mth sample in the decision sample set of the potential operable quantity of characteristic power sources in the tth horizontal year.

[0099] Step 2: For each horizontal year, use the labeled potential operational quantity decision sample set of characteristic power sources and the potential operational quantity decision variables of characteristic power sources as feature variables to train the oblique decision tree model for that horizontal year.

[0100] Since the training method for each level year is the same, the training process of the oblique decision tree is explained by taking the t-th level year (t∈{1,…,T}) of the planning period as an example. The specific steps are as follows:

[0101] Step 2.1: Create the root node of the oblique decision tree, set the current node depth d to 0, and set the t-th level year labeled feature power source potential operational quantity decision sample set D t Assign to the root node and set the root node as the current node, denoted as p;

[0102] Step 2.2: Determine whether the depth d of the current node of the decision tree exceeds the maximum depth H. If d>H, the node is set as a leaf node, and its label value is set to the majority sample label value among the samples assigned to the current node (when the decision value of the potential operable quantity of the characteristic power source meets the decision rule represented by the root node to this leaf node, the decision tree determines whether the decision value meets the reliability index requirement based on the leaf node label. If the label is 1, it meets the requirement, and 0, it does not meet the requirement). Otherwise, execute step 2.3.

[0103] Step 2.3: If the feature power potential runnable quantity decision samples assigned to the current node have the same label, the current node is set as a leaf node and the label is recorded as the sample label. Otherwise, execute step 2.4.

[0104] Step 2.4: Randomly initialize the initial values ​​of the segmentation parameter vector and the segmentation constant, denoted as θ t,0 and b t,0 , using the L-BFGS algorithm, with the goal of minimizing the weighted information entropy function shown in formula (15), solve the large-scale unconstrained optimization problem and obtain the current decision tree node segmentation parameter vector θ t and the partition constant b t , forming a linear partition function

[0105] E(θ t ,b t )=W L (θ t ,b t )H L (θ t ,b t )+W R (θ t ,b t )H R (θ t ,b t ) (15)

[0106] Linear segmentation function is a vector of decision variables of the potential operational quantity of the characteristic power source, and For the meaning of , see step 1.2.3.

[0107] In formula (15): W L (θ t ,b t ) and W R (θ t ,b t ), H L (θ t ,b t ) and H R (θ t ,b t ) represent the sum of the corresponding weights and information entropy when all decision samples of the potential operable quantity of all characteristic power sources at the current node are pre-allocated to the left and right child nodes, respectively. The specific formulas are shown in equations (16)-(19):

[0108]

[0109] In formulas (16)-(19): Np is the number of decision samples of the potential operational quantity of the characteristic power source at the current node p, W L,0 (θ t ,b t ) is the sum of the weights of all decision samples of the potential operable quantity of the characteristic power source with the reliability evaluation label of 0 at the current node when all are pre-assigned to the left child node, W L,1 (θ t ,b t ) is the sum of the weights of all decision samples of the potential operable quantity of the characteristic power source with the reliability evaluation label 1 at the current node when all are pre-assigned to the left child node, W R,0 (θ t ,b t ) and W R,1 (θ t ,b t ) means the corresponding parameter when it is assigned to the right child node; and are the mth sample at the current node respectively. The weights of (decision samples of the potential number of operable power sources) when pre-allocated to the left and right child nodes are shown in equations (20) and (21), respectively:

[0110]

[0111] Step 2.5: Use the linear partition function of the current node p According to formula (22) and formula (23), the decision samples of the potential operable quantity of the characteristic power source at the current node are allocated to the left and right child nodes to form the training sample sets of the left and right child nodes respectively. and

[0112]

[0113] Step 2.6: Construct the left child node, set it as the current node p, and let the training sample set of this node be d=d+1, execute step 2.2.

[0114] Step 2.7: Construct the right child node, set it as the current node p, and let the training sample set of this node be d=d+1, execute step 2.2.

[0115] Step 3: Based on the oblique decision tree model of each level year, recursively extract the feasible domain of reliability assessment for each year.

[0116] For a certain level year, select the leaf node with the label value of 1 in the oblique decision tree, and reach the leaf node from the root node through the intermediate nodes and the branch paths between different nodes. The linear inequality corresponding to each node is The logical decision value corresponding to the branch path below the node indicating whether the linear inequality is established can determine a linear constraint on the characteristic variable of the oblique decision tree (i.e., the potential number of characteristic power sources that can be operated) (logical value 1: represents The linear constraints determined are Logical value 0: represents It does not hold, and the corresponding linear constraint is Right now All linear constraints can be enclosed in a polyhedral feasible region. Given the potential operable number of non-characteristic power sources, if the potential operable number of characteristic power sources is within this feasible region, the LOLH index of the new power system will meet the generation reliability requirements.

[0117] In a certain horizontal year, starting from the root node, traversing all leaf nodes with label values ​​of 1 in the oblique decision tree of that year, all polyhedral feasible domains that meet the requirements of power generation reliability indicators in that horizontal year can be found. The construction of the above feasible domain can be carried out in a depth-first search manner, and the specific steps are as follows:

[0118] Step 3.1: Set the root node of the oblique decision tree as the current node, and record the current node as T node , the number of leaf nodes with label 1 is recorded as N lab_1 , the number of leaf nodes with label 1 currently searched is k = 0, and the segmentation parameter matrix is ​​recorded as Q t , the branch selection matrix is ​​denoted as U t and the partition constant vector is denoted as γ t , Q t , U t and γ t Set to empty, i.e. Q t =[],U t =[],γ t =[];

[0119] Step 3.2: If T node If it is not a leaf node, go to step 3.8; otherwise, go to step 3.3;

[0120] Step 3.3: If the current node T node If the label is 1, the reliability assessment feasible domain coefficient matrix R is returned. t,k =U t Q t and constant vector b t,k =U t γ t , k=k+1, execute step 3.4; otherwise, execute step 3.4 directly;

[0121] Step 3.4: If all leaf nodes with label 1 have been searched, that is, k = Nlab_1 , go to step 3.11; otherwise, go to step 3.5;

[0122] Step 3.5: From U t Remove the last diagonal element and the row and column where the element is located, and execute step 3.6;

[0123] Step 3.6: If the current node T node is the left child of its parent node, then the current node T node Set it as its parent node and execute step 3.10; otherwise, set the current node T node Set it as its parent node and execute step 3.7;

[0124] Step 3.7: From U t Remove the last diagonal element and the row and column where the element is located from the partition parameter matrix Q t Remove the last row and split the constant vector γ t Remove the last element and return to step 3.6;

[0125] Step 3.8: Transpose the row vector of the split parameters of the current node Add to the segmentation parameter matrix Q t In the example, the division constant b t Add to the segmentation constant vector γ t , execute step 3.9;

[0126] Step 3.9: Select the square matrix U in the branch t Add a diagonal element at the end of the last diagonal element, set its value to 1, fill the remaining element positions of the row and column corresponding to the new diagonal element with 0, and search for its left child node That is Execute step 3.2;

[0127] Step 3.10: Select the square matrix U in the branch t Add a diagonal element at the end of the last diagonal element, set its value to -1, fill the remaining element positions of the row and column corresponding to the new diagonal element with 0, and search for its right child node That is Execute step 3.2;

[0128] Step 3.11: After completing the depth-first traversal of all leaf nodes, the final reliability evaluation feasible domain can be expressed in the following logical or constraint form:

[0129]

[0130] In the formula, ∨ represents the logical OR operator; W t,kis a binary logical variable whose value represents yes or no, indicating whether the kth reliability assessment feasible domain (mathematical polyhedron) of the tth level year is selected; represents the number of feasible regions for reliability assessment in the t-th level year (i.e., the number of leaf nodes with label 1 in the oblique decision tree in that year); R t,k represents the coefficient matrix corresponding to the kth feasible region in the tth level year; b t,k Represents the constant vector corresponding to the feasible region of the kth reliability assessment in the tth level year.

[0131] Step 4: Build embedded H L A new power system power planning model based on the feasible domain of hourly reliability assessment is proposed. The specific steps are as follows:

[0132] Step 4.1: For each level year, the feasible domain of reliability assessment in the form of logic or constraints for that year is transformed into mixed integer linear constraints using the convex hull modeling method, as shown in equations (25)-(28):

[0133]

[0134] In the formula express Decomposition vector in the feasible region of the kth polyhedron; and They respectively represent the boundary vectors composed of the upper and lower bounds of the potential operable quantity variables of different types of characteristic power sources in the t-th level year, as shown in equations (29) to (34).

[0135]

[0136] The vector elements in equations (31) and (32) are derived from the decision interval of the potential operable quantity of the characteristic power source set in step 1.3.2.

[0137] Step 4.2: For each type of characteristic new power source and characteristic old power source, the potential operational quantity decision variables are constructed as shown in equations (35) and (36).

[0138]

[0139] In formulas (35) and (36), is the set of sub-region names where new power sources with the i-th characteristic can be built in the entire planning area. for The name of the rth sub-region in is the name set of the sub-region where the i-th characteristic old power source is located, for The name of the rth sub-region in .

[0140] Step 4.3: For each type of non-characteristic new power source and non-characteristic old power source, the sum of its potential operable quantity in each sub-area is fixed to the decision value obtained in step 1.1.3 based on any backup margin step, as shown in equations (37) and (38).

[0141]

[0142] In formulas (37) and (38): is the name set of the non-characteristic new power supply (obtained in step 1.2.3), The number of non-characteristic new power supply names, for The name of the i-th type of non-characteristic new power source, is the name set of the sub-areas to be constructed corresponding to the i-th type of non-characteristic new power source, for The name of the rth sub-region in is the potential operable number of the i-th type of non-characteristic new power source in the r-th sub-region in the t-th horizontal year, is the decision value of the potential operable quantity of the i-th type of non-characteristic new power source (obtained in step 1.1.3); is the name set of non-characteristic old power supplies (obtained from step 1.2.3), is the number of non-characteristic old power supply names, for The name of the i-th type of non-characteristic old power supply, is the name set of the sub-area where the i-th type of non-characteristic old power source is located, for The name of the rth sub-region in; is the potential operable number of the i-th type of non-characteristic old power sources in the r-th sub-region in the t-th horizontal year, is the decision value of the potential operable quantity of the i-th type of non-characteristic old power source (obtained in step 1.1.3).

[0143] Step 4.4: Set the reserve margin of each level in the integrated optimization model of new and old power supply input and output decision and unit combination described in the literature "Mixed-integer linear programming models and algorithms for generation and transmission expansion planning of power systems" to zero (that is, set the parameters of the model to zero). Then, constraints (25)-(28) and (35)-(38) are added to the integrated optimization model of new and old power supply switching decision and unit combination, thus forming an embedded H LA new power system power planning model based on feasible domain of hourly reliability assessment.

[0144] Step 5: Use Benders decomposition technology (such as the method introduced in the literature [P.Liu, L.Cheng, J.Zhang, and J.Yu, "Customized Benders Decomposition for Unit Commitment Integrated GenerationExpansion Planning," presented at the 2023International Conference on PowerSystem Technology (Power Con), 2023.]) to solve the embedded H L The integrated optimization model of new and old power supply investment and withdrawal decision and unit combination in the feasible domain of reliability assessment is used to obtain the new and old power supply investment and withdrawal plan that meets the reliability requirements at each level in the planning period and has the lowest comprehensive investment and operation and maintenance cost.

[0145] Example

[0146] The following is combined with Figure 2 To Attachment Fig.10 Taking the 20-year medium- and long-term power planning problem in the ERCOT region of the United States as an example, the present invention is further elaborated to verify its effectiveness. The ERCOT region consists of five sub-regions, namely Northeast, West, Coast, South and Panhandle, of which the Panhandle region has no load and is a power transmission sub-region.

[0147] This embodiment uses the integrated optimization model of new and old power supply decision-making and unit combination provided in Appendix A of the document [C. Li, AJ Conejo, P. Liu, BP Mell, JD Siirola, and I. E. Grossmann, "Mixed-integer linear programming models and algorithms for generation and transmission expansion planning of power systems," European Journal of Operational Research, vol. 297, no. 3, pp. 1071-1082, 2022.]. For detailed introduction, please refer to the document, which can be downloaded at [https: / / doi.org / 10.1016 / j.ejor.2021.06.024]. The original data required to build this model comes from the above-mentioned literature and literature [C.L. Lara, D.S. Mallapragada, D.J. Papageorgiou, A. Venkatesh, and I.E. Grossmann, "Deterministic electric power infrastructure planning: Mixed-integer programming model and nested decomposition algorithm," European Journal of Operational Research, vol. 271, no. 3, pp. 1037-1054, 2018.].

[0148] In the embodiment, the old power types include: (1) coal power: steam turbine (coal-st-old); (2) gas power: steam turbine (ng-st-old), gas turbine (ng-ct-old) and steam-gas combined cycle (ng-cc-old); (3) nuclear power: steam turbine (nuc-st-old); (4) solar power generation: photovoltaic (pv-old); (5) wind power: wind turbine (wind-old).

[0149] New power sources include: (1) coal-fired power: steam turbine (coal-st-new), steam turbine + carbon capture (coal-st-ccs-new); (2) gas-fired power: gas turbine (ng-ct-new), gas-steam combined cycle (ng-cc-new) and gas-steam combined cycle with carbon capture (ng-cc-ccs-new); (3) nuclear power: steam turbine (nuc-st-new); (4) solar power generation: photovoltaic power generation (pv-new); (5) wind power: wind turbine (wind-new).

[0150] Therefore, the collection of old and new power supply names and They are:

[0151] and

[0152] The power supply forced outage probability information used in the embodiment is shown in Table 1, which is derived from the literature ["Generic Database Characteristics," https: / / www.ercot.com / calendar / 04052011-Long-Term-Study-Task (accessed Jul.4, 2023)] and [SLGbadamosi and NINwulu,"Reliability assessment of composite generation and transmission expansion planning incorporating renewable energy sources," Journal of Renewable and Sustainable Energy, vol.12, no.2, 2020.].

[0153] The annual growth rate of load in the embodiment is set to 1.4%, and the historical time series information of wind and solar power output and per-unit load for 8760 hours is used to evaluate the power generation reliability index LOLH, that is, H L = 8760. These data are based on 8760 hours of actual load and wind and solar power output data in the ERCOT region (available for download from https: / / www.ercot.com / gridinfo / load / load_hist and https: / / www.ercot.com / mp / data-products / data-product-details?id=PG7-126-M).

[0154] Table 1: Power forced outage rate

[0155]

[0156] Implementation Case 1: Taking a 20-year planning period as an example, analyze the specific effects of the present invention

[0157] Step 1: Generate a sample data set for oblique decision tree training for each level year of the planning period.

[0158] Step 1.1: For the integrated optimization model of new and old power supply investment and withdrawal decision and unit combination, different reserve margin levels are set, and the model is solved to obtain the optimal decision of the potential operable number of new and old power supplies that meet the LOLH index. The specific process is as follows:

[0159] Step 1.1.1: Set the parameters of the integrated optimization model of new and old power supply decision-making and unit combination The Gurobi solver is called to solve the model using Benders decomposition optimization technology to obtain the decision values ​​of the potential operable quantity of different types of new and old power sources in each horizontal year, as shown in Tables 2 to 5.

[0160] Table 2 Decision values ​​of the potential operable quantity of old power sources from the 1st to the 10th year (the reserve margin in each year is zero)

[0161]

[0162] For old power sources with the same primary energy source and prime mover type, these old power sources are divided into different types because their nameplate capacities in different sub-areas are different. In order to clearly distinguish these types, Tables 2 and 3 append the sub-area information of the old power source after its name. In the embodiment, each type of new power source can be invested and constructed in each sub-area. Therefore, the new power source names in Tables 4 and 5 are not followed by sub-area information. For new power sources with the same primary energy source and prime mover type, considering the typical design and construction standards, the embodiment only considers one representative nameplate capacity.

[0163] Note: The present invention only utilizes the most basic branch and cut module (Branch and Cut) and linear programming module (Linear Programming) in commercial solvers, which are configured in mainstream commercial solvers (such as Gurobi, CPLEX, COPT and Xpress, etc.), so the present invention does not rely on a specific type of commercial solver.

[0164] Table 3 Decision values ​​of the potential operable quantity of old power sources from the 11th to the 20th year (the reserve margin in each year is zero)

[0165]

[0166] Step 1.1.2: For each horizontal year in the planning period, the decision value information of the potential operable number of different types of new and old power sources obtained in step 1.1.1, the capacity information and random fault outage probability information of each conventional power source (each generator set), each wind power source (each wind farm), and each photovoltaic power source (each photovoltaic power station), the historical time series data of wind and solar power generation capacity coefficient and 8760 hours of per unit load in the planned area, and the annual peak load forecast data are used to evaluate the power generation reliability index LOLH corresponding to the horizontal year. The results are shown in Table 6.

[0167] Step 1.1.3: Set the benchmark step sizes from small to large as 0.01, 0.02, 0.03, 0.04, and 0.05; for each benchmark step, gradually increase the reserve margin of each level year from zero, and use the Benders decomposition optimization technology to re-solve the model to obtain the potential operable number of different types of power sources in each level year, and evaluate the LOLH index of the power system to be expanded in each level year. Repeat this process until the LOLH index of all years meets the reliability requirements, and save the decision value of the potential operable number of different types of power sources in each level year.

[0168] Table 4 Decision values ​​of the potential operable quantity of new power sources from the 1st to the 10th year (the reserve margin in each year is zero)

[0169]

[0170] Table 5 Decision values ​​of the potential operable quantity of new power sources from the 11th to the 20th year (the reserve margin in each year is zero)

[0171]

[0172] Table 6 LOLH index corresponding to the decision value of the potential operable quantity of new and old power sources in each level year (the reserve margin in each year is zero)

[0173]

[0174] Step 1.2: For each horizontal year t∈{1,…,20}, determine the characteristic variables of the oblique decision tree based on the five groups of decision information on the potential operable quantities of new and old power sources obtained in step 1.1.

[0175] Step 1.2.1: Under different reserve margin reference step sizes of 0.01, 0.02, 0.03, 0.04, and 0.05, the decision values ​​of the potential operable quantities of different types of new and old power sources obtained in step 1.1.3 for each horizontal year (t∈{1,…,20}) are used to form a decision matrix, which is listed in Tables 7 to 20. In order to save space, the embodiment combines the old power sources in the horizontal years with the same decision values ​​of the potential operable quantities, see Tables 17 to 20 for details.

[0176] Step 1.2.2: Based on Tables 7 to 20, select the power source type whose decision value of the potential operable quantity changes with the change of the reserve margin step size in one or some level years.

[0177] Table 7 Decision values ​​for the potential number of new power sources that can be operated from the first year to the second year

[0178]

[0179] Table 8 Decision values ​​for the potential number of new power sources that can be operated from the 3rd to the 4th year

[0180]

[0181] Table 9 Decision values ​​for the potential number of new power sources that can be operated from the 5th to the 6th year

[0182]

[0183] Table 10 Decision values ​​for the potential number of new power sources that can be operated from the 7th to the 8th year

[0184]

[0185] Table 11 Decision values ​​for the potential number of new power sources that can be operated from the 9th to the 10th year

[0186]

[0187] Table 12 Decision values ​​for the potential number of new power sources that can be operated from year 11 to year 12

[0188]

[0189] The new power source types and change year information are as follows: wind-new: 12th to 20th year; pv-new: 1st to 20th year; ng-cc-new: 2nd to 20th year; ng-cc-ccs-new: 3rd to 20th year; ng-ct-new: 20th year; these new power source names form a set Number of power supply types

[0190] The old power source type and change year information are as follows: ng-cc-old (Coastal): 4th to 20th year, ng-st-old (South): 13th to 20th year; the old power source name composition set Number of old and medium power types

[0191] Table 13 Decision values ​​for the potential number of new power sources that can be operated from year 13 to year 14

[0192]

[0193] Table 14 Decision values ​​for the potential number of new power sources that can be operated from the 15th to the 16th year

[0194]

[0195] Step 1.2.3: For each horizontal year of the planning period, wind-new, pv-new, ng-cc-new, ng-cc-ccs-new and ng-ct-new are selected from step 1.2.2 as characteristic new power sources, ng-cc-old (Coastal) and ng-st-old (South) are selected as characteristic old power sources, and the potential operable number of characteristic power sources is used as the characteristic variable of the oblique decision tree; But not The new power supply is a non-characteristic new power supply (there are 3 types) and will belong to But not The old power supplies are treated as non-characteristic old power supplies (a total of 23 types).

[0196] Table 15 Decision values ​​for the potential number of new power sources that can be operated from year 17 to year 18

[0197]

[0198] Table 16 Decision values ​​for the potential number of new power sources that can be operated from the 19th to the 20th year

[0199]

[0200] Step 1.3: Generate a large number of decision samples of the potential operational quantity of characteristic power sources for each horizontal year from the 1st to the 4th year and from the 17th to the 20th year.

[0201] It can be seen from Table 6 that when the reserve margin of each horizontal year is zero, the LOLH index from the 1st to the 4th year and from the 17th to the 20th year is obviously abnormal. After calculation, we found that only using the reliability assessment feasible domain of the first 4 years and the last 4 years of the planning period can ensure that each year of the entire planning period meets the power generation reliability requirements, and further adding the reliability assessment feasible domain of the 5th to the 16th year will not affect the power planning results. Therefore, starting from step 1.3, the embodiment only needs to focus on the 1st to the 4th year and the 17th to the 20th year, construct the corresponding assessment feasible domain, and based on this, establish a new and old power supply investment and withdrawal decision and unit combination optimization model integrating the 8760 reliability assessment feasible domain. Although the embodiment only uses the reliability assessment feasible domain of the 1st to the 4th year and the 17th to the 20th year, the results are exactly the same as the evaluation results using all 20 horizontal years, because after further adding the feasible domain of the 5th to the 16th year, we found that these constraints are redundant. Therefore, this embodiment is sufficient to verify the correctness of the method proposed in the invention by constructing the feasible domain of reliability assessment for the first 4 years and the last 4 years.

[0202] Table 17 Decision values ​​for the potential operable quantity of old power sources from the 1st to the 6th year

[0203]

[0204] The specific steps include:

[0205] Step 1.3.1: According to Tables 7 to 20, calculate the upper and lower bounds of the decision value of the potential operable quantity of each type of characteristic new and old power sources in each horizontal year from the 1st to the 4th year and from the 17th to the 20th year under different reserve margin benchmark step conditions.

[0206] Taking the characteristic power sources pv-new and ng-cc-old (Coastal) as examples, the upper and lower bounds of the potential operable quantity decision value in the first level year are and They are [24,31] and [12,12] respectively. The upper and lower bounds of the remaining characteristic power sources in the 1st to 4th year and the 17th to 20th year can also be obtained from Tables 7 to 20.

[0207] Table 18 Decision values ​​for the potential operable quantity of old power sources from the 7th to the 8th year

[0208]

[0209] Step 1.3.2: Based on the upper and lower bounds of the decision values ​​of the potential operable quantity of each type of characteristic new and old power sources in step 1.3.1, the quantity of characteristic old power sources that already existed in the initial year of the planning period, and the remaining life of characteristic old power sources in each horizontal year of the planning period, set the decision intervals for the potential operable quantity of characteristic power sources in each horizontal year from the 1st to the 4th year and from the 17th to the 20th year.

[0210] In the embodiment, the number of ng-cc-old (Coastal) and ng-st-old (South) existing in the initial year of the planning period is 12 and 3 respectively, that is, The remaining life information of the characteristic old power sources in each level year of the planning period is derived from the literature [C.Li, AJConejo, P.Liu, BPomell, JDSiirola, and I.E.Grossmann, "Mixed-integer linear programming models and algorithms for generation and transmission expansion planning of power systems," European Journal of Operational Research, vol. 297, no. 3, pp. 1071-1082, 2022.], as shown in Table 21.

[0211] Table 19 Decision values ​​for the potential operable quantity of old power sources from the 9th to the 11th year

[0212]

[0213] Taking the decision interval setting of the potential operable quantity of pv-new and ng-cc-old (Coastal) in the first level year as an example, the conditions that the decision interval setting should meet are discussed.

[0214] According to equations (8) and (9), the lower bound of the decision interval of the potential number of operations pv-new is Not greater than the lower bound of the potential runnable quantity decision value in step 1.3.1 Right now: The upper bound of the potential runnable number of pv-new decisions Not less than the upper bound of the potential operable quantity decision value in step 1.3.1 Right now To this end, pv-new is the decision range of the potential number of operations in the first level year Set to [0, 48].

[0215] Table 20 Decision values ​​for the potential operable quantity of old power sources from the 12th to the 20th year

[0216]

[0217] According to equations (10) and (11), the upper and lower bounds of the decision interval of the potential number of operations of ng-cc-old (Coastal) are

[0218] As shown in Table 21, ng-cc-old(Coastal) has three power supplies reaching their lifespan in the first year, namely Considering that the initial number is 12, that is

[0219] Table 21 Characteristic information of remaining life of old power supply

[0220]

[0221] According to equations (12) and (13), Therefore, Therefore, the decision range of the potential operable number of ng-cc-old(Coastal) in the first level year is set to [9, 12].

[0222] Similarly, according to equations (8) to (13), the conditions that should be met for the decision intervals of the remaining characteristic power sources in each year from the 1st to the 4th year and from the 17th to the 20th year can be determined. Based on these conditions, the decision intervals for the potential number of characteristic power sources that can be operated are set, as shown in Tables 22 and 23.

[0223] Table 22 Decision interval of potential operable quantity of characteristic power sources

[0224]

[0225] In Table 23, the decision range of the potential operable number of ng-st-old(South) from the 1st to the 4th year is [3, 3], which has degenerated into a specific real number. The reason is that ng-st-old(South) has 3 power sources in the initial year of the planning period, and these 3 power sources have reached their service life in the 8th and 14th years respectively (see Table 21). Therefore, from the 1st to the 4th year, the potential operable number of ng-st-old(South) is always 3.

[0226] Table 23 Decision interval of potential operable quantity of characteristic power sources

[0227]

[0228] Step 1.3.3: For each horizontal year from the 1st to the 4th year and from the 17th to the 20th year, within the decision interval of the potential operable number of the seven types of characteristic power sources shown in Tables 22 and 23, with a step size of 1 for the number of single units, generate a sample set of decision on the potential operable number of characteristic power sources by cyclically enumerating the potential operable number of each type of characteristic power source.

[0229] Step 1.4: For each horizontal year from the 1st to the 4th year and from the 17th to the 20th year, evaluate the LOLH index corresponding to each decision sample, determine the label value of each decision sample, and together with the corresponding decision sample, form a labeled feature power source potential operational quantity decision sample set.

[0230] It can be seen from Table 7, Table 8, Table 15 and Table 16 that the potential operable number of the three non-characteristic new power sources is zero. Since the potential operable number of non-characteristic old power sources does not change with the cumulative amount of reserve margin, the specific values ​​of the 23 non-characteristic old power sources can be the decision values ​​obtained by any reserve margin reference step in Table 17 and Table 20.

[0231] Step 2: For each horizontal year from the 1st to the 4th year and from the 17th to the 20th year, use the labeled potential operational quantity decision sample set of characteristic power sources and the potential operational quantity decision variables of characteristic power sources as feature variables to train the oblique decision tree model for that horizontal year.

[0232] The maximum depth of the oblique decision tree model is set to 6, and the maximum number of iterations of the L-BFGS algorithm is set to 100. For the specific training process, see steps 2.1 to 2.7 of the specific implementation method 1. In the embodiment, the decision interval of the potential runnable number of ng-st-old (South) from the 1st to the 4th year is [3, 3], which has degenerated into a specific real number value 3. Therefore, when training the oblique decision tree from the 1st to the 4th year, the split parameter corresponding to the potential runnable number of ng-st-old (South) needs to be set to 0.

[0233] Step 3: Based on the oblique decision tree model for each level year from the 1st to the 4th year and from the 17th to the 20th year, recursively extract the feasible domain of reliability assessment for each level year.

[0234] Taking the first level year as an example, by executing steps 3.1 to 3.11 in the first specific implementation mode, the reliability assessment feasible domain of the year can be constructed, which includes 13 polyhedral feasible domains, as shown in Table 24. In Table 24: the columns where r1 to r6 are located represent the characteristic variable coefficients, the column where β is located is the constant term, and the constraints represented by each row of data have the following form:

[0235]

[0236] If the decision value of the potential operable quantity of characteristic new and old power sources in the first horizontal year is in any of the 13 polyhedral feasible regions shown in Table 24, then the power system composed of these decision values ​​and the decision values ​​of non-characteristic power sources in the first year (see Table 7 and Table 17) will meet the power generation reliability index requirements. In other words, there is an OR logical relationship between the 13 polyhedral feasible regions in Table 24.

[0237] Table 24 Feasible domain of reliability assessment for the first level year

[0238]

[0239] Table 24 Feasible domain of reliability assessment for the first level year - continued from Table 1

[0240]

[0241] Table 24 Feasible domain of reliability assessment for the first level year - continued from Table 2

[0242]

[0243] By using the oblique decision trees for the years 2 to 4 and 17 to 20, and executing steps 3.1 to 3.11 of the first specific implementation method, the feasible domains of reliability assessment for the corresponding years can be constructed respectively. In the years 2 to 4 and 17 to 20, the number of polyhedral feasible domains contained in the feasible domain of horizontal year reliability assessment is 20, 23, 25, 17, 20, 19 and 19, respectively. Due to space limitations, the specific data will not be listed in detail in the embodiment.

[0244] Step 4: Construct a new power system power planning model with an embedded 8760-hour reliability assessment feasible domain. The specific process is as follows:

[0245] Step 4.1: For the 1st to 4th year and the 17th to 20th year, the reliability assessment feasible domain of the annual logic or constraint relationship of each level is transformed into mixed integer linear constraints using the convex hull modeling method, specifically:

[0246]

[0247] Where:

[0248] ,

[0249] express The decomposition vector in each polyhedron feasible region of the reliability assessment feasible region at the t-th level, W t,k is a binary variable that determines whether to select the feasible region of the kth polyhedron, and are the upper and lower bounds of the decision interval for the potential operational quantity of characteristic power sources. For specific values, see Table 22 and Table 23.

[0250]

[0251] R t,k is the coefficient matrix of the feasible region of the kth polyhedron b t,k is the constant vector of the feasible region of the kth polyhedron. Taking the first level year as an example, R 1,k Specifically, the data corresponding to the row where the kth polyhedron feasible region is located and the columns where r1~r6 are located in Table 24, b 1,k Specifically, the data corresponding to the row where the kth polyhedron feasible region is located and the column where β is located in Table 24. t,k and b t,k This can be obtained by executing steps 3.1 to 3.11 of the first specific implementation mode, which will not be given here again due to space limitations. The numbers of polyhedral feasible regions included in the feasible region for reliability assessment at each level are 13, 20, 23, 25, 17, 20, 19 and 19, respectively.

[0252] Step 4.2: In each horizontal year from the 1st to the 4th year and from the 17th to the 20th year, for each type of characteristic new power source and characteristic old power source potential operable quantity decision variable, construct the connection constraints shown in Equations (43) and (44).

[0253] Features The connection constraints for the new power supply are as follows:

[0254]

[0255] Where: A new power name set for the feature, is the number of new power types, is the name of the new power source with the i-th characteristic, which are: is the set of sub-regions where the i-th type of characteristic new power source can be built; in the embodiment, the characteristic new power source can be built in the five sub-regions of ERCOT, so

[0256] for The names of the r-th sub-regions in are:

[0257] The connection constraints of the characteristic old power supply are as follows:

[0258]

[0259] Where: A collection of old power supply names for the feature, is the number of characteristic old power types, are the names of the i-th type characteristic old power supplies, which are: is the set of sub-regions where the i-th type of characteristic old power source is located. In the embodiment, each type of characteristic old power source is located in only one sub-region. for The rth region name in In terms of right In terms of

[0260] Step 4.3: In each horizontal year from the 1st to the 4th year and from the 17th to the 20th year, for each type of non-characteristic new power source and non-characteristic old power source, the sum of its potential operable quantity in each sub-region is fixed to the decision value obtained based on any backup margin step in Table 7, Table 8, Table 15, Table 16, Table 17 and Table 20, as shown in the following formula:

[0261]

[0262] In formula (45): A collection of names for non-characteristic new power supplies. is the number of non-characteristic new power types, for The names of the i-th category non-characteristic new power sources are: The values ​​of are shown in Table 7, Table 8, Table 15 and Table 16. is the set of sub-areas where the i-th type of non-characteristic new power source can be built. In the embodiment, all non-characteristic new power sources can be built in five sub-areas. for The names of the rth sub-areas in are:

[0263] In formula (46): For a collection of non-characteristic old power supplies, is the number of non-characteristic old power types, for The names of the i-th category non-characteristic old power sources are:

[0264]

[0265]

[0266] is the sub-area set where the i-th type of non-characteristic old power source is located, for In the embodiment, each type of non-characteristic old power source exists in only one sub-region, and the sub-region name is marked in brackets after the old power source name. For details, please refer to the above-mentioned old power source type name information.

[0267] Step 4.4: Set the annual reserve margin of each level in the integrated optimization model of new and old power supply investment and decommissioning decision-making and unit combination to zero, and then add constraints (39)-(46) to the integrated optimization model of new and old power supply investment and decommissioning decision-making and unit combination to form a new power system power planning model with an embedded 8760-hour reliability assessment feasible domain.

[0268] Step 5: Use Benders decomposition technology to solve the new power system power planning model with an embedded 8760-hour reliability assessment feasible domain, and obtain the new and old power supply investment and withdrawal plan that meets the reliability requirements at each level of the planning period and has the lowest comprehensive investment and operation and maintenance costs.

[0269] The integrated optimization model of power investment and unit combination with embedded 8760 reliability assessment feasible domain is solved to obtain the comprehensive investment and operation cost of the new and old power supply investment and withdrawal scheme, which is 264.66 ($10 9 ), compared with the step 1 where the reserve margin level is gradually increased by the minimum benchmark step of 0.01 until the reliability index is met, the investment and operation and maintenance comprehensive cost is 265.31 ($10 9 ), decreased by 0.65 ($10 9 ). The LOLH of each level year is shown in Table 25. The LOLH of each level year is less than the reliability standard of 2.4 hours and can meet the probabilistic reliability requirements.

[0270] Table 25 Reliability index LOLH for each level year of Case 1

[0271]

[0272] The actual capacity margin using the reliability assessment feasible region method and the reserve margin dynamic adjustment method is as follows: Figure 2 Compared with the dynamic adjustment method of reserve margin, the absolute and relative reductions of the actual capacity margin of the reliability assessment feasible region method are shown in Table 26.

[0273] As shown in Table 26, except for the 1st to 2nd year and the 4th to 6th year, the reliability assessment feasible region method can effectively reduce the actual capacity margin in each level year. Among them, in the 3rd year and the 7th to 16th year, the former is slightly lower than the latter; in the 17th to 20th year, the former is significantly lower than the latter; in the 19th year, the difference between the two is the largest, with the absolute decrease of the former compared with the latter being 0.0573, and the relative decrease reaching 19.62%. Figure 2 As shown in the figure, except for the 1st to 2nd year and the 4th to 6th year, the capacity margin curve of the reliability assessment feasible domain method is below the actual margin curve of the backup margin method, indicating that the optimal solution of the new and old power supply investment and withdrawal decision-making and unit combination integrated optimization model embedded in the 8760 reliability assessment feasible domain can further effectively squeeze out the redundant water in the capacity margin of the backup margin method on the premise of meeting the reliability requirements, so that the capacity margin in the corresponding level year will be further reduced, thereby reducing the comprehensive investment and operation and maintenance costs accordingly.

[0274] The investment cost and operation and maintenance cost of each level are shown in Figure 3 and Figure 4 In the 4th, 13th, 16th and 20th years, the investment cost of the reliability assessment feasible region method is higher than that of the reserve margin method; in the 3rd, 7th, 17th and 18th years, the investment cost of the reliability assessment feasible region method is lower than that of the reserve margin method. Due to the effective reduction of the overall capacity margin, the total investment cost of the reliability assessment feasible region method is 1.66 ($10 9 ).

[0275] Table 26 Reduction of annual capacity margin at each level in Example 1

[0276]

[0277] In the first to 17 years, the operation and maintenance costs of the reliability assessment feasible region method are very close to those of the reserve margin method; in the 18th to 20th years, the operation and maintenance costs of the reliability assessment feasible region method are higher than those of the reserve margin method. The total operation and maintenance costs of the reliability assessment feasible region method are 0.95 ($10 9 ), because the reliability assessment feasible region method effectively reduces the capacity margin level in the horizontal year, thereby compressing the optimization space at the operation level to a certain extent. However, due to the greater reduction in overall investment costs, the comprehensive investment and operation cost of the reliability assessment feasible region method is lower.

[0278] The installed capacity of power generation by primary energy in each year is as follows: Figure 5As shown. The proportion of wind power and photovoltaic power in the overall energy structure increases year by year. In the first 10 years, the installed capacity of gas-fired power increased slowly, but due to the increase in carbon tax and the retirement of old units, the installed capacity of gas-fired power began to decline after the 10th year; due to the increase in load demand, the installed capacity began to increase year by year in the 16th year. In the first 10 years, the installed capacity of coal-fired power remained unchanged, but due to the retirement of old units and the increase in carbon tax, the installed capacity of coal-fired power gradually declined in the following 10 years. During the 20-year planning period, the installed capacity of nuclear power remained unchanged. In the 20th year, the installed capacity of gas-fired power accounted for 34.2%, the installed capacity of coal-fired power accounted for 8.6%, the installed capacity of nuclear power accounted for 2.8%, the installed capacity of wind power accounted for 35.2%, and the installed capacity of photovoltaic power accounted for 19.2%.

[0279] Figure 6 This is the 24-hour operation dispatch diagram for the peak load day of the 20th year under N-3 conditions (assuming that the three conventional units with the largest capacity fail). As can be seen from the figure, there is no power shortage under N-3 conditions, and the system can still reliably meet the load demand. Due to the influence of transmission line power loss, the total power output is slightly higher than the load demand. The nuclear power source has a slow ramp rate and a high startup cost, so it operates continuously for 24 hours and generates electricity at full rated power, with a power of 1.35GW. Gas power is the main regulating unit to cope with wind and light fluctuations. With the reduction of wind power and photovoltaic power output, when the gas power output cannot meet the load demand, the coal-fired power gradually joins in regulating the output to meet the load demand. In the 21st period, the coal-fired power output reached a peak of 14.33GW, and the gas power reached its maximum output of 62.8GW.

[0280] Implementation Case 2: Impact Analysis of Changes in Decision Ranges of Characteristic Power Source Types and Potentially Operable Quantity

[0281] In order to further analyze the impact of changes in the decision range of characteristic power types and potential operable quantities, the characteristic power types and the decision range of the potential operable quantities of each type of characteristic power sources are gradually expanded to observe the changes in actual capacity margin and investment and operation costs. Similar to Implementation Case 1, Implementation Case 2 also takes the construction of the feasible domain of reliability assessment from the 1st to the 4th year and from the 17th to the 20th year as an example for discussion.

[0282] For characteristic power sources, the decision values ​​corresponding to the reference step size of 0.01 in Table 7, Table 8, Table 15, Table 16, Table 17 and Table 20 are used to gradually expand the decision interval of the potential operable quantity to both sides. For non-characteristic power source types, the potential operable quantity is fixed to the value corresponding to the reference step size of 0.01 in the above tables. The names of the characteristic power source sets and their corresponding power source types are:

[0283] G1: {ng-cc-new};

[0284] G2: {ng-cc-new, ng-cc-ccs-new};

[0285] G3: {ng-cc-new, ng-cc-ccs-new, ng-gt-new};

[0286] G4: {ng-cc-new, ng-cc-ccs-new, ng-gt-new, pv-new};

[0287] G5: {ng-cc-new, ng-cc-ccs-new, ng-gt-new, pv-new, wind-new};

[0288] G6: {ng-cc-new, ng-cc-ccs-new, ng-gt-new, pv-new, wind-new, ng-st-old(South)};

[0289] G7: {ng-cc-new, ng-cc-ccs-new, ng-gt-new, pv-new, wind-new, ng-st-old(South), ng-cc-old(Coast)}.

[0290] The number of characteristic power sources that can potentially be operated is divided into four scenarios according to different decision intervals. The larger the scenario number, the wider the decision interval, as shown in Tables 27 to 33. The average capacity margin, total investment cost, total operation and maintenance cost, and comprehensive investment and operation and maintenance cost of each level year change with the expansion of the characteristic power source type and the decision interval of the potential number of operations as shown in Tables 27 to 33. Figure 7-10 shown.

[0291] Table 27 ng-cc-new potential operational quantity decision interval

[0292]

[0293] Table 28 ng-cc-ccs-new potential operational quantity decision interval

[0294]

[0295] Table 29 ng-gt-new potential runnable quantity decision interval

[0296]

[0297] observe Figure 7It can be seen that under the same scenario conditions, with the increase of characteristic power types, the actual capacity margin and total investment cost show an overall downward trend; under the same characteristic power type conditions, with the increase of the decision interval of the potential number of operable power sources of each type, the actual capacity margin shows an overall downward trend. Figure 8 and Fig. 9 As shown in the figure, due to the decrease in actual capacity margin, the overall investment cost and capacity margin show the same change pattern; the decrease in capacity margin has compressed the optimization space at the operation level to a certain extent, resulting in an increase in operation and maintenance costs, making the overall operation and maintenance costs show the opposite change pattern to the capacity margin and total investment cost. Fig.10 It can be seen that from the perspective of integrated optimization of the investment, withdrawal and operation of new and old power sources, expanding the decision interval of the characteristic power source types and the potential number of each type of characteristic power source that can be operated can create a larger optimization space for further reducing investment and operation costs. Therefore, the increase in operation and maintenance costs is offset by the reduction in investment costs, so that the comprehensive investment and operation costs decrease with the increase in actual capacity margin. The above results further verify the effectiveness of the new power system power planning method proposed in this invention.

[0298] Table 30 Decision interval of potential operable quantity of pv-new

[0299]

[0300] Table 31 Decision interval of potential operable quantity of wind-new

[0301]

[0302] Table 32 ng-st-old (South) potential operational quantity decision interval

[0303]

[0304] Table 33 ng-cc-old (Coastal) potential operational quantity decision interval

[0305]

[0306] The above calculation examples of the present invention are only used to explain the calculation model and calculation process of the present invention in detail, and are not intended to limit the implementation methods of the present invention. For ordinary technicians in the relevant field, other different forms of changes or modifications can be made based on the above description. It is impossible to list all the implementation methods here. All obvious changes or modifications derived from the technical solution of the present invention are still within the scope of protection of the present invention.

Claims

1. A new power system power planning method based on reliability assessment feasible region, characterized by: The following steps are involved: Step 1: Generate a sample data set for oblique decision tree training for each level year of the planning period. The specific process includes: Step 1.1: For the integrated optimization model of new and old power supply investment and withdrawal decision and unit combination, different reserve margin levels are set, and the model is solved to obtain the optimal decision values ​​of the potential operable number of new and old power supplies that meet the LOLH index, including: Step 1.1.1: In the integrated optimization model of new and old power supply investment and decommissioning decision and unit combination, the reserve margin of each level year is calculated. Set it to zero, use Benders decomposition optimization technology to solve the integrated optimization model of new and old power supply investment and withdrawal decision and unit combination, and obtain the decision value of the potential operable number of different types of new and old power supplies in each horizontal year; New power sources refer to power sources that are to be decided and constructed in each year, and old power sources refer to power sources that existed before the start of the planning period. The names of all types of new power sources are grouped into a set. Group the names of all types of legacy power supplies into one set The meanings of the potential operational quantities of new and old power supplies are as follows: For a certain horizontal year, the sum of the cumulative number of new power sources of a certain type invested and constructed in that year and the previous years is called the potential operable number of such new power sources in that horizontal year; the sum of the cumulative number of old power sources of a certain type retired in that year and the previous years is called the cumulative number of old power sources retired in that horizontal year, and the difference between the initial number of old power sources and the cumulative number of retired power sources is called the potential operable number of such old power sources in that horizontal year; Step 1.1.2: For each horizontal year in the planning period, based on the decision values ​​of the potential operable number of different types of new and old power sources obtained in step 1.1.1, the capacity information of conventional power sources, wind and solar power sources, the random failure and outage probability information of conventional power sources, wind and solar power sources, the wind and solar power generation capacity coefficient and the per unit load H in the planned area are used to calculate the decision value of the potential operable number of new and old power sources of different types. L Hourly historical time series data and annual peak load forecast data are used to evaluate the power generation reliability index LOLH corresponding to the horizontal year; Step 1.1.3: Set the benchmark step lengths from small to large. satisfy For each benchmark step, the reserve margin of each level year is gradually increased from zero, and the model is re-solved using the Benders decomposition optimization technology to obtain the potential operable number of different types of power sources in each level year, and the LOLH index of the power system to be expanded in each level year is evaluated. This process is repeated until the LOLH index of all years meets the reliability requirements, and the decision value of the potential operable number of different types of power sources in each level year is saved; Step 1.2: For each horizontal year, based on the decision values ​​of the potential operable quantities of different types of new and old power sources in each horizontal year obtained in step 1.1.3, the decision value matrices of the potential operable quantities of new and old power sources are constructed respectively, with the power source type information as rows and the benchmark step information as columns; based on the decision value matrices of the potential operable quantities of new and old power sources, the name sets corresponding to each type of new and old power sources whose decision values ​​of the potential operable quantities change with the change of the backup margin benchmark step are recorded as Will The corresponding new power sources will be used as the characteristic new power sources of the year. The corresponding types of old power sources are used as the characteristic old power sources of the year, and the potential operable quantity of characteristic new and old power sources is calculated. As a feature variable of oblique decision tree; At the same time, the type belongs to But not The power supply of this year is regarded as a non-characteristic new power supply, and the type belongs to But not The power supply is regarded as the non-characteristic old power supply, and the names of the non-characteristic new power supplies are combined into a set and recorded as The names of non-characteristic old power supplies are combined into a set and recorded as Step 1.3: For each level year t, generate a decision sample of the potential operable quantity of characteristic power sources; Step 1.4: For each horizontal year t, based on the generated value data of the potential operable quantity of characteristic power sources and the decision value data of the potential operable quantity of non-characteristic power sources corresponding to each decision sample, the corresponding LOLH index is evaluated, and the label value of each decision sample is determined according to the relationship between the LOLH index and the threshold value, and together with the corresponding decision sample, a labeled decision sample set D of the potential operable quantity of characteristic power sources is formed. t ; Step 2: For each horizontal year t, use the labeled decision sample set of the potential operational quantity of characteristic power sources and the decision variable of the potential operational quantity of characteristic power sources as the characteristic variable to train the oblique decision tree model for the horizontal year; Step 3: Based on the oblique decision tree model of each level year, recursively extract the feasible domain of reliability assessment for each year; Step 4: Based on the reliability assessment feasible domain of each year in step 3, construct the embedded H L A new power system power planning model based on the feasible domain of hourly reliability assessment; Step 5: Use Benders decomposition technique to solve the embedded H L The integrated optimization model of new and old power supply investment and withdrawal decision and unit combination in the feasible domain of hourly reliability assessment is used to obtain the new and old power supply investment and withdrawal plan that meets the reliability requirements at each level of the planning period and has the lowest comprehensive investment and operation and maintenance cost.

2. According to the novel power system power supply planning method based on reliability assessment feasible region according to claim 1, it is characterized in that: Step 1.1.1 The process of obtaining the decision value of the potential operable quantity of different types of new and old power sources in each level year includes: The names of all types of new power supplies are combined into a set, recorded as The number of names in is the number of new power types, recorded as The name of the new power supply of type i is The names of each sub-region where the i-th type of new power source can be built are grouped into a set, denoted as The number of names in is the number of sub-regions, denoted by The name of the rth sub-region in The potential operable number of new power sources of type i in sub-region r in the t-th horizontal year is recorded as The names of all types of old power supplies are combined into a set, recorded as The number of names in is the number of types of old power supplies, recorded as The name of the i-th type of old power supply is recorded as The names of each sub-area where the i-th type of old power source is located are grouped into a set, denoted as The number of names in is the number of sub-regions, denoted by The name of the rth sub-region in The potential operable quantity of the i-th type of old power source in the r-th sub-region in the t-th horizontal year is recorded as and The decision value of and Decision value of the potential operable quantity of the i-th type of new power source in the t-th level year Decision value of the potential operable quantity of the i-th type of old power source in the t-th level year 3. According to claim 2, the new power system power supply planning method based on reliability assessment feasible region is characterized in that: The process of evaluating the power generation reliability index LOLH corresponding to the level year described in step 1.1.2 includes: (1): Based on the decision values ​​of the potential operable quantities of different types of new and old power sources obtained in step 1.1.1, the decision value information of the potential operable quantities of conventional power sources, wind power sources, and photovoltaic power sources are respectively calculated; (2): Using the decision value information of the potential operable quantity of conventional power sources and wind and solar power sources, the capacity information of each conventional power source, each wind power source, and each photovoltaic power source, and the random fault outage probability information, the power outage capacity probability distribution tables of conventional power sources, wind power sources, and photovoltaic power sources in the year to be evaluated are generated respectively; the steps for generating the power outage capacity probability distribution tables of conventional power sources, wind power sources, and photovoltaic power sources are the same, and the steps for generating the power outage capacity probability distribution table of conventional power sources include: 1) Set the capacity step Δ con , input the decision value information of the potential operable number of conventional power sources in the year to be evaluated, the capacity information of each conventional power source and the random forced power outage probability information Among them C n and r n is the capacity of the nth conventional power source and the probability of random forced power outage, N con It is the sum of the decision values ​​of the potential operable quantity of all types of conventional power sources in the year to be evaluated; 2) Generate the probability distribution table of the power outage capacity of the first power source, p(0) = 1-r1, p(C1) = r1, the power source number is set to n = 1, and the current maximum power outage capacity C max =C1, the power outage probability when the current power outage capacity is X is recorded as p(X); 3) If the set G con If there is a power source that has not participated in generating the power outage capacity probability distribution table, let n=n+1 and execute step 4); otherwise, obtain the probability distribution table corresponding to different power outage capacities; 4) From the set G con Extract the forced power outage probability and unit capacity information of the nth power source from max =C max +C n , with Δ con As the step length, set the power outage capacity X = m·Δ con , increase m from 0 to C max / Δ con , calculate the power outage capacity X = m·Δ con The probability p′(X)=(1-r n )·p(X)+r n ·p(XC n ), use p′(X) to update the current power outage capacity probability distribution table and execute step 3); (3): Loop through all the power outage capacities and their corresponding probabilities in the power outage capacity probability distribution table of conventional power sources, wind power sources, and photovoltaic power sources, and calculate the power outage capacity coefficient of the wind and solar power generation capacity in the planned area and the per-unit load H. L The hourly historical time series data and annual peak load forecast data are used to calculate the annual expected load loss time index LOLH for each level year.

4. The novel power system power supply planning method based on reliability assessment feasible region according to claim 3 is characterized in that: In step 1.1.3, when the reserve margin reference step is Under the condition of , the decision value vector of the potential operable quantity of various new and old power sources in the t-th level year is recorded as: is a vector of decision values ​​of the potential number of new and old power sources that can be operated. is the decision value of the potential number of new power sources of type i that can be operated, is the decision value of the potential operable quantity of the i-th type of old power source; is the backup margin reference step length Under the condition, by dynamically increasing the reserve margin level of each level year and finally making the LOLH index of each level year meet the requirements, the decision value of the potential operable quantity of the i-th type of new power source in the r-th sub-area is obtained by solving the integrated optimization model of the new and old power source investment and withdrawal decision and unit combination. is the decision value of the potential operable quantity of the i-th type of old power source in the r-th sub-area.

5. The novel power system power planning method based on reliability assessment feasible region according to claim 4 is characterized in that: The process of determining the characteristic variables of the oblique decision tree based on the multiple groups of decision information on the potential operable quantities of new and old power sources obtained in step 1.1.3 in step 1.2 includes: Step 1.2.1: Step at different reserve margin benchmarks Under these conditions, the decision values ​​of the potential number of new and old power sources of different types that can be operated in each horizontal year obtained in step 1.1.3 are used to construct the decision matrix shown in equations (2)-(3) with power source type information as rows and benchmark step information as columns: In formula (2), the i-th row of the t-th matrix represents the decision value of the potential operable number of the i-th type of new power source in the t-th horizontal year under different backup margin reference step conditions; in formula (3), the i-th row of the t-th matrix represents the decision value of the potential operable number of the i-th type of old power source in the t-th horizontal year under different backup margin reference step conditions; since the cumulative amount of backup margin in each horizontal year is determined by the reference step in step 1.1.3, the change law of the decision value of the potential operable number of each type of new and old power source in formula (2) and formula (3) is jointly affected by the change of the cumulative amount of backup margin in the current and previous years under different reference step conditions; Step 1.2.2: For each type of new power source, use formula (2) to select the new power source type whose decision value of the potential operable quantity changes with the change of the reserve margin benchmark step in one or some horizontal years, and record the name of each type of new power source whose decision value of the potential operable quantity changes with the change of the reserve margin benchmark step in one or some horizontal years as a set For each type of old power source, use formula (3) to select the type of old power source whose potential operable quantity decision value changes with the change of the reserve margin benchmark step in one or some horizontal years, and the names of various types of old power sources whose potential operable quantity decision value changes with the change of the reserve margin benchmark step in one or some horizontal years are combined into a set and recorded as Step 1.2.3: For each level year t, the set obtained in step 1.2.2 and Represent the name set of the new power source and the name set of the old power source respectively. The potential operable number of the new and old power sources is composed into vectors, which are recorded as and is the potential number of new power sources of the i-th type, as one of the characteristic variables of the oblique decision tree, is the name of the new power source with the i-th characteristic, is the potential operable quantity of the old power source of the i-th type, as one of the characteristic variables of the oblique decision tree, is the name of the old power source with the i-th characteristic, 6. The novel power system power planning method based on reliability assessment feasible region according to claim 5 is characterized in that: The process of generating a decision sample of the potential operable quantity of a characteristic power source as described in step 1.3 includes: Step 1.3.1: For each level year t of the planning period, traverse each row of the potential operable quantity decision value matrix of new and old power sources, and count the upper and lower bounds of the potential operable quantity decision value of each type of characteristic new power source and characteristic old power source in the t-th level year. The upper and lower bounds of the potential operable quantity decision value of the i-th type of characteristic new power source are recorded as The upper and lower bounds of the decision value of the potential operable quantity of the i-th type of characteristic old power source are recorded as Step 1.3.2: Based on the upper and lower bounds of the potential operational quantity decision values ​​of each type of new and old power sources in step 1.3.1, the quantity of old power sources in the initial year of the planning period, and the remaining life of old power sources in each level of the planning period, set the decision interval corresponding to the i-th type of new power source in the t-th level year as The decision interval corresponding to the i-th type of characteristic old power supply is recorded as The decision interval of the potential operable quantity of new and old power sources satisfies the characteristic The upper and lower bounds of the number of potential operating characteristics of old power sources meet is the number of old power sources with the i-th characteristic existing in the initial year of the planning period, is the number of old power sources with the i-th characteristic reaching the end of their service life in the k-th horizontal year; Step 1.3.3: For each horizontal year t, within the decision range of the potential operable quantity of each type of characteristic power source, with the number of single units as 1 as the step length, the potential operable quantity of each type of characteristic power source is cyclically enumerated, which is called the generated value of the potential operable quantity of characteristic power sources, and a decision sample containing the information of the potential operable quantity of each type of characteristic power source is generated to construct a decision sample set of the potential operable quantity of characteristic power sources; The decision sample set of the potential operable quantity of characteristic power sources in the t-th horizontal year is: M t is the total number of decision samples, and the mth sample is recorded as: is a vector of generated values ​​of the potential number of new power sources that can be run for all features in the mth sample, is the generated value of the potential number of new power sources with the i-th characteristic in the m-th sample that can be operated, Generate a vector of values ​​for the number of potentially operational old power sources for all features in the mth sample, is the generated value of the potential number of new power sources with the i-th characteristic in the m-th sample that can be operated.

7. The novel power system power planning method based on reliability assessment feasible region according to any one of claims 1 to 6 is characterized in that: The process of training the oblique decision tree model for the horizontal year in step 2 includes: Step 2.1: Create the root node of the oblique decision tree, set the current node depth d to 0, and take the t-th level year labeled feature power source potential operational quantity decision sample set D t Assign to the root node and set the root node as the current node, denoted as p; Step 2.2: Determine whether the depth d of the current node of the decision tree exceeds the maximum depth H. If d>H, the node is set as a leaf node, and its label value is set to the majority sample label value among the samples assigned to the current node; otherwise, execute step 2.3; Step 2.3: If the feature power source potential runnable quantity decision samples assigned to the current node have the same label, the current node is set as a leaf node, and the label is recorded as the sample label; otherwise, execute step 2.4; Step 2.4: Randomly initialize the initial values ​​of the segmentation parameter vector and the segmentation constant, denoted as θ t,0 and b t,0 , using the L-BFGS algorithm, with the goal of minimizing the weighted information entropy function shown in formula (15), solve the large-scale unconstrained optimization problem and obtain the current decision tree node segmentation parameter vector θ t and the partition constant b t , forming a linear partitioning function E(θ t ,b t )=W L (i t ,b t )H L (i t ,b t )+W R (i t ,b t )H R (i t ,b t ) (15) in, is a vector of decision variables for the potential operational quantity of characteristic power sources; W L (θ t ,b t ) and W R (θ t ,b t ), H L (θ t ,b t ) and H R (θ t ,b t ) represent the sum of the corresponding weights and information entropy when all decision samples of the potential operable quantity of all feature power sources at the current node are pre-allocated to the left and right child nodes respectively; Step 2.5: Use the linear partition function of the current node p According to formula (22) and formula (23), the decision samples of the potential operable quantity of the characteristic power source at the current node are allocated to the left and right child nodes to form the training sample sets of the left and right child nodes respectively. and Step 2.6: Construct the left child node, set it as the current node p, and let the training sample set of this node be d=d+1, execute step 2.2; Step 2.7: Construct the right child node, set it as the current node p, and let the training sample set of this node be D t R , d=d+1, execute step 2.

2.

8. The novel power system power planning method based on reliability assessment feasible region according to claim 7 is characterized in that: The process of recursively extracting the feasible domain of reliability assessment for each year based on the oblique decision tree model for each level year in step 3 includes: For a certain level year, select the leaf node with the label value of 1 in the oblique decision tree. The label value of 1 means that the LOLH index in step 1.4 is less than or equal to the label value corresponding to the threshold. Starting from the root node, the leaf node is reached through the intermediate nodes and the branch paths between different nodes. The linear inequality corresponding to each node is The logical decision value indicating whether the linear inequality holds true or not, corresponding to the branch path below the node, determines a linear constraint on the characteristic variables of the oblique decision tree, and all linear constraints enclose a polyhedral feasible domain; given the potential operable number of non-characteristic power sources, if the potential operable number of characteristic power sources is within this feasible domain, the LOLH index of the new power system will meet the power generation reliability requirements; in a certain horizontal year, starting from the root node, traverse all leaf nodes with a label value of 1 in the oblique decision tree of that year to find all polyhedral feasible domains that meet the power generation reliability index requirements in that horizontal year.

9. The novel power system power planning method based on reliability assessment feasible region according to claim 8 is characterized in that: The construction of the feasible domain of the polyhedron is carried out in a depth-first search manner, and the specific steps include: Step 3.1: Set the root node of the oblique decision tree as the current node, and record the current node as T node , the number of leaf nodes with label 1 is recorded as N lab_1 , the number of leaf nodes with label 1 currently searched is k = 0, and the segmentation parameter matrix is ​​recorded as Q t , the branch selection matrix is ​​denoted as U t and the partition constant vector is denoted as γ t , Q t , U t and γ t Set to empty, i.e. Q t =[],U t =[],γ t =[]; Step 3.2: If T node If it is not a leaf node, go to step 3.8; otherwise, go to step 3.3; Step 3.3: If the current node T node If the label is 1, the reliability assessment feasible domain coefficient matrix R is returned. t,k =U t Q t and constant vector b t,k =U t γ t , k=k+1, execute step 3.4; otherwise, execute step 3.4 directly; Step 3.4: If all leaf nodes with label 1 have been searched, that is, k = N lab_1 , go to step 3.11; otherwise, go to step 3.5; Step 3.5: From U t Remove the last diagonal element and the row and column where the element is located, and execute step 3.6; Step 3.6: If the current node T node is the left child of its parent node, then the current node T node Set it as its parent node and execute step 3.10; otherwise, set the current node T node Set it as its parent node and execute step 3.7; Step 3.7: From U t Remove the last diagonal element and the row and column where the element is located from the partition parameter matrix Q t Remove the last row and split the constant vector γ t Remove the last element and return to step 3.6; Step 3.8: Transpose the row vector of the split parameters of the current node Add to the segmentation parameter matrix Q t In the example, the division constant b t Add to the segmentation constant vector γ t , execute step 3.9; Step 3.9: Select the square matrix U in the branch t Add a diagonal element at the end of the last diagonal element, set its value to 1, fill the remaining element positions of the rows and columns corresponding to the newly added diagonal element with 0, and search for the left child node That is Execute step 3.2; Step 3.10: Select the square matrix U in the branch t Add a diagonal element at the end of the last diagonal element, set its value to -1, fill the remaining element positions of the row and column corresponding to the new diagonal element with 0, and search for its right child node That is Execute step 3.2; Step 3.11: After the depth-first traversal completes all leaf nodes, the final reliability evaluation feasible domain is expressed in the following logical or constraint form: In the formula, ∨ represents the logical OR operator; W t,k is a binary logical variable whose value represents yes or no, indicating whether the kth reliability assessment feasible region of the tth level year is selected; represents the number of feasible regions for reliability assessment in the t-th level year; R t,k represents the coefficient matrix corresponding to the kth feasible region in the tth level year; b t,k Represents the constant vector corresponding to the feasible region of the kth reliability assessment in the tth level year.

10. The novel power system power supply planning method based on reliability assessment feasible region according to claim 9 is characterized in that: Step 4: Building the embedded H L The process of the new power system power planning model with hourly reliability assessment feasible domain includes: Step 4.1: For each level year, the convex hull modeling method is used to transform the feasible domain of reliability assessment for that year into mixed integer linear constraints, as shown in equations (25)-(28): In the formula express Decomposition vector in the feasible region of the kth polyhedron; and They represent the boundary vectors composed of the upper and lower bounds of the potential operable quantity variables of different types of characteristic power sources in the t-th level year, as shown in equations (29) to (34); Step 4.2: For each type of characteristic new power source and characteristic old power source potential operable quantity decision variable, construct the connection constraints shown in equations (35) and (36); In formulas (35) and (36), is the set of sub-region names where new power sources with the i-th characteristic can be built in the entire planning area. for The name of the rth sub-region in is the name set of the sub-region where the i-th characteristic old power source is located, for The name of the rth sub-region in; Step 4.3: For each type of non-characteristic new power source and non-characteristic old power source, the sum of its potential operable quantity in each sub-area is fixed to the decision value obtained in step 1.1.3 based on any backup margin step, as shown in equations (37) and (38); in, is the decision value of the potential operable quantity of the i-th type of non-characteristic new power source; is the decision value of the potential operable quantity of the i-th type of non-characteristic old power source; is the name set of the sub-areas to be constructed corresponding to the i-th type of non-characteristic new power source, for The name of the rth sub-region in is the name set of the sub-area where the i-th type of non-characteristic old power source is located, for The name of the rth sub-region in; is the potential operable number of the i-th type of non-characteristic new power source in the r-th sub-region in the t-th horizontal year, is the potentially operable number of the i-th type of non-characteristic old power sources in the r-th sub-region in the t-th horizontal year. Step 4.4: Set the annual reserve margin of each level in the integrated optimization model of new and old power supply switching decision-making and unit combination to zero, and then add constraints (25)-(28) and (35)-(38) to the integrated optimization model of new and old power supply switching decision-making and unit combination to form an embedded H L A new power system power planning model based on feasible domain of hourly reliability assessment.

Images