Multi-type energy storage planning method and device considering continuous low output scene

Abstract

The invention provides a multi-type energy storage planning method and device considering a continuous low output scene, and belongs to the field of power system planning. The method comprises the steps that a stochastic programming framework considering the new energy continuous low-output Markov process is established, the framework comprises a programming stage and multiple operation decision stages, and the multiple operation decision stages are obtained by dividing the operation year of a power system; based on the Markov process planning framework, a multi-type energy storage planning model considering a continuous low output scene is established, and the planning model comprises a planning stage sub-model and a multi-stage operation sub-model considering a Markov conversion process; and solving the multi-type energy storage planning model to obtain a multi-type energy storage planning scheme. According to the method, the evolution dynamic state of the continuous low output process of the new energy can be described more truly, the robustness and rationality of energy storage planning are improved, and the load loss risk of the system is reduced.

Description

Technical Field

[0001] This invention belongs to the field of power system planning, and specifically relates to a multi-type energy storage planning method and device that considers scenarios with continuous low power output. Background Technology

[0002] With a high proportion of renewable energy sources, the power system may face a more severe risk of power shortages due to extreme weather conditions. For example, prolonged periods of no wind and no sunlight combined with a surge in high-temperature loads could prevent the system from meeting load demands in the short term.

[0003] Traditional planning methods primarily consider the operational scheduling and planning problems of single extreme weather events on a multi-day scale, making it difficult to characterize the timing and frequency of sustained low-output events of renewable energy sources within an annual period. With the addition of long-duration energy storage, the impact of the temporal distribution of sustained low-output PLOS events on system scheduling will be further amplified. Therefore, it is necessary to consider energy storage planning for sustained low-output scenarios on an annual timescale. On the other hand, existing studies mostly describe the sustained low-output process of renewable energy sources using static scenario sets, making it difficult to characterize the uncertainties in its occurrence time, frequency, and duration, and failing to describe the evolutionary dynamics of PLOS, thus limiting the rational configuration of energy storage systems.

[0004] In power systems with long-duration energy storage, the operation of energy storage is highly correlated with the time-series information of wind and solar loads. If the temporal characteristics of PLOS cannot be accurately characterized, energy storage resources may be unable to provide support at critical moments due to insufficient stored power. Most existing studies adopt a two-stage model, assuming that the scheduling layer of the planning model knows in advance the timing, frequency, and duration of continuous low-output events throughout the year. This assumption is too idealistic, failing to reflect the suddenness of continuous low-output processes, and failing to describe the phased decision-making process as future information is gradually revealed. Furthermore, it does not specifically address the dynamic transition and correlation between the system in normal operating conditions and extreme scenarios. Summary of the Invention

[0005] The purpose of this invention is to overcome the shortcomings of existing technologies and propose a multi-type energy storage planning method and device that considers scenarios of continuous low output. This invention can more realistically depict the dynamic evolution of continuous low output processes of new energy sources, increase the robustness and rationality of energy storage planning, and reduce the risk of system load failure.

[0006] A first aspect of this invention proposes a multi-type energy storage planning method considering scenarios with sustained low power output, comprising:

[0007] A stochastic programming framework is established that takes into account the Markov process of continuous low output of new energy sources. The framework includes a planning phase and multiple operation decision phases, which are obtained by dividing the power system operation year.

[0008] Based on the Markov process planning framework, a multi-type energy storage planning model that takes into account the continuous low output scenario is established. The planning model includes a planning stage sub-model and a multi-stage operation sub-model that considers the Markov transformation process.

[0009] Solving the multi-type energy storage planning model yields planning schemes for the multi-type energy storage.

[0010] In a specific embodiment of the present invention, the stochastic programming framework expression for the Markov process that takes into account the continuous low output of new energy sources is as follows:

[0011]

[0012] Among them, let the random process The process evolves dynamically between stages, where t represents the stage number and T is the stage number of the last stage; s represents the state, s=1 indicates that the stage is in a normal state, and s=2 indicates that the stage is in a state of continuous low output of new energy; ξ s,t Let be a random variable in stage t under state s, including renewable energy output and load power. Let Q be the set of random variables in stage t and state s; t,s =i(·) represents the future cost function with current uncertainty state i at stage t; x t-1 c is the decision variable for stage t-1; t (ξ s＝i,t Let ξ be a random variable. s＝i,t In this case, the cost vector of stage t; P state,i,j Q(t) represents the time-varying state transition probability of the Markov model. T+1 (x T )≡0,x T Q is the decision variable for the final stage. T+1 This represents the total cost of the final stage.

[0013] ρ(·) is a nested risk measure, where λ, α ∈ [0, 1], λ is the risk weight, and α is the gradation value for the worst-case scenario; A t B is the constraint coefficient matrix related to stage t. t b is the constraint coefficient matrix related to stage t-1. t (ξ s,t () represents the constraint coefficient vector associated with the random variable;

[0014] In equation (1), This indicates the cost at the current stage. This represents the expected future cost considering Markov transformation probabilities.

[0015] In a specific embodiment of the present invention, the expression for the planning stage sub-model is as follows:

[0016]

[0017]

[0018] In the formula, the superscript z represents different energy storage types, the subscript k represents the k-th node in the power system; Q0 represents the objective function of the planning stage, Q1 represents the total cost of the first operation decision stage; C inv For the investment cost of energy storage, C om For the operation and maintenance costs of energy storage; A y,z It is the discount rate for the investment in energy storage. and Let z represent the investment cost of the power capacity and the investment cost of the energy capacity of energy storage z, respectively. and These are the unit operation and maintenance costs related to energy storage power capacity and energy capacity, respectively. Let z be the power capacity of the energy storage to be planned at the k-th node. Let z be the energy capacity of the planned energy storage at the k-th node. and Let P be the maximum power capacity and maximum energy capacity of the planned energy storage z-th node; 0,j Let be the probability that the first operational decision stage is state j. To take the expectation of the variable in parentheses, Q1(x0,ξ) s,1 Let x0 be the total cost of the first operational decision-making phase, x0 be the planning result output of the planning phase, and ξ be the total cost of the first operational decision-making phase. s,1 Represents the random variable in the first operational decision-making stage; This represents the expected total cost of the first operational decision-making stage when the state is j.

[0019] In a specific embodiment of the present invention, the objective function expression of the multi-stage running sub-model considering the Markov transformation process is as follows:

[0020]

[0021] In the formula, Q t,s＝i C represents the objective function in state s = i during the t-th running decision stage; shed,k Indicates the penalty cost for abandoning load; C start,k and C gen,k S represents the start-up cost and power generation cost of the thermal power unit located at node k, respectively; rated,k γ represents the rated capacity of the thermal power unit located at node k; w and γ pvThese represent the unit penalty cost for curtailing wind and solar power; and These represent the startup status and generating power of the thermal power unit located at node k, respectively; the subscript d represents days, and τ represents the minimum scheduling time interval. Represents the set of all days in the t-th stage of the decision-making process; and P represents the curtailed wind power and curtailed solar power during the time period τ. shed,d,τ,k Indicates the power of abandoned load;

[0022] The constraints of the multi-stage operating sub-model considering the Markov transformation process include intra-stage constraints and inter-stage constraints.

[0023] The intra-stage constraints include:

[0024]

[0025]

[0026] P l,d,τ =(δ k1,d,τ -δ k2,d,τ ) / X k1,k2 (twenty one)

[0027]

[0028] δ 0,d,τ =0 (23)

[0029] In the formula, This indicates the online status of the thermal power unit at node k, where 1 indicates the unit is online and 0 indicates the unit is offline. These represent the start-up and shutdown states of the unit at node k, respectively, both being 0-1 variables; μ k These represent the maximum and minimum generating capacities of the thermal power unit at node k, respectively; RU k and RD k These represent the maximum uphill and downhill ramp rates of the thermal power unit at node k; T on,k and T off,k Minimum start-up and shutdown times for the thermal power unit at node k; These represent the self-discharge, charging, and discharging efficiencies of the energy storage system, respectively. and Let z represent the power capacity and energy capacity of energy storage z at node k, respectively. These represent the charging power and discharging power of the energy storage system, respectively. These represent the intraday and interday energy of the energy storage system, respectively; Ω yΩ m Ω d These represent sets of energy storage types applicable to different balance cycles; M start (m) and M end (m) represents the first and last day of month m, respectively; |τ| represents the number of time periods in a day; P w,d,τ ,P PV,d,τ These represent the generating capacity of wind and solar power, respectively; the symbols marked with an asterisk (*) indicate the available generating capacity of wind and solar power, respectively; β up ,α up ,β dn ,α dn It is the peak shaving reserve factor related to new energy sources and load; P 1oad,d,τ,k Let P be the electrical power consumption of the load at node k. shed,d,τ,k P represents the power of the load being discarded at node k; l,d,τ δ represents the power flow on branch l. k,d,τ The phase angle at node k is represented; the superscripts min and max represent the minimum and maximum allowable values, respectively; δ 0,d,τ X represents the phase angle of the reference node. k1,k2 This represents the reactance of the tie line from node k1 to node k2;

[0030] The inter-stage constraints include:

[0031]

[0032] In the formula, set Ω on,t,t-1 and Ω off,t,t-1 These represent the last T of stage t-1, respectively. on,k Time period and T off,k Time period; d end For a set of dates The last day; Ω t It is the set of all time intervals in stage t.

[0033] In one specific embodiment of the present invention, it further includes:

[0034] The intra-stage constraints and inter-stage constraints are expressed in a compact form as shown in equation (27):

[0035] B t x t-1 +A t x t ≥b t (27)

[0036] Among them, B t A is a coefficient matrix related to the decision variables of the previous stage. t b is the coefficient matrix related to the decision variables at the current stage.t x is a constant vector representing the constraints; t-1 and x t Let be the decision variables for stages t-1 and t.

[0037] In one specific embodiment of the present invention, it further includes:

[0038] The compact form of the multi-type energy storage planning model that takes into account the continuous low output scenario is as follows:

[0039]

[0040] B t x t-1 +A t x t ≥b t

[0041] In a specific embodiment of the present invention, solving the multi-type energy storage planning model includes:

[0042] 1) Reconstruct multi-stage conditional risk values ​​by nesting multiple types of energy storage planning models;

[0043] According to the conditional risk value expression in equation (29), ρ(·) in equation (1) is expressed as shown in equation (30);

[0044] Let V be a discrete random variable, and let each random variable take the value v. l v l Let V represent the specific values ​​of the l-th random variable, and there are a total of L values. Then the expected value of the discrete random variable V is expressed as (1 / L)Σv. l In the formula, b is the auxiliary variable introduced to linearize the conditional risk value function; this is achieved by introducing auxiliary constraints and the auxiliary variable δ. l With the auxiliary variable υ, equation (30) is reconstructed into a linear programming form as shown in equation (31), where v l The future cost function in equation (1);

[0045]

[0046] 2) Reconstruct the model after step 1) using a Markov process to transform the original model into a multi-stage mixed integer linear programming problem;

[0047] First, ignore the outer ρ(·) in equation (1); let the stochastic programming problem have a finite support set in each stage, and let each data sample be numbered by the subscript o. Then the conditional probability can be discretized and expressed in the form of equation (32), where ξ s,t,o It is the o-th data sample of state s in stage t. θ is the total number of data samples in state s at stage t+1. t+1,j,o As an auxiliary variable, it replaces the discretized future cost function;

[0048]

[0049] st A t x t ≥b t (ξ s＝i,t,o )-B t x t-1 (32)

[0050] θ t+1,j,o ≥Q t+1,s＝j,o ,x t ≥0

[0051] Estimating the future cost function Q using the cutting plane t+1,s＝j,o The lower bound of equation (32) is further constructed into the form of equations (33)-(34); where, It is the attempted solution in the t-th iteration, π t,s＝i,o (ξ s＝i,t,o ) is the dual variable of the cross-stage constraint; This is a relaxation problem of the original problem (32); Let be the set of all cutting planes for state s in stage t+1 and data sample o;

[0052]

[0053] By reintroducing ρ(·) and substituting equations (29)-(31) into equations (33)-(34), we obtain the form shown in equation (35). Thus, the original problem equation (1) is reconstructed into a multi-stage mixed integer linear programming problem.

[0054]

[0055] 3) Solve the multi-stage mixed-integer linear programming problem obtained in step 2) using the mixed-integer stochastic dual dynamic programming algorithm;

[0056] First, we introduce auxiliary constraint equation (36), where z t-1 For continuous variables, x is a state variable. t-1 The local copy in stage t; the constraints of each stage are updated as shown in equation (37);

[0057] z t-1 =x t-1 ,[π t,i,d (ξ s＝i,t,o (36)

[0058] A t xt ≥b t (ξ s＝i,t,o )-B t z t-1 (37)

[0059] By relaxing the integer variables of the original problem into continuous variables and solving the linear programming problem The Benders' cutting plane is obtained as shown in equation (38), including the dual variables. and objective function The Lagrange cut plane is used to solve a max-min bilevel optimization problem (40) in backward backtracking, namely the Lagrange relaxation problem. The outer optimization maximizes the objective function by finding the optimal dual variable, while the inner optimization is the original integer optimization problem of each stage; the generated cutting plane contains a combination of the optimal dual variable and the optimal objective function value as shown in Equation (41);

[0060]

[0061] In a specific embodiment of the present invention, the specific steps of the mixed integer random dual dynamic programming algorithm are as follows:

[0062] 3-1) Input the initial state x ini and random data sample ξ s,t Set the iteration count ite = 0, the lower bound LB = -∞, and the flag = 0;

[0063] 3-2) Set ite = ite + 1; For all stages, from stage 1 to T, perform forward sampling, first performing state sampling: based on the state s of stage t-1. t-1 And Markov transition probability simulation or sampling of t-stage state s t Then perform scene sampling: from A sample of uncertain parameter ξ is randomly selected from the samples. s,t,o And solve equation (35) to obtain an attempted solution.

[0064] 3-3) Determine if ite is N ite Integer multiples of:

[0065] If yes, proceed to step 3-4); otherwise, proceed to step 3-5.

[0066] 3-4) Perform a backward backtracking for all possible states s from stage T to stage 2. t as well as Each data sample ξ s,t,o Solve The corresponding dual variables and objective function values ​​are obtained, and the cutting plane expression (39) is generated and passed to stage t-1;

[0067] 3-5) Perform a backward backtracking from stage T to all stages t in stage 2, for all possible states s. t as well as Each data sample ξ s,t,o Solve The corresponding dual variables and objective function values ​​are obtained, and the cutting plane expression (38) is generated and passed to stage t-1;

[0068] 3-6) Update the lower bound by taking the larger value of the total cost of the current iterative planning stage and the current lower bound LB;

[0069] 3-7) If the convergence criterion is met, set flag = 1 and output the planning scheme x0.

[0070] In one specific embodiment of the present invention, it further includes:

[0071] The solution efficiency of the model can be improved by the following methods:

[0072] 1) Add regularization term (43) to the objective function of equation (35) to punish attempts that deviate significantly from the current optimal solution;

[0073]

[0074] In the formula, It is a positive semi-definite matrix; R t As an auxiliary variable, it represents the portion B in the stage t constraint that is related to stage t-1. t x t-1 That is, coefficient matrix B t Compared with the decision variable x in the previous stage t-1 The product; R in the previous loop t The value of ;

[0075] 2) A dynamic sampling method is adopted to continuously adjust the sampling probability of the scene and state during the algorithm operation;

[0076] At the beginning of the algorithm's operation, the transition probabilities of the normal state and the PLOS state of continuous low output of new energy are considered to be equal; after the lower bound increases significantly, sampling is performed based on the actual transition probabilities of the two states.

[0077] 3) Parallel computing is used to improve computing speed in different sampling paths of forward sampling and different data sample scenarios of backward backtracking.

[0078] A second aspect of the present invention provides a multi-type energy storage planning device considering continuous low-output scenarios, comprising:

[0079] A Markov process programming framework construction module is used to establish a stochastic programming framework that takes into account the continuous low output Markov process of new energy sources. The framework includes a planning stage and multiple operation decision stages, which are obtained by dividing the power system operation year.

[0080] The energy storage planning model construction module is used to establish a multi-type energy storage planning model that takes into account the continuous low output scenario based on the Markov process planning framework. The planning model includes a planning stage sub-model and a multi-stage operation sub-model that considers the Markov transformation process.

[0081] The planning scheme generation module is used to solve the multi-type energy storage planning model to obtain planning schemes for the multi-type energy storage.

[0082] A third aspect of the present invention provides an electronic device comprising:

[0083] At least one processor; and a memory communicatively connected to said at least one processor;

[0084] The memory stores instructions that can be executed by the at least one processor, and the instructions are configured to execute the above-described multi-type energy storage planning method considering a continuous low-output scenario.

[0085] A fourth aspect of the present invention provides a computer-readable storage medium storing computer instructions for causing the computer to execute the above-described multi-type energy storage planning method considering a continuous low-output scenario.

[0086] The features and beneficial effects of this invention are as follows:

[0087] This invention divides the operating scenario into a normal state and a continuous low-output state, models the inter-stage evolution dynamics of continuous low-output events, and integrates this evolution dynamics into multi-stage operating decisions to construct an energy storage planning model that reflects the occurrence time, duration, and frequency of PLOS on an annual timescale. Simultaneously, by introducing stochastic dual dynamic integer programming, it solves a multi-stage stochastic programming model containing integers and addresses multi-stage optimization problems with numerous scenario evolution paths.

[0088] 1) This invention can effectively decompose complex annual problems into multiple small-scale subproblems that can be solved in parallel, thereby improving the computational efficiency of large-scale system planning.

[0089] 2) By explicitly modeling the phase transition probability between PLOS and the normal state, this invention can capture the dynamic characteristics of the evolution of system boundary conditions over time, and describe the uncertainty of the occurrence time, frequency and duration of the continuous low output process of new energy.

[0090] 3) This invention enhances the ability to avoid low-probability, high-impact events by nesting CVaR risk modeling mechanisms, making the final planning scheme more robust and operable in practice.

[0091] 4) This invention takes into account the unpredictable nature of annual power system operation, enabling the operation layer to better align with actual power system dispatching and making the planning results more robust. It provides flexibility for situations where dispatching decisions differ from optimal operating decisions due to incomplete information during actual dispatching operations. Attached Figure Description

[0092] Figure 1 This is a flowchart of a multi-type energy storage planning method considering a scenario of continuous low output, according to an embodiment of the present invention. Detailed Implementation

[0093] This invention proposes a multi-type energy storage planning method and device that considers scenarios with continuous low power output. The following is a detailed description in conjunction with the accompanying drawings and specific embodiments.

[0094] A first aspect of this invention proposes a multi-type energy storage planning method considering scenarios with sustained low power output, comprising:

[0095] A stochastic programming framework is established that takes into account the Markov process of continuous low output of new energy sources. The framework includes a planning stage and multiple operation decision stages, which are obtained by dividing the power system operation year.

[0096] Based on the Markov process planning framework, a multi-type energy storage planning model that takes into account the continuous low output scenario is established. The planning model includes a planning stage sub-model and a multi-stage operation sub-model that considers the Markov transformation process.

[0097] Solving the multi-type energy storage planning model yields planning schemes for the multi-type energy storage.

[0098] In a specific embodiment of the present invention, the overall process of the multi-type energy storage planning method considering continuous low-output scenarios is as follows: Figure 1 As shown, it includes the following steps:

[0099] 1) Establish a stochastic programming framework that takes into account the continuous low-output Markov process of new energy sources.

[0100] In this embodiment, the Markov Mixed Integer Stochastic Programming (MMiSP) framework includes a planning phase and multiple operational decision phases. The planning phase considers the total investment and operation and maintenance costs of the energy storage system, as well as the expected operating costs for future operational decision phases. The power system operating year is divided into multiple operational decision phases, each of which may contain a number of operating days. These operational decision phases aim to evaluate system operating costs, renewable energy utilization rates, and the reliability of load supply.

[0101] Specifically, the operational scenarios (i.e., load and renewable energy output scenarios) for each operational decision-making stage can be divided into normal state and continuous low renewable energy output PLOS state. Between different operational decision-making stages, the operational state can transition between each other; it can transition from normal state to PLOS, from PLOS to normal state, from PLOS back to PLOS, or from normal state to normal state. The transition process between states is determined by Markov transition probabilities, and the Markov scenario tree expands exponentially with the number of stages. The resulting Markov scenario tree can contain a massive number of scenarios and also reflects the evolutionary dynamics of PLOS, describing the uncertainty of the timing, frequency, and duration of continuous low output processes. Simultaneously, the exchange of state variables between stages satisfies unpredictability, meaning that the decision-making basis does not include accurate future wind and solar load information, requiring the operational process to be prepared for possible PLOS processes in the next stage at any time, thus improving the robustness of the planning scheme. The state variables transferred between stages include the unit start / stop status and the SOC of the energy storage system.

[0102]

[0103] Assuming a stochastic process The process evolves dynamically between stages, where t represents the stage number, T is the stage number of the last stage, and s represents the state (normal state and PLOS, s=1 indicates the stage is in a normal state, s=2 indicates the stage is in a state of continuous low output of new energy). ξ s,t Let be a random variable in stage t under state s, including renewable energy output and load power. Let Q be the set of random variables in stage t under state s. t,s＝i (·) represents the future cost function with current uncertainty state i at stage t. t-1 Let c be the decision variable for stage t-1. t (ξ s＝i,t Let ξ be a known random variable. s＝i,t In this case, the cost vector at stage t. The time-varying state transition probability of the Markov model is P. state,i,j (t), and Q is also agreed upon T+1 (x T )≡0,x TQ is the decision variable for the final stage. T+1 This represents the total cost of the final stage.

[0104] To mitigate losses from high-impact, low-probability events, this embodiment employs a nested risk metric ρ(·) in equation (1), where λ, α ∈ [0, 1], λ is the risk weight, and α is the worst-case gradation, representing the worst-case loss at α% to describe the risk cost. This risk description integrates the expected cost and the conditional risk value function CVaR. α , which can be represented as a linear combination of them. A t B is the constraint coefficient matrix related to stage t. t b is the constraint coefficient matrix related to stage t-1. t (ξ s,t ) represents the constraint coefficient vector associated with the random variable. Equation (1) represents the compact model for each stage, and the objective function consists of two parts, the first part... This is the cost at the current stage, Part Two. This considers the Markov transition probabilities and the expected future costs. The latter half of the cost function... Specifically, it means: based on the possible states s in the next stage and the expected cost in the corresponding state, and also based on the state transition probability P. state,i,j We use (t) to estimate the expected future cost of the next stage. This form is difficult to solve directly. The decision objective at each stage is to rationally decide on the current use of resources and their reservation for the future under Markov transformation uncertainty, in order to minimize the total cost of the current stage and the expected future costs.

[0105] 2) Based on the Markov mixed-integer stochastic programming framework of step 1), establish a multi-type energy storage planning model that takes into account the scenario of continuous low output. The specific steps are as follows:

[0106] 2-1) Construct the planning phase sub-model.

[0107] In this embodiment, the first stage of the multi-type energy storage system planning model considering the PLOS Markov conversion process is the planning stage, whose objective is to minimize the expected values ​​of the annual total investment cost, operation and maintenance cost, and future operating cost. This stage, as the first stage of the multi-stage stochastic programming model, has an objective function as shown in equation (2), which includes the annual value of the energy storage system's investment cost, daily maintenance cost, and expected values ​​of future operating cost. The expected value of future operating cost is a linear combination of the expected operating costs under different states in the first operation decision stage, P... 0,j It is the probability of state j occurring in the first operational decision-making stage, which is also the probability of each state occurring in the first stage of the Markov process. Constraint equation (4) limits the maximum value of the energy storage to be planned.

[0108]

[0109] In the formula, the superscript z represents different energy storage types, and the subscript k represents the k-th node in the power system. Q0 represents the objective function for the planning stage, and Q1 represents the total cost of the first operational decision stage. C inv For the investment cost of energy storage, C om The operation and maintenance costs of energy storage. A y,z It is the discount rate for the investment in energy storage. and These represent the investment cost of the power capacity and the investment cost of the energy capacity of energy storage z, respectively. and These are the unit operation and maintenance costs related to energy storage power capacity and energy capacity, respectively. Let z be the power capacity of the energy storage to be planned at the k-th node. Let z be the energy capacity of the planned energy storage at the k-th node. and Let ρ(·) be the maximum power capacity and maximum energy capacity of the planned energy storage z at the k-th node. ρ(·) is a nested risk metric, and P... 0,j Let be the probability that the first operational decision stage is state j. To take the expectation of the variable in parentheses, Q1(x0,ξ) s,1 Let x0 be the total cost of the first operational decision-making phase, x0 be the planning result output of the planning phase, and ξ be the total cost of the first operational decision-making phase. s,1 This represents a random variable in the first operational decision-making stage. This means the expected total cost of the first operational decision-making stage when the state is j.

[0110] 2-2) Establish a multi-stage operational sub-model that considers the Markov transformation process.

[0111] In this embodiment, the multi-stage operation simulation divides the annual operation problem of the power system into several stages to consider the unexpected operation and the Markov transition process of the operation scenario. The objective function of each operation decision stage is Equation (5), which minimizes the sum of the current operation decision stage's operating cost and the expected future cost. The operating cost calculation expression for the current operation decision stage is shown in Equation (6). The operating cost of the current operation decision stage includes the power generation cost, start-up and shutdown cost, wind and solar curtailment cost, and load curtailment cost of the thermal power unit, while also considering unit operation constraints and power flow constraints. By considering the uncertainty of the occurrence time, duration, and frequency of PLOS through the Markov process, each subsequent operation decision stage may be a normal scenario or a PLOS scenario, which is determined by the state transition probability. Therefore, during the operation, preventive scheduling is required for the possible occurrence of PLOS in the next stage. The expression of the multi-stage operation sub-model is as follows:

[0112]

[0113] In the formula, Q t,s＝i Let C represent the objective function in state s = i during the t-th stage of the decision-making process. shed,k This represents the penalty cost for abandoning load. (C) start,k and C gen,k S represents the start-up cost and power generation cost of the thermal power unit located at node k, respectively. rated,k γ is the rated capacity of the thermal power unit located at node k. w and γ pv These represent the unit penalty cost for curtailing wind and solar power. and These represent the startup status and generating power of the thermal power unit located at node k, respectively. The subscript d represents days, and τ represents the minimum scheduling time interval. Let t represent the set of all days in the t-th stage of the decision-making process. and P represents the curtailed wind power and curtailed solar power during the time period τ. shed,d,τ,k This indicates the power of the abandoned load.

[0114] In this embodiment, the constraints of each operational decision-making stage can be divided into intra-stage constraints and inter-stage constraints. Intra-stage constraints involve variables that are only related to the current stage, while inter-stage constraints involve state variables across stages. Specifically:

[0115] In-phase constraints:

[0116]

[0117]

[0118] P l,d,τ =(δ k1,d,τ -δ k2,d,τ ) / X k1,k2 (twenty one)

[0119]

[0120] δ 0,d,τ =0 (23)

[0121] In the formula, This indicates the online status of the thermal power unit at node k, where 1 indicates the unit is online and 0 indicates the unit is offline. These represent the start-up and shutdown states of the unit at node k, respectively, both being 0-1 variables. μ k These represent the maximum and minimum generating capacities of the thermal power unit at node k, respectively. RU k and RDk T represents the maximum uphill and downhill ramp rates of the thermal power unit at node k. on,k and T off,k The minimum start-up and shutdown times of the thermal power unit at node k. These represent the self-discharge, charging, and discharging efficiencies of the energy storage system, respectively. and Let z represent the power capacity and energy capacity of energy storage z at node k, respectively. These represent the charging power and discharging power of the energy storage system, respectively. These represent the intraday and interday energy levels of the energy storage system, respectively. Ω y Ω m Ω d These represent sets of energy storage types applicable to different balance cycles. M start (m) and M end (m) represents the first and last day of month m, respectively. |τ| represents the number of time periods in a day. P w,d,τ ,P PV,d,τ These represent the power generation capacity of wind and solar energy, respectively. The superscript symbols (*) indicate the available power generation capacity of wind and solar energy, determined by wind speed and solar irradiance, respectively. β up ,α up ,β dn ,α dn It is the peak shaving reserve coefficient related to new energy sources and load. P 1oad,d,τ,k Let P be the electrical power consumption of the load at node k. shed,d,τ,k P represents the power of the load being discarded at node k. l,d,τ δ represents the power flow on branch l. k,d,τ This represents the phase angle at node k. The superscripts min and max represent the minimum and maximum allowable values, respectively. δ 0,d,τ X represents the phase angle of the reference node. k1,k2 This represents the reactance of the tie line from node k1 to node k2.

[0122] Constraints (8)-(10) define the combination of thermal power units, where equation (8) calculates the online status of the unit through unit start-up and shutdown, equation (9) limits the up-and-down ramping capability of the thermal power unit, and equation (10) limits the frequency of frequent start-up and shutdown of the unit. Constraints (11)-(16) define the joint operation of multiple types of energy storage, which are divided into annual balancing type, monthly balancing type and daily balancing type. Daily balancing energy storage is used to handle intraday energy imbalance, while monthly balancing type and annual balancing type energy storage are used to handle the cross-seasonal imbalance of new energy and load. More importantly, they provide load support during PLOS by leveraging their larger energy capacity.

[0123] In this embodiment, the stored energy is divided into intraday energy and interday energy. Intraday energy is balanced daily, while interday energy is balanced in different balancing cycles according to the type of energy storage. Equation (11) calculates the intraday energy change of the stored energy through the energy storage charging and discharging power. Equations (12)-(13) limit the upper and lower limits of the energy storage charging and discharging power and the stored energy. Equation (14) specifies that the intraday energy is balanced once a day. Equation (15) links the intraday energy with the interday energy, that is, the intraday energy is the remaining value of the interday energy of the previous day plus the intraday unbalanced energy of the current day. Equation (16) defines the balancing cycle of different types of energy storage. Equation (17) sets the upper and lower limits of the power generation of new energy sources. Equations (18)-(19) ensure the spinning reserve capacity of the system. Equation (20) is the system node power balance constraint. Equations (21)-(23) are the system DC power flow constraints.

[0124] Inter-phase constraints:

[0125]

[0126] In the formula, set Ω on,t,t-1 and Ω off,t,t-1 These represent the last T of stage t-1, respectively. on,k Time period and T off,k The time periods encompass all time periods within stage t-1 related to the operation of the thermal power unit located at node k in stage t. end For a set of dates The last day. Ω t It is the set of all time intervals in stage t. Constraint (10) and constraint (25) together form the minimum start-up constraint and minimum shutdown duration constraint of the thermal power unit. Constraint (24) and constraint (26) ensure that the start-up / shutdown state and energy storage energy of the unit do not change abruptly between stages, and are also an important guarantee for meeting the unpredictability of operation.

[0127] In this embodiment, all intra-stage and inter-stage constraints can be uniformly represented in a compact form as shown in equation (27). t A is a coefficient matrix related to the decision variables of the previous stage. t b is the coefficient matrix related to the decision variables at the current stage. t x is a constant vector representing the constraints. t-1 and x t Let be the decision variables for stages t-1 and t.

[0128] B t x t-1 +A t x t ≥b t (27)

[0129] 2-3) Based on the results of steps 2-1) and 2-2), establish a complete multi-type energy storage planning model that takes into account the scenario of continuous low output.

[0130] In this embodiment, based on the results of steps 2-1) and 2-2), a complete multi-type energy storage planning model considering the continuous low-output scenario is obtained. This model includes the first stage (planning stage) sub-model equations (2)-(4), and subsequent multiple operation decision stage sub-model equations (5)-(26) considering operational unpredictability and the dynamic evolution of the continuous low-output process. Each stage of the entire planning model can be written in a compact form as shown in equation (28):

[0131]

[0132] B t x t-1 +A t x t ≥b t

[0133] 3) Solve the model established in step 2) using the mixed integer stochastic dual dynamic programming algorithm.

[0134] The model established in step 2) is a multi-stage mixed integer stochastic programming (MMiSP) problem with a large number of scenarios. The solution steps are as follows:

[0135] 3-1) Reconstruct the nested multi-stage conditional risk values ​​of the model established in step 2).

[0136] In this embodiment, based on the conditional risk value (CVaR) expression in equation (29), ρ(·) in equation (1) can be further expressed as shown in equation (30). Assume V is a discrete random variable, and each random variable takes the specific value v. l v l Let V represent the specific values ​​of the l-th random variable, and there are a total of L values. Then the expectation of the discrete random variable V can be expressed as (1 / L)Σv l In the formula, b is the auxiliary variable introduced to linearize the conditional risk value function, and α is the scale value for the worst-case scenario, meaning the loss at the worst α% level. This is achieved by introducing auxiliary constraints and the auxiliary variable δ. l With the auxiliary variable υ, equation (30) can be reconstructed into a linear programming form, as shown in equation (31), where v l This is the future cost function in equation (1).

[0137]

[0138] 3-2) Reconstruct the model after step 3-1) using a Markov process to transform the original model into a multi-stage mixed integer linear programming problem.

[0139] In this embodiment, in order to introduce the inter-stage dependency of random variables and Markov processes into multi-stage stochastic optimization, the outer layer ρ(·) in equation (1) is first ignored. Assuming that the stochastic programming problem has a finite support set in each stage, and each data sample is numbered by the subscript o, the conditional probability can be discretized and expressed in the form of equation (32), where ξ s,t,o It is the o-th data sample of state s in stage t. θ is the total number of data samples in state s at stage t+1. t+1,j,o As an auxiliary variable, it is used to replace the discretized future cost function. Equation (32) means that when the current stage state s = i is known, all possibilities of the next stage state are considered to minimize the total expected cost.

[0140]

[0141] Then, the future cost function Q can be estimated using the cutting plane. t+1,s＝j,o The lower bound of the cost function. This embodiment uses a multi-cutting plane lower bound estimation method. By introducing the cutting plane to estimate the future cost function, equation (32) can be further constructed into the form of equations (33)-(34). It is the attempted solution in the t-th iteration, π t,s＝i,o (ξ s＝i,t,o ) is the dual variable of the cross-stage constraint. The solution to the relaxation problem is the lower bound of the solution to the original problem (32) because the cutting plane is used instead of the future cost function. Let be the set of all cut planes for state s and data sample o in stage t+1. For stage t, the variables with subscripts t-1 and t+1 are parameters known in stage t, which have been fed back through forward sampling and backward backtracking. Forward sampling passes the state variables of stage t-1 to stage t, and backward backtracking passes the cut planes generated in t+1 to stage t.

[0142]

[0143] By reintroducing ρ(·) and substituting equations (29)-(31) into equations (33)-(34), we can derive the form shown in equation (35). Thus, the original problem equation (1) is reconstructed into a multi-stage mixed integer linear programming problem.

[0144]

[0145] 3-3) Solve the multi-stage mixed integer linear programming problem obtained in step 3-2) using the mixed integer stochastic dual dynamic programming algorithm.

[0146] In equation (35), the presence of integer variables hinders the generation of the dual variable, making it difficult to estimate the lower bound of the future cost function. To address this issue, it is necessary to modify the generation process of the dual variable and the method for generating the cutting plane. First, an auxiliary constraint equation (36) is introduced, where z t-1 For continuous variables, x is a state variable. t-1 A local copy in stage t. The constraints of each stage are further updated as shown in equation (37). This step is to centralize the representation of cross-stage variables to facilitate the extraction of dual information.

[0147] z t-1 =x t-1 ,[π t,i,d (ξ s＝i,t,o (36)

[0148] A t x t ≥b t (ξ s＝i,t,o )-B t z t-1 (37)

[0149] The general approach to solving multi-stage mixed integer problems is to construct a relaxation problem of the original problem, thereby obtaining the objective function and corresponding dual variable of the relaxation problem, and ultimately generating a cutting plane. Different methods of constructing relaxation problems correspond to different types of cutting planes, and the tightness of different cutting planes varies.

[0150]

[0151] By relaxing the integer variables of the original problem into continuous variables and solving the linear programming problem The Benders' cutting plane can be obtained as shown in equation (38), and the dual variables are obtained. and objective function The Lagrange cut plane requires solving a max-min bilevel optimization problem (40) in the backward backtracking, also known as the Lagrange relaxation problem. The outer optimization maximizes the objective function by finding the optimal dual variable, while the inner optimization is the original integer optimization problem for each stage. The generated cutting plane contains a combination of the optimal dual variable and the optimal objective function value (Equation (41)).

[0152] Furthermore, since solving the Lagrange problem is computationally expensive, a combination of the Lagrange problem and the linear relaxation problem can be used to generate the enhanced Benders' cutting plane formula (39). First, the dual variables are obtained from the linear relaxation, and then they are substituted into the Lagrange relaxation problem to directly solve the inner minimization problem in order to calculate the optimal objective function value. The problem is called Although the enhanced Benders' cut plane is not as tight as the Lagrange cut plane, its generation process is computationally faster, and its overall convergence speed is worse than using the Lagrange cut plane alone. Since the Lagrange relaxation problem is computationally slow, this invention combines the Benders' cut plane and the enhanced Benders' cut plane to efficiently solve the multi-stage integer programming problem.

[0153] Therefore, the MMiSP proposed in this invention can be solved through forward sampling and backward backtracking. In forward sampling, sampling is performed sequentially from stage 1 to T based on Markov transition probabilities. The basic method is to obtain the state of stage t, and then obtain the state of stage t+1 based on the transition probabilities from stage t to stage t+1 and the state of stage t, and so on. The state transition probabilities can be estimated using maximum likelihood estimation. For each stage, once the normal state or PLOS state is determined through sampling, specific data samples of load and renewable energy generation can be selected from the scenario set of the corresponding state. This invention uses an equal probability sampling method to sample from the state scenario set. The sampled specific scenario (implemented with uncertain information) and the final value x of the state variable from the previous stage are compared. t-1 The inputs are combined into the optimization problem (35) to obtain a trial solution. In the backtracking, a cutting plane is constructed for each trial solution to provide an estimate of the lower bound of the future cost function, which is then fed back to the previous stage. The above process is repeated until the algorithm converges. In each iteration, the methods to deal with the uncertainty of the future stage are stored in the form of cutting planes, approximating the real future cost while optimizing the running decision process to deal with the risk of unload. Since the reconstructed problem (35) is a multi-stage mixed integer linear programming problem, the convergence of SDDiP can be guaranteed. However, considering the nested CVaR, the estimation of the upper bound becomes very complicated due to the mismatch between the sampling probability and the probability distribution of the extreme scenarios considered by CVaR. Therefore, this embodiment sets the convergence criterion as: when the relative change of the lower bound in 20 iterations is less than 10. -4 The algorithm is considered convergent when the time is reached. The solution algorithm is as follows:

[0154] 3-3-1) Input the initial state x ini and random data sample ξ s,t Set the iteration count ite = 0, the lower bound LB = -∞, and the flag = 0.

[0155] 3-3-2) Set ite = ite + 1; For all stages, from stage 1 to T, perform forward sampling, first performing state sampling: based on the state s of stage t-1. t-1 And Markov transition probability simulation or sampling of t-stage state s t Then perform scene sampling: from A sample of uncertain parameter ξ is randomly selected from the samples. s,t,o And solve equation (35) to obtain an attempted solution.

[0156] 3-3-3) Determine if ite is N ite Integer multiples of:

[0157] If yes, then proceed to step 3-3-4; otherwise, proceed to step 3-3-5.

[0158] 3-3-4) Perform a backward backtracking from stage T to all stages t in stage 2, for all possible states s. t as well as Each data sample ξ s,t,o Solve The corresponding dual variables and objective function values ​​are obtained, and the cutting plane formula (39) is generated and passed to stage t-1.

[0159] 3-3-5) Perform a backward backtracking from stage T to all stages t in stage 2, for all possible states s. t as well as Each data sample ξ s,t,o Solve The corresponding dual variables and objective function values ​​are obtained, and the cutting plane expression (38) is generated and passed to stage t-1.

[0160] 3-3-6) Update the lower bound by taking the larger of the total cost of the current iterative planning stage and the current lower bound LB.

[0161] 3-3-7) If the convergence criterion is met, set flag = 1 and output the planning scheme x0.

[0162] Furthermore, to improve the efficiency of solving mixed-integer multi-stage problems, the method described in this embodiment introduces three acceleration methods: adding a regularization term to the objective function to prioritize exploring neighboring regions; using dynamic sampling to capture high-risk, low-probability events and explore possible combinations of extreme event occurrences; and using parallelization to reduce the runtime of each iteration. It should be noted that the following acceleration strategies are optional steps, but implementing all of them will help reduce the algorithm's solution time; one or more acceleration methods can also be selected.

[0163] a) Regularization

[0164] A regularization term (43) is added to the objective function of equation (35) to penalize attempts that deviate significantly from the current optimal solution. The core idea of ​​this method is to prioritize exploring feasible regions with more current information in order to find new optimal solutions. The penalty coefficient γ gradually decreases with the iteration process, thereby improving the convergence speed of the algorithm without affecting the optimality and feasibility of the solution.

[0165]

[0166] In the formula, It is a positive semi-definite matrix, usually a diagonal matrix, and needs to be determined according to R. t The range of values ​​for R assigns different weights to data of different dimensions. t As an auxiliary variable, its meaning is the part B in the stage t constraint that is related to stage t-1. t x t-1 That is, coefficient matrix B t Compared with the decision variable x in the previous stage t-1 The product of. R in the previous loop t The value of .

[0167] b) Dynamic sampling and parallelization

[0168] Due to the low probability nature of PLOS events, sampling solely based on their transition probabilities may limit scenario exploration, failing to fully exploit the uncertainties in PLOS occurrence time, duration, and frequency, while also slowing down convergence. Therefore, a dynamic sampling method is necessary, continuously adjusting the sampling probabilities of scenarios and states during algorithm execution: initially, the transition probabilities of normal and PLOS states are considered equal. Then, after a significant increase in the lower bound, sampling is performed based on the actual transition probabilities of the two states. Parallel computation can be employed in different sampling paths during forward sampling and in different data sample scenarios during backward backtracking to significantly improve computational speed.

[0169] To achieve the above embodiments, a second aspect of the present invention proposes a multi-type energy storage planning device considering continuous low-output scenarios, comprising:

[0170] A Markov process programming framework construction module is used to establish a stochastic programming framework that takes into account the continuous low output Markov process of new energy sources. The framework includes a planning stage and multiple operation decision stages, which are obtained by dividing the power system operation year.

[0171] The energy storage planning model construction module is used to establish a multi-type energy storage planning model that takes into account the continuous low output scenario based on the Markov process planning framework. The planning model includes a planning stage sub-model and a multi-stage operation sub-model that considers the Markov transformation process.

[0172] The planning scheme generation module is used to solve the multi-type energy storage planning model to obtain planning schemes for the multi-type energy storage.

[0173] It should be noted that the foregoing explanation of an embodiment of a multi-type energy storage planning method considering a scenario of continuous low output also applies to a multi-type energy storage planning device considering a scenario of continuous low output in this embodiment, and will not be repeated here. According to an embodiment of the present invention, a multi-type energy storage planning device considering a scenario of continuous low output is proposed by establishing a stochastic programming framework that considers the Markov process of continuous low output of new energy sources. The framework includes a planning stage and multiple operational decision stages, which are obtained by dividing the power system operating year. Based on the Markov process programming framework, a multi-type energy storage planning model considering the scenario of continuous low output is established. The planning model includes a planning stage sub-model and a multi-stage operational sub-model considering the Markov transformation process. Solving the multi-type energy storage planning model yields planning schemes for multiple types of energy storage. This allows for a more realistic depiction of the dynamic evolution of the continuous low output process of new energy sources, increases the robustness and rationality of energy storage planning, and reduces the risk of system load shedding.

[0174] To implement the above embodiments, a third aspect of the present invention provides an electronic device, comprising:

[0175] At least one processor; and a memory communicatively connected to said at least one processor;

[0176] The memory stores instructions that can be executed by the at least one processor, and the instructions are configured to execute the above-described multi-type energy storage planning method considering a continuous low-output scenario.

[0177] To implement the above embodiments, a fourth aspect of the present invention provides a computer-readable storage medium storing computer instructions for causing the computer to execute the above-described multi-type energy storage planning method considering a continuous low-output scenario.

[0178] It should be noted that the computer-readable medium described in this disclosure can be a computer-readable signal medium or a computer-readable storage medium, or any combination thereof. A computer-readable storage medium can be, for example,—but not limited to—an electrical, magnetic, optical, electromagnetic, infrared, or semiconductor system, apparatus, or device, or any combination thereof. More specific examples of a computer-readable storage medium may include, but are not limited to: an electrical connection having one or more wires, a portable computer disk, a hard disk, random access memory (RAM), read-only memory (ROM), erasable programmable read-only memory (EPROM or flash memory), optical fiber, portable compact disk read-only memory (CD-ROM), optical storage device, magnetic storage device, or any suitable combination thereof. In this disclosure, a computer-readable storage medium can be any tangible medium containing or storing a program that can be used by or in connection with an instruction execution system, apparatus, or device. In this disclosure, a computer-readable signal medium can include a data signal propagated in baseband or as part of a carrier wave, carrying computer-readable program code. Such propagated data signals can take various forms, including but not limited to electromagnetic signals, optical signals, or any suitable combination thereof. A computer-readable signal medium can be any computer-readable medium other than a computer-readable storage medium, which can send, propagate, or transmit a program for use by or in connection with an instruction execution system, apparatus, or device. The program code contained on the computer-readable medium can be transmitted using any suitable medium, including but not limited to: wires, optical fibers, RF (radio frequency), etc., or any suitable combination thereof.

[0179] The aforementioned computer-readable medium may be included in the aforementioned electronic device; or it may exist independently and not assembled into the electronic device. The aforementioned computer-readable medium carries one or more programs, which, when executed by the electronic device, cause the electronic device to perform a multi-type energy storage planning method considering a continuous low-output scenario as described in the above embodiments.

[0180] Computer program code for performing the operations of this disclosure can be written in one or more programming languages ​​or a combination thereof, including object-oriented programming languages ​​such as Java, Smalltalk, and C++, and conventional procedural programming languages ​​such as the "C" language or similar programming languages. The program code can be executed entirely on the user's computer, partially on the user's computer, as a standalone software package, partially on the user's computer and partially on a remote computer, or entirely on a remote computer or server. In cases involving remote computers, the remote computer can be connected to the user's computer via any type of network—including a local area network (LAN) or a wide area network (WAN)—or can be connected to an external computer (e.g., via the Internet using an Internet service provider).

[0181] In the description of this specification, the references to terms such as "one embodiment," "some embodiments," "example," "specific example," or "some examples," etc., refer to specific features, structures, materials, or characteristics described in connection with that embodiment or example, which are included in at least one embodiment or example of this application. In this specification, the illustrative expressions of the above terms do not necessarily refer to the same embodiment or example. Furthermore, the specific features, structures, materials, or characteristics described may be combined in any suitable manner in one or more embodiments or examples. Moreover, without contradiction, those skilled in the art can combine and integrate the different embodiments or examples described in this specification, as well as the features of different embodiments or examples.

[0182] Furthermore, the terms "first" and "second" are used for descriptive purposes only and should not be construed as indicating or implying relative importance or implicitly specifying the number of technical features indicated. Thus, a feature defined as "first" or "second" may explicitly or implicitly include at least one of that feature. In the description of this application, "multiple" means at least two, such as two, three, etc., unless otherwise explicitly specified.

[0183] Any process or method described in the flowchart or otherwise herein can be understood as representing a module, segment, or portion of code comprising one or more executable instructions for implementing a particular logical function or process, and the scope of the preferred embodiments of this application includes additional implementations in which functions may be performed not in the order shown or discussed, including substantially simultaneously or in reverse order depending on the function involved, as will be understood by those skilled in the art to which embodiments of this application pertain.

[0184] The logic and / or steps represented in the flowchart or otherwise described herein, for example, can be considered as a sequenced list of executable instructions for implementing logical functions, and can be embodied in any computer-readable medium for use by, or in conjunction with, an instruction execution system, apparatus, or device (such as a computer-based system, a processor-included system, or other system that can fetch and execute instructions from, an instruction execution system, apparatus, or device). For the purposes of this specification, "computer-readable medium" can be any means that can contain, store, communicate, propagate, or transmit programs for use by, or in conjunction with, an instruction execution system, apparatus, or device. More specific examples (a non-exhaustive list) of computer-readable media include: an electrical connection having one or more wires (electronic device), a portable computer disk drive (magnetic device), random access memory (RAM), read-only memory (ROM), erasable and editable read-only memory (EPROM or flash memory), fiber optic devices, and portable optical disc read-only memory (CDROM). Furthermore, computer-readable media can even be paper or other suitable media on which programs can be printed, because programs can be obtained electronically, for example, by optically scanning the paper or other media, followed by editing, interpreting, or otherwise processing as necessary, and then stored in computer memory.

[0185] It should be understood that various parts of this application can be implemented using hardware, software, firmware, or a combination thereof. In the above embodiments, multiple steps or methods can be implemented using software or firmware stored in memory and executed by a suitable instruction execution system. For example, if implemented in hardware, as in another embodiment, it can be implemented using any one or a combination of the following techniques known in the art: discrete logic circuits having logic gates for implementing logical functions on data signals, application-specific integrated circuits (ASICs) having suitable combinational logic gates, programmable gate arrays (PGAs), field-programmable gate arrays (FPGAs), etc.

[0186] Those skilled in the art will understand that all or part of the steps of the methods in the above embodiments can be implemented by a program instructing related hardware. The program can be stored in a computer-readable storage medium, and when executed, the program includes one or a combination of the steps of the method embodiments.

[0187] Furthermore, the functional units in the various embodiments of this application can be integrated into a processing module, or each unit can exist physically separately, or two or more units can be integrated into a module. The integrated module can be implemented in hardware or as a software functional module. If the integrated module is implemented as a software functional module and sold or used as an independent product, it can also be stored in a computer-readable storage medium.

[0188] The storage medium mentioned above can be a read-only memory, a disk, or an optical disk, etc. Although embodiments of this application have been shown and described above, it is understood that the above embodiments are exemplary and should not be construed as limiting this application. Those skilled in the art can make changes, modifications, substitutions, and variations to the above embodiments within the scope of this application.

Claims

1. A multi-type energy storage planning method considering scenarios with continuous low output, characterized in that, include: A stochastic programming framework is established that takes into account the Markov process of continuous low output of new energy sources. The framework includes a planning stage and multiple operation decision stages, which are obtained by dividing the power system operation year. Based on the Markov process planning framework, a multi-type energy storage planning model that takes into account the continuous low output scenario is established. The planning model includes a planning stage sub-model and a multi-stage operation sub-model that considers the Markov transformation process. Solving the multi-type energy storage planning model yields planning schemes for the multi-type energy storage.

2. The method according to claim 1, characterized in that, The stochastic programming framework expression for the Markov process that takes into account the continuous low output of new energy sources is as follows: Among them, let the random process The process evolves dynamically between stages, where t represents the stage number and T is the stage number of the last stage; s represents the state, s=1 indicates that the stage is in a normal state, and s=2 indicates that the stage is in a state of continuous low output of new energy; ξ s,t Let be a random variable in stage t under state s, including renewable energy output and load power. Let Q be the set of random variables in stage t and state s; t,s＝i (·) represents the future cost function with current uncertainty state i at stage t; x t-1 c is the decision variable for stage t-1; t (ξ s＝i,t Let ξ be a random variable. s＝i,t In this case, the cost vector of stage t; P state,i,j Q(t) represents the time-varying state transition probability of the Markov model. T+1 (x T )≡0,x T Q is the decision variable for the final stage. T+1 This represents the total cost of the final stage. ρ(·) is a nested risk measure, where λ, α ∈ [0, 1], λ is the risk weight, and α is the gradation value for the worst-case scenario; A t B is the constraint coefficient matrix related to stage t. t b is the constraint coefficient matrix related to stage t-1. t (ξ s,t () represents the constraint coefficient vector associated with the random variable; In equation (1), This indicates the cost at the current stage. This represents the expected future cost considering Markov transformation probabilities.

3. The method according to claim 2, characterized in that, The expression for the planning phase sub-model is as follows: In the formula, the superscript z represents different energy storage types, the subscript k represents the k-th node in the power system; Q0 represents the objective function of the planning stage, Q1 represents the total cost of the first operation decision stage; C inv For the investment cost of energy storage, C om For the operation and maintenance costs of energy storage; A y,z It is the discount rate for the investment in energy storage. and Let z represent the investment cost of the power capacity and the investment cost of the energy capacity of energy storage z, respectively. and These are the unit operation and maintenance costs related to energy storage power capacity and energy capacity, respectively. Let z be the power capacity of the energy storage to be planned at the k-th node. Let z be the energy capacity of the planned energy storage at the k-th node. and Let P be the maximum power capacity and maximum energy capacity of the planned energy storage z-th node; 0,j Let be the probability that the first operational decision stage is state j. To take the expectation of the variable in parentheses, Q1(x0,ξ) s,1 Let x0 be the total cost of the first operational decision-making phase, x0 be the planning result output of the planning phase, and ξ be the total cost of the first operational decision-making phase. s,1 Represents the random variable in the first operational decision-making stage; This represents the expected total cost of the first operational decision-making stage when the state is j.

4. The method according to claim 3, characterized in that, The objective function expression for the multi-stage operational sub-model considering the Markov transformation process is as follows: In the formula, Q t,s＝i C represents the objective function in state s = i during the t-th running decision stage; shed,k Indicates the penalty cost for abandoning load; C start,k and C gen,k S represents the start-up cost and power generation cost of the thermal power unit located at node k, respectively; rated,k γ represents the rated capacity of the thermal power unit located at node k; w and γ pv These represent the unit penalty cost for curtailing wind and solar power; and These represent the startup status and generating power of the thermal power unit located at node k, respectively; the subscript d represents days, and τ represents the minimum scheduling time interval. Represents the set of all days in the t-th stage of the decision-making process; and P represents the curtailed wind power and curtailed solar power during the time period τ. shed,d,τ,k Indicates the power of abandoned load; The constraints of the multi-stage operating sub-model considering the Markov transformation process include intra-stage constraints and inter-stage constraints. The intra-stage constraints include: P l,d,τ =(δ k1,d,τ -d k2,d,τ ) / X k1,k2 (21) d 0,d,τ =0 (23) In the formula, This indicates the online status of the thermal power unit at node k, where 1 indicates the unit is online and 0 indicates the unit is offline. These represent the start-up and shutdown states of the unit at node k, respectively, both being 0-1 variables; μ k These represent the maximum and minimum generating capacities of the thermal power unit at node k, respectively; RU k and RD k These represent the maximum uphill and downhill ramp rates of the thermal power unit at node k; T on,k and T off,k Minimum start-up and shutdown times for the thermal power unit at node k; These represent the self-discharge, charging, and discharging efficiencies of the energy storage system, respectively. and Let z represent the power capacity and energy capacity of energy storage z at node k, respectively. These represent the charging power and discharging power of the energy storage system, respectively. These represent the intraday and interday energy of the energy storage system, respectively; Ω y Ω m Ω d These represent sets of energy storage types applicable to different balance cycles; M start (m) and M end (m) represents the first and last day of month m, respectively; |τ| represents the number of time periods in a day; P w,d,τ ,P PV,d,τ These represent the generating capacity of wind and solar power, respectively; the symbols marked with an asterisk (*) indicate the available generating capacity of wind and solar power, respectively; β up ,α up ,β dn ,α dn It is the peak shaving reserve factor related to new energy sources and load; P 1oad,d,τ,k Let P be the electrical power consumption of the load at node k. shed,d,τ,k P represents the power of the load being discarded at node k; l,d,τ δ represents the power flow on branch l. k,d,τ The phase angle at node k is represented; the superscripts min and max represent the minimum and maximum allowable values, respectively; δ 0,d,τ X represents the phase angle of the reference node. k1,k2 This represents the reactance of the tie line from node k1 to node k2; The inter-stage constraints include: In the formula, set Ω on,t,t-1 and Ω off,t,t-1 These represent the last T of stage t-1, respectively. on,k Time period and T off,k Time period; d end For a set of dates The last day; Ω t It is the set of all time intervals in stage t.

5. The method according to claim 4, characterized in that, Also includes: The intra-stage constraints and inter-stage constraints are expressed in a compact form as shown in equation (27): B t x t-1 +A t x t ≥b t (27) Among them, B t A is a coefficient matrix related to the decision variables of the previous stage. t b is the coefficient matrix related to the decision variables at the current stage. t x is a constant vector representing the constraints; t-1 and x t Let be the decision variables for stages t-1 and t.

6. The method according to claim 5, characterized in that, Also includes: The compact form of the multi-type energy storage planning model that takes into account the continuous low output scenario is as follows:

7. The method according to claim 6, characterized in that, The solution to the multi-type energy storage planning model includes: 1) Reconstruct multi-stage conditional risk values ​​by nesting multiple types of energy storage planning models; According to the conditional risk value expression in equation (29), ρ(·) in equation (1) is expressed as shown in equation (30); Let V be a discrete random variable, and let each random variable take the value v. l v l Let V represent the specific values ​​of the l-th random variable, and there are a total of L values. Then the expected value of the discrete random variable V is expressed as (1 / L)Σv. l In the formula, b is the auxiliary variable introduced to linearize the conditional risk value function; this is achieved by introducing auxiliary constraints and the auxiliary variable δ. l With the auxiliary variable υ, equation (30) is reconstructed into a linear programming form as shown in equation (31), where v l The future cost function in equation (1); 2) Reconstruct the model after step 1) using a Markov process to transform the original model into a multi-stage mixed integer linear programming problem; First, ignore the outer ρ(·) in equation (1); let the stochastic programming problem have a finite support set in each stage, and let each data sample be numbered by the subscript o. Then the conditional probability can be discretized and expressed in the form of equation (32), where ξ s,t,o It is the o-th data sample of state s in stage t. θ is the total number of data samples in state s at stage t+1. t+1,j,o As an auxiliary variable, it replaces the discretized future cost function; Estimating the future cost function Q using the cutting plane t+1,s＝j,o The lower bound of equation (32) is further constructed into the form of equations (33)-(34); where, It is the attempted solution in the t-th iteration, π t,s＝i,o (ξ s＝i,t,o ) is the dual variable of the cross-stage constraint; This is a relaxation problem of the original problem (32); Let be the set of all cutting planes for state s in stage t+1 and data sample o; By reintroducing ρ(·) and substituting equations (29)-(31) into equations (33)-(34), we obtain the form shown in equation (35). Thus, the original problem equation (1) is reconstructed into a multi-stage mixed integer linear programming problem. 3) Solve the multi-stage mixed-integer linear programming problem obtained in step 2) using the mixed-integer stochastic dual dynamic programming algorithm; First, we introduce auxiliary constraint equation (36), where z t-1 x is a continuous variable and a state variable. t-1 The local copy in stage t; the constraints of each stage are updated as shown in equation (37); z t-1 =x t-1 ,[p t,i,d (x) s＝i,t,o )] (36) A t x t ≥b t (ξ s＝i,t,o )-B t z t-1 (37) By relaxing the integer variables of the original problem into continuous variables and solving the linear programming problem The Benders' cutting plane is obtained as shown in equation (38), including the dual variables. and objective function The Lagrange cut plane is used to solve a max-min bilevel optimization problem (40) in backward backtracking, namely the Lagrange relaxation problem. The outer optimization maximizes the objective function by finding the optimal dual variable, while the inner optimization is the original integer optimization problem of each stage; the generated cutting plane contains a combination of the optimal dual variable and the optimal objective function value as shown in Equation (41); 8. The method according to claim 7, characterized in that, The specific steps of the mixed-integer random dual dynamic programming algorithm are as follows: 3-1) Input the initial state x ini and random data sample ξ s,t Set the iteration count ite = 0, the lower bound LB = -∞, and the flag = 0; 3-2) Set ite = ite + 1; For all stages, from stage 1 to T, perform forward sampling, first performing state sampling: based on the state s of stage t-1. t-1 And Markov transition probability simulation or sampling of t-stage state s t Then perform scene sampling: from A sample of uncertain parameter ξ is randomly selected from the samples. s,t,o And solve equation (35) to obtain an attempted solution. 3-3) Determine if ite is N ite Integer multiples of: If yes, proceed to step 3-4); otherwise, proceed to step 3-5. 3-4) Perform a backward backtracking for all possible states s from stage T to stage 2. t as well as Each data sample ξ s,t,o Solve The corresponding dual variables and objective function values ​​are obtained, and the cutting plane expression (39) is generated and passed to stage t-1; 3-5) Perform a backward backtracking from stage T to all stages t in stage 2, for all possible states s. t as well as Each data sample ξ s,t,o Solve The corresponding dual variables and objective function values ​​are obtained, and the cutting plane expression (38) is generated and passed to stage t-1; 3-6) Update the lower bound by taking the larger value of the total cost of the current iterative planning stage and the current lower bound LB; 3-7) If the convergence criterion is met, set flag = 1 and output the planning scheme x0.

9. The method according to claim 8, characterized in that, Also includes: The solution efficiency of the model can be improved by the following methods: 1) Add regularization term (43) to the objective function of equation (35) to punish attempts that deviate significantly from the current optimal solution; In the formula, It is a positive semi-definite matrix; R t As an auxiliary variable, it represents the portion B in the stage t constraint that is related to stage t-1. t x t-1 That is, coefficient matrix B t Compared with the decision variable x in the previous stage t-1 The product; R in the previous loop t The value of ; 2) A dynamic sampling method is adopted to continuously adjust the sampling probability of the scene and state during the algorithm operation; At the beginning of the algorithm's operation, the transition probabilities of the normal state and the PLOS state of continuous low output of new energy are considered to be equal; after the lower bound increases significantly, sampling is performed based on the actual transition probabilities of the two states. 3) Parallel computing is used to improve computing speed in different sampling paths of forward sampling and different data sample scenarios of backward backtracking.

10. A multi-type energy storage planning device considering continuous low-output scenarios, characterized in that, include: A Markov process programming framework construction module is used to establish a stochastic programming framework that takes into account the continuous low output Markov process of new energy sources. The framework includes a planning stage and multiple operation decision stages, which are obtained by dividing the power system operation year. The energy storage planning model construction module is used to establish a multi-type energy storage planning model that takes into account the continuous low output scenario based on the Markov process planning framework. The planning model includes a planning stage sub-model and a multi-stage operation sub-model that considers the Markov transformation process. The planning scheme generation module is used to solve the multi-type energy storage planning model to obtain planning schemes for the multi-type energy storage.

Images