Risk-driven high-proportion new energy power system electric quantity standby decision-making method and system

Abstract

The invention discloses a risk-driven high-proportion new energy power system electric quantity reserve decision-making method and system, and belongs to the technical field of power system operation. The method comprises the following steps: carrying out operation risk definition of a VRE admission domain for a power system, and learning a mapping relation between an uncertain set boundary and an operation risk for the power system by using a conservative sparse neural network CSNN after definition; constructing a risk-driven adaptive uncertainty set; based on the linearization model and the adaptive uncertainty set, establishing a risk look-ahead robust scheduling framework through offline training and online mapping modes; and using the risk prospective robust scheduling framework to identify the operation risk of the power system through an online mapping mode, generating a current power reserve decision based on the operation risk, and executing the power reserve decision. The method has good expandability and engineering feasibility, and can be directly integrated to an existing scheduling platform.

Description

Technical Field

[0001] This invention relates to the field of power system operation technology, and more specifically, to a risk-driven decision-making method and system for high-proportion renewable energy power system power reserve. Background Technology

[0002] Accelerating the grid connection of renewable energy sources is a key path for the low-carbon transformation of the power system. According to relevant research, wind and solar power capacity in China is expected to exceed 60% by 2050. However, the integration of large-scale variable renewable energy sources (VREs), which are characterized by variability and intermittency, poses risks to power system operation, such as the Spanish blackout in 2025. By 2025, Spain's VRE share reached 67%, but its system regulation capacity was severely insufficient. A sudden drop in solar power generation caused grid power imbalance, triggering frequency oscillations and ultimately leading to a disconnection from the European grid. To address the risks posed by the randomness of VREs, sufficient flexible reserves are needed to ensure real-time power balance, thereby reducing system operational risks. Therefore, this paper quantifies the operational risks caused by the uncertainty of VRE output, explores the flexible regulation capabilities of various resource types, and proposes a new energy-reserve coordinated dispatch method to address a new challenge in power system energy management.

[0003] High-penetration VRE integration leads to increased demand for spinning reserves in power systems. Traditional deterministic reserve decision-making methods rely on accurate load forecasting, making it difficult to balance the flexibility and economy of high-penetration VRE power systems. Therefore, endogenous reserve determination methods based on uncertainty optimization, especially robust optimization, have been extensively studied. Endogenous reserve determination methods embed various uncertain events into the unit portfolio (UC) model to obtain generation and reserve allocation schemes.

[0004] Traditional robust energy-reserve dispatch methods rely on uncertain set modeling. An excessively large predefined set can lead to conservative dispatch schemes, while an excessively small set may result in overly optimistic ones. In particular, if the actual output of VREs exceeds the uncertain set, it will pose operational risks to the power grid (VRE reduction or load shedding). However, in power system operation, the probability distribution information of VRE prediction errors is often difficult to obtain accurately.

[0005] Flexibility resources are crucial for power systems to cope with uncertainty and reduce operational risks. With the gradual phasing out of thermal power units, energy storage has become an important flexibility resource for power systems. However, previous studies have mainly treated energy storage as a market parameter, failing to fully recognize its dynamic role in reserve dispatch. Summary of the Invention

[0006] To address the above problems, this invention proposes a risk-driven decision-making method for high-proportion renewable energy power system power reserve, comprising:

[0007] The operational risk of the VRE admission domain is defined for the power system. After the definition, a conservative sparse neural network (CSNN) is used to learn the mapping relationship between the uncertainty set boundary and the operational risk for the power system.

[0008] Based on the mapping relationship, the pre-trained CSNN risk model is transformed into a linear model using the Big-M method, and a risk-driven adaptive uncertainty set is constructed.

[0009] Based on the linearized model and the adaptive uncertainty set, a risk-proactive robust scheduling framework is established through offline training and online mapping.

[0010] Using the aforementioned risk-proactive robust scheduling framework, operational risks of the power system are identified through online mapping. Based on these operational risks, a current power reserve decision is generated and executed.

[0011] Optionally, the expressions for the wind curtailment risk (WCR) and load shedding risk (LSR) defined in the operational risk definition are as follows:

[0012]

[0013] Among them, R w,t For system operation risks, t is the time index and w is the wind farm index; Contribute to wind power forecasting, As the upper bound of the acceptable domain, As the lower bound of the acceptable domain, e represents the wind power prediction error. w,t This is to predict the probability distribution of the error.

[0014] Optionally, the expression for the mapping relationship is as follows:

[0015]

[0016] in, To mitigate the risk of wind power curtailment in renewable energy sources, For an uncertain upper boundary set, To mitigate the risk of load shedding, This represents the lower boundary of an uncertain set.

[0017] Optionally, the expression for the linear optimization model is as follows:

[0018]

[0019] in, This represents the output of neuron i in the l-th layer of the neural network. The weights between neurons i and j M is the bias of neuron i in layer l. U and M L These are the positive and negative large values ​​used to relax the ReLU function.

[0020] Optional, the expression for an indeterminate set is as follows:

[0021]

[0022] in, For wind power with uncertain output, Contribute to wind power forecasting, and These are binary auxiliary variables, For budgets with uncertain timeframes, Budget for spatial uncertainty.

[0023] Optional, risk-proactive robust scheduling frameworks include:

[0024] Energy storage backup model for variable operating conditions, risk-driven energy backup collaborative scheduling model, and ESS backup model for variable operating conditions.

[0025] Optionally, the objective function of the risk-driven energy reserve coordinated scheduling model is as follows:

[0026]

[0027] in, and The operating and standby costs of conventional generating units are separate. κ represents the backup cost of the ESS, and κ is the risk coefficient. and VCR and LSR, respectively. i.k P represents the incremental cost of the consumption characteristics of unit i in the k-th stage. i,k,t For the segmented output of conventional units, and These are the start-up and shutdown costs for conventional units. These represent adjustments to the standby cost coefficient for conventional generating units, one upward and one downward. and These represent the upward / downward adjustment of the backup cost coefficient for energy storage.

[0028] Optional, risk-driven energy reserve coordinated scheduling model, including: two-stage constraints:

[0029] The pre-scheduling constraints in the first phase and the rescheduling constraints in the second phase.

[0030] Optionally, using the aforementioned risk-prone robust scheduling framework, operational risks of the power system are identified through online mapping, and based on these operational risks, a current power reserve decision is generated, including:

[0031] Using the aforementioned risk-prone robust scheduling framework, the inner-layer problem is transformed into a single-layer problem by performing a dual transformation, resulting in the worst-case scenario for VRE processing to generate C&CG cut constraints. Based on the C&CG cut constraints, the operational risks of the power system are identified through online mapping, and based on the operational risks, the current power reserve decision is generated.

[0032] Furthermore, this invention also proposes a risk-driven high-proportion renewable energy power system energy reserve decision-making system, comprising:

[0033] The learning unit is used to define the operational risk of the VRE admission domain for the power system. After the definition, the conservative sparse neural network CSNN is used to learn the mapping relationship between the uncertainty set boundary and the operational risk for the power system.

[0034] The modeling unit is used to transform the pre-trained CSNN risk model into a linear model using the Big-M method based on the mapping relationship, and to construct a risk-driven adaptive uncertainty set.

[0035] The framework unit is used to establish a risk-proactive robust scheduling framework based on the linearized model and the adaptive uncertainty set through offline training and online mapping.

[0036] The output unit is used to identify the operational risks of the power system through online mapping using the aforementioned risk-forward robust scheduling framework, and based on the operational risks, generate the current power reserve decision and execute the power reserve decision.

[0037] Optionally, the expressions for the wind curtailment risk (WCR) and load shedding risk (LSR) defined in the operational risk definition are as follows:

[0038]

[0039] Among them, R w,t For system operation risks, t is the time index and w is the wind farm index; Contribute to wind power forecasting, As the upper bound of the acceptable domain, As the lower bound of the acceptable domain, e represents the wind power prediction error. w,t This is to predict the probability distribution of the error.

[0040] Optionally, the expression for the mapping relationship is as follows:

[0041]

[0042] in, To mitigate the risk of wind power curtailment in renewable energy sources, For an uncertain upper boundary set, To mitigate the risk of load shedding, This represents the lower boundary of an uncertain set.

[0043] Optionally, the expression for the linear optimization model is as follows:

[0044]

[0045] in, This represents the output of neuron i in the l-th layer of the neural network. The weights between neurons i and j M is the bias of neuron i in layer l. U and M L These are the positive and negative large values ​​used to relax the ReLU function.

[0046] Optional, the expression for an indeterminate set is as follows:

[0047]

[0048] in, For wind power with uncertain output, Contribute to wind power forecasting, and These are binary auxiliary variables, For budgets with uncertain timeframes, Budget for spatial uncertainty.

[0049] Optional, risk-proactive robust scheduling frameworks include:

[0050] Energy storage backup model for variable operating conditions, risk-driven energy backup collaborative scheduling model, and ESS backup model for variable operating conditions.

[0051] Optionally, the objective function of the risk-driven energy reserve coordinated scheduling model is as follows:

[0052]

[0053] in, and The operating and standby costs of conventional generating units are separate. κ represents the backup cost of the ESS, and κ is the risk coefficient. and VCR and LSR, respectively. i.k P represents the incremental cost of the consumption characteristics of unit i in the k-th stage. i,k,tFor the segmented output of conventional units, and These are the start-up and shutdown costs for conventional units. These represent adjustments to the standby cost coefficient for conventional generating units, one upward and one downward. and These represent the upward / downward adjustment of the backup cost coefficient for energy storage.

[0054] Optional, risk-driven energy reserve coordinated scheduling model, including: two-stage constraints:

[0055] The pre-scheduling constraints in the first phase and the rescheduling constraints in the second phase.

[0056] Optionally, using the aforementioned risk-prone robust scheduling framework, operational risks of the power system are identified through online mapping, and based on these operational risks, a current power reserve decision is generated, including:

[0057] Using the aforementioned risk-prone robust scheduling framework, the inner-layer problem is transformed into a single-layer problem by performing a dual transformation, resulting in the worst-case scenario for VRE processing to generate C&CG cut constraints. Based on the C&CG cut constraints, the operational risks of the power system are identified through online mapping, and based on the operational risks, the current power reserve decision is generated.

[0058] In another aspect, the present invention also provides a computing device, comprising: one or more processors;

[0059] A processor is used to execute one or more programs;

[0060] When the one or more programs are executed by the one or more processors, the method described above is implemented.

[0061] In another aspect, the present invention also provides a computer-readable storage medium having a computer program stored thereon, which, when executed, implements the method described above.

[0062] Compared with the prior art, the beneficial effects of the present invention are as follows:

[0063] This invention provides a risk-driven method for high-proportion renewable energy power system reserve decision-making, comprising: defining the operational risk of the VRE (Vehicle Renewable Energy) acceptance domain for the power system; using a conservative sparse neural network (CSNN) to learn the mapping relationship between the uncertainty set boundary and the operational risk; based on the mapping relationship, using the Big-M method to transform the pre-trained CSNN risk model into a linearized model and constructing a risk-driven adaptive uncertainty set; based on the linearized model and the adaptive uncertainty set, establishing a risk-proactive robust scheduling framework through offline training and online mapping; using the risk-proactive robust scheduling framework, identifying the operational risk of the power system through online mapping, generating a current reserve decision based on the operational risk, and executing the reserve decision. This invention has good scalability and engineering feasibility and can be directly integrated into existing scheduling platforms. Attached Figure Description

[0064] Figure 1 This is a flowchart of the method of the present invention;

[0065] Figure 2 This is a schematic diagram of a data-driven risk modeling method according to an embodiment of the present invention;

[0066] Figure 3 This is a risk-driven energy-reserve collaborative scheduling framework diagram for an embodiment of the method of the present invention;

[0067] Figure 4 A linear backup model diagram of ESS switching between two working modes in an embodiment of the method of the present invention;

[0068] Figure 5 This is a comparison chart of system operation risks for models 1-3 in the embodiments of the method of the present invention;

[0069] Figure 6 Diagrams showing the backup capacity and system operation risk of ESS under different discharge durations in embodiments of the present invention. Detailed Implementation

[0070] Exemplary embodiments of the invention will now be described with reference to the accompanying drawings. However, the invention may be embodied in many different forms and is not limited to the embodiments described herein. These embodiments are provided to fully and completely disclose the invention and to fully convey its scope to those skilled in the art. The terminology used in the exemplary embodiments illustrated in the drawings is not intended to limit the invention. In the drawings, the same units / elements are referred to by the same reference numerals.

[0071] Unless otherwise stated, the terms used herein (including technical terms) have their common meaning as understood by one of ordinary skill in the art. Furthermore, it is understood that terms defined in commonly used dictionaries should be understood to have a meaning consistent with the context of their relevant field, and not to be interpreted as having an idealized or overly formal meaning.

[0072] Example 1:

[0073] This invention proposes a risk-driven decision-making method for high-proportion renewable energy power systems, such as... Figure 1 As shown, it includes:

[0074] Step 1: Define the operational risk of the VRE admission domain for the power system. After definition, use a conservative sparse neural network (CSNN) to learn the mapping relationship between the uncertainty set boundary and the operational risk for the power system.

[0075] Step 2: Based on the mapping relationship, the pre-trained CSNN risk model is transformed into a linearized model using the Big-M method, and a risk-driven adaptive uncertainty set is constructed.

[0076] Step 3: Based on the linearized model and the adaptive uncertainty set, establish a risk-proactive robust scheduling framework through offline training and online mapping.

[0077] Step 4: Using the aforementioned risk-forward robust scheduling framework, identify the operational risks of the power system through online mapping, generate the current power reserve decision based on the operational risks, and execute the power reserve decision.

[0078] The following provides further explanation of steps 1-4 above:

[0079] Specifically, it includes:

[0080] First, a conservative sparse neural network (CSNN) is used to learn the mapping relationship between the uncertainty set boundary and operational risk. Then, the trained CSNN risk model is transformed into a linearized model based on the Big-M method. Next, a risk-driven adaptive uncertainty set is constructed to achieve synergistic optimization of operational risk and the uncertainty set. Based on this, a risk-prospective robust scheduling framework combining offline training and online mapping is developed. Operational risks are identified through rapid online mapping, thereby guiding energy and reserve decisions. Furthermore, a reserve scheduling model for an energy storage system (ESS) considering variable operating conditions is proposed to fully explore the adjustment capability of rapid start-up and shutdown resources. For the 0-1 variables introduced by the variable operating conditions of the ESS, an improved C&CG algorithm is proposed to solve the robust optimization problem of mixed-integer linear programming in the second stage. Numerical examples show that the proposed data-driven method can effectively assess system operational risk and achieve a balance between reserve capacity and operational risk through adaptive adjustment of the uncertainty set. In addition, the proposed linearized reserve model explores the reserve adjustment potential of the ESS and improves the VRE acceptability.

[0081] A data-driven risk modeling method based on deep learning is proposed, including:

[0082] S11. Definition of operational risks based on VRE admission domain:

[0083] The concept of CvaR can be used to calculate operational risk. When the probability distribution of wind power forecast error is known, the operational risk can be obtained by integrating the shaded area. The specific expressions for wind curtailment risk (WCR) and load shedding risk (LSR) are shown in formula (1).

[0084]

[0085] In the formula: R w,t For system operation risks, t is the time index and w is the wind farm index; Contribute to wind power forecasting, As the upper bound of the acceptable domain, As the lower bound of the acceptable domain, e represents the wind power prediction error. w,t This is to predict the probability distribution of the error.

[0086] It is worth noting that the risk model in Equation (1) assumes that the prediction error follows a certain distribution function. Generally, it is assumed that the prediction error follows a normal distribution. However, the probability information of the VRE prediction error distribution is difficult to obtain. This embodiment analyzes the prediction error distribution under different output ranges based on 60,000 sets of wind power prediction-actual data provided by Belgian transmission operator Elia. It can be found that: 1) the prediction error of wind power output does not completely follow a normal distribution, and the positive and negative deviations are not symmetrically distributed; 2) the prediction error distribution is different under different wind power output ranges, and the prediction error tends to decrease as the wind farm output increases. Therefore, the risk model (1) driven by the model based on the prediction error distribution has certain errors, which may lead to a larger risk in the scheduling result. In addition, the risk model in Equation (1) is a strongly nonlinear model, which needs to be linearized twice to transform it into a piecewise linearized model, which further reduces the accuracy of the model. Therefore, it is urgent to study a risk modeling method that combines historical data to reasonably describe the relationship between the wind power acceptance domain and the operational risk.

[0087] S12. Data-driven risk model based on deep learning:

[0088] Based on the above analysis, operational risk is directly related to the acceptable domain boundary. Therefore, this section proposes a data-driven risk assessment method using historical wind power data, constructing an operational risk model based on deep learning. Specific steps are as follows: Figure 2 As shown.

[0089] S121. Data Binning Processing: Due to dynamic changes caused by weather, errors under low and high power generation conditions often follow different distribution patterns. Therefore, the data is first sorted in ascending order of the predicted values ​​to divide equal-width power bins for effective analysis of local error characteristics. Secondly, the historical predicted-actual data are assigned to the corresponding bins, and the prediction error sample set is calculated according to equation (2). The prediction error samples of different power output bins are then divided into K error bins ξ. k .

[0090]

[0091] Where, N k For the interval ξ k The number of samples within. n,k Indicates the prediction error. and These are the actual output and the predicted output, respectively.

[0092] S122, Operational Risk Quantification: We propose a data-based risk quantification method, which transforms continuous integrals into discrete summations, i.e., transforming equation (1) into equation (4), to improve the ability to capture the true error distribution. Specifically, a) Calculate each error interval [e] based on equation (3). k-1 ,e k The empirical probability p k b) Calculate the risk value corresponding to each error interval k; c) Accumulate the risk values ​​of each interval outside the acceptable domain boundary to obtain the system operation risk.

[0093]

[0094] In the formula, N total Let be the total number of samples, and E be the conditional expectation of the risk corresponding to interval k.

[0095] S123. Risk Modeling Based on Deep Learning: First, an M dataset of acceptable domain boundaries is generated using the Latin hypercube sampling method (Equation (5)). Then, a training dataset of acceptable domain boundaries and operational risk values ​​is constructed using Equation (4). Based on this, a Conservative Sparse Neural Network (CSNN) is used to construct the mapping relationship between acceptable domain boundaries and operational risks, as shown in Equation (6).

[0096]

[0097] The trained CSNN risk model (6) is a general form of an L-layer feedforward neural network with ReLU as the activation function (as shown in equation (7)). The activation function ReLU(z) = max{0,z} is nonlinear. For each layer of ReLU in equation (7), ReLU(x) = max{0,x}, it can be linearized using the Big-M method in equation (8). The linearized CSNN risk model can be embedded into a power system optimization model.

[0098]

[0099]

[0100] In the formula: It is the output of neuron i in the l-th layer of the neural network. These are the weights between neurons i and j. It is the bias of neuron i in layer l, M U and M L These are the positive and negative large values ​​used to relax the ReLU function.

[0101] S13, Risk-driven adaptive uncertainty set:

[0102] The ARV introduced above corresponds to the uncertain set in robust optimization. When the VRE output exceeds the range of the uncertain set, it will cause grid operation risks. However, an overly conservative uncertain set will also reduce operational economy. To this end, this embodiment constructs a risk-driven adaptive uncertain set (AUS), as shown in Equation (8). Compared with the traditional uncertain set, the boundary of AUS is... and It is a decision variable. By adjusting the boundaries of AUS, a balance can be achieved between operational risk and operational cost.

[0103]

[0104] In the formula, For wind power with uncertain output, It contributes to wind power forecasting. and These are binary auxiliary variables, For budgets with uncertain timeframes, Budget for spatial uncertainty.

[0105] A backup scheduling framework for risk prediction in offline learning and online mapping was constructed, as follows: Figure 3 As shown;

[0106] S21. Backup modeling for resources that can be quickly started and stopped:

[0107] In traditional power systems, spinning reserve is mainly provided by conventional power sources such as thermal power. Unlike conventional units that only provide backup service while online, fast start-stop power sources (such as energy storage and fast start-stop units) can provide backup service by changing operating conditions. Taking ESS as an example, this embodiment constructs an energy storage backup model operating under varying conditions.

[0108] The operational constraints of ESS are shown in equations (11)-(15).

[0109]

[0110]

[0111] in, and These represent the discharge and charge states of an energy storage system (ESS), respectively. and These represent the discharge and charging power of the energy storage system, respectively; e b,t Indicates the energy state of the energy storage system; Indicates the initial energy state of the energy storage system; η ch and η dis These represent the energy conversion efficiency during the charging and discharging processes, respectively.

[0112] Two sets of auxiliary variables are introduced to describe the switching between charging and discharging modes of the ESS, namely: and If the ESS switches from charging mode to discharging mode Otherwise, it is 0; if the ESS switches from discharge mode to charge mode, then The logical relationship between auxiliary variables and state variables is shown in equation (16).

[0113]

[0114] Based on the constructed auxiliary variables, a linear standby model for ESS switching between two operating modes is established, such as... Figure 4 As shown. In discharge mode, the ESS can provide upward reserve by increasing the power generation, or downward reserve by decreasing the power generation or switching to charging mode. Equations (17)-(19) are the reserve constraints of the ESS in discharge mode. In charging mode, the ESS can provide downward reserve by increasing the charging power, or upward reserve by decreasing the charging power or switching to power generation mode. Equations (20)-(21) are the reserve constraints of the ESS in charging mode.

[0115]

[0116] In the formula, and These refer to the upward and downward adjustment of reserve capacity under power generation mode; Reserved for use under changing operating conditions; and These are the up and down adjustments for charging mode, respectively. For use in adjusting to different operating conditions.

[0117] S22, Risk-Driven Energy-Spare Coordination Scheduling Model:

[0118] This section establishes a risk-driven robust energy-reserve dispatch model, which comprises two phases: a day-ahead phase for formulating generation and reserve plans; and an intraday phase for responding to uncertainties based on the reserve capacity determined in the day-ahead phase. This model accurately characterizes the operation of slow-response conventional units and fast-response energy storage in the coordinated energy and reserve market.

[0119] S221, Objective Function:

[0120] The objective function (18) aims to minimize operating costs and operating risks. Operating costs include the generation, start-up, shutdown, and standby costs of conventional units (19), and the standby costs of ESS (20). Operating risks include VRE reduction and load shedding.

[0121]

[0122] In the formula: and Operating and standby costs for conventional generating units; κ represents the backup cost of the ESS; κ represents the risk coefficient. and These are VCR and LSR, respectively; c i.k P represents the incremental cost of the consumption characteristics of unit i in the k-th segment; i,k,t For the segmented output of conventional units; and These are the start-up and shutdown costs for conventional generating units; These represent adjustments to the standby cost coefficient for conventional generating units, respectively. These represent the upward / downward adjustment of the backup cost coefficient for energy storage.

[0123] S222, Phase 1 Pre-scheduling Constraints:

[0124] The constraints of the pre-scheduling model include: system node power balance constraints (26), line power flow constraints (27)-(28), and operational risk constraints (29) with f. csnn Equations (7)-(8) can be converted into MILP constraints. Conventional unit operation constraints include: power generation constraints and reserve constraints (30)-(31), as well as minimum start-up and shutdown time constraints, start-up and shutdown cost constraints, and ramp-up constraints. Equation (32) is the segmented output model of the unit. The operation and reserve constraints of ESS are as shown in Equations (11)-(12). Equation (33) is the wind farm output. Equation (34) is the boundary constraint of the wind power uncertainty set.

[0125]

[0126] S223, Second-phase rescheduling constraints:

[0127] In the second stage, power fluctuations in any wind power output scenario within the uncertain set are balanced by utilizing the reserve capacity of various resources. This ensures the feasibility of reserve decisions and the acceptability of VREs. The uncertain set of VRE output is shown in equations (9)-(10). The constraints in the second stage include: equation (35) is the node power balance constraint. Line power flow constraints are not given in detail. Equation (36) is the unit reserve scheduling constraint; equations (37)-(40) are the ESS flexible adjustment constraints under uncertain scenarios. In addition, energy limits for ESS under uncertain scenarios are also included, as shown in equations (13)-(15), to ensure the deliverability of reserves.

[0128]

[0129] In the formula, P i,t P w,t P l,t P d,t To predict the output of conventional generating units, VRE output, line transmission power, and node load under the given scenario; i,t Indicates the unit's start-up and shutdown status, P i min and P i max These are the lower and upper limits of the unit's output, respectively. and These are the restrictions on increasing and decreasing the reserve of the generating unit, respectively. θ m Let θ be the voltage phase angle at node m. ref The phase angle of the reference node, x l P is the reactance of line l; l max This represents the upper limit of line transmission power. Vc This indicates wind curtailment penalty dilution, c Ls This is the load shedding penalty coefficient. These are, respectively, the output of conventional generating units, the output of VRE, the power transmission power of the line, and the discharge / charging power of ESS under uncertain scenarios; and These are the ESS operating modes under uncertain scenarios.

[0130] S3. An ESS backup model considering flexible operation under varying conditions is proposed:

[0131] S31, Matrix form of the robust UC model:

[0132] To facilitate the solution, the model proposed by S22 is rewritten in the following matrix form.

[0133]

[0134] In the formula: u and p are the 0-1 variable and continuous variable vectors of the first stage, respectively; Q represents the operational risk variable; pb is the boundary of the uncertainty set; v and p u These are the 0-1 variables and continuous variable vectors for the second stage, respectively. Represents an uncertain variable vector; a, b, g, l, A, B, H, F, G, L are the corresponding symbolic constant coefficient matrices.

[0135] S32. Improved column and constraint generation algorithm solution process:

[0136] Based on the C&CG decomposition idea, the original problem (41) is decomposed into a first-stage main problem (42) and a second-stage sub-problem (43). The main problem includes equations (11)-(34) and the C&CG cutting plane returned by solving the sub-problem. The main problem is a MILP problem, which can be solved using a commercial solver.

[0137]

[0138] After solving the main problem, the solution p iter (Such as planned output and backup plan) are passed into the second-stage subproblem. The subproblem can be expressed as a max-min problem as shown in equation (43), where s is a vector of slack variables introduced to ensure that the subproblem has a feasible solution under uncertain scenarios. The proposed subproblem solution method is detailed in S33.

[0139]

[0140] By solving the rescheduling problem (43), the objective function value Φ and the worst-case scenario are obtained. If Φ > 0, increment the iteration count iter by 1 (i.e., iter = iter + 1), based on Create new variables and generate the C&CG cutting plane (44), then return to the main problem. If Φ = 0, output the optimal solution and end the iteration process.

[0141]

[0142] Solving the S33, MILP-form rescheduling subproblem:

[0143] For the second-stage subproblem in equation (43), the introduction of 0-1 variables in the ESS variable operating condition operation prevents the dual transformation of the inner-level problem of the max-min model. To solve this problem, this embodiment adopts a method of introducing auxiliary variables. The specific improvement method is as follows: First, introduce 0-1 auxiliary variables into equation (43), define the 0-1 variable v in the original problem as a continuous variable, and constrain v to be equal to Equation (43) is transformed into the model shown in Equation (45). In Equation (45), the outer layer is a max problem under an uncertain scenario, the middle layer is a min problem with auxiliary variables as the variables to be optimized, and the inner layer is transformed into a linear min problem.

[0144]

[0145] Where η and λ are the dual variable vectors of constraint (52).

[0146] After introducing auxiliary variables, the inner minimization problem in equation (43) is transformed into a linear programming problem. Therefore, the inner problem can be dualized, as shown in equation (46).

[0147]

[0148] Based on the minimum-maximum inequality, the formula of equation (46) can be transformed as follows:

[0149]

[0150] In equation (47), the variable The optimal value is determined by enumerating all possible combinations of 0 and 1, which can be represented as follows:

[0151]

[0152] Where Ω={σ≤λ,σ≤0}. Substituting equation (48) into (47) can eliminate the inner min problem. Then, by merging the outer max problem with the inner max problem, the objective function of equation (46) can be transformed into (49).

[0153]

[0154] Through the above processing, equation (46) is transformed into a single-layer problem. Furthermore, η exists in equation (49). T and The nonlinear terms involved in the multiplication can be handled using the Big M method. By solving the dual problem of the subproblem, the worst-case scenario for VRE processing is obtained, thereby generating C&CG cut constraints.

[0155] The invention will now be illustrated using the IEEE RTS-24 system and the IEEE 118-node system as examples:

[0156] To verify the effectiveness of the proposed method, it is compared with existing studies, as shown in Table I. Model 1 is a RERS model based on a predefined uncertainty set, with the uncertainty set boundary consistent with the prediction error at an 85% confidence level. Model 2 is a RERS model considering operational risk, using the risk model of Equation (1). It is important to note that both Model 1 and Model 2 assume that the VRE prediction error follows a normal distribution, and its parameters are obtained by fitting historical data. Models 3 and 4 use data-driven risk models. In Model 3, the ESS provides backup only under fixed operating conditions; Model 4 uses the proposed ESS variable operating condition linear backup model.

[0157] Table 1

[0158]

[0159] The improved IEEE RTS-24 system comprises 10 conventional turbine units, one 1000MW wind farm, and two 50MW / 400MWh energy storage stations. The standby capacity price for energy storage is ¥10 / MW. The total dispatch time is 24 hours, with a 1-hour time interval. Curtailment penalties and load shedding penalties are set at ¥100 / MWh and ¥1000 / MWh, respectively. VRE uncertainty budget Γ T Set to 10.

[0160] Data-driven risk model analysis based on DNN:

[0161] This section analyzes the learning process and training results of the data-driven risk model, taking VRE risk reduction as an example. In the learning phase, 10,000 samples are generated to construct the model training dataset. Subsequently, based on CSNN, the mapping relationship between AUS and operational risk is learned by normalizing the maximum value of the feature vector to the [0,1] interval. As the number of training rounds increases, the mean squared error (MSE) value continuously decreases to 10⁻⁴, indicating effective training. Based on this, the weights and biases of the trained CSNN are extracted, and the activation function is processed using the Big-M method, transforming it into a MILP model.

[0162] Analysis of energy-reserve scheduling results from different models:

[0163] Table 2 compares the operating costs and risks of the four models. The uncertainty set boundary of Model 1 is predetermined. As shown in Table 2, Model 1 has a higher operating risk, indicating that traditional uncertainty set modeling methods cannot effectively mitigate system operating risks. Compared to Model 1, Model 2 reduces operating risk through coordinated optimization of operational flexibility and wind power acceptance domain. However, Model 2's operating and reserve costs are significantly higher, possibly due to an overly conservative uncertainty set. Models 3 and 4 effectively measure the actual system adjustment needs through data-driven risk models. While mitigating operating risks, Model 4 reduces operating and reserve costs by 9.5% and 24.5%, respectively, compared to Model 2. Compared to Model 3, Model 4 considers that the ESS (Electronic Energy Storage System) provides reserve capacity by flexibly adjusting its operating mode, improving the ESS's reserve capability and reducing the system's total reserve cost by 9.5%. Furthermore, the system operating risk is also reduced.

[0164] Table 2

[0165]

[0166] Comparative analysis of wind power uncertainty set and operational risks:

[0167] This section further analyzes the effectiveness of the proposed method in conjunction with the operational risk analysis of Models 1, 2, and 4. Figure 5 The system operation risks for different models at various time periods are presented. A wind power prediction error distribution band is constructed based on historical data, and the uncertainty set boundaries for different models are given. Specifically, the uncertainty set for Model 1 is determined using the 85% confidence level of the prediction error distribution. It can be seen that during the time period from period 5 to period 20, the uncertainty set boundaries all fall within the prediction error distribution band, leading to significant risks of wind curtailment and load shedding, such as... Figure 5 As shown. Furthermore, during periods of high wind power output, the uncertainty boundary extends beyond the prediction error band, leading to increased operating costs.

[0168] Models 2 and 4 achieve adaptive optimization of the uncertain set based on the risk model. However, Model 2 uses a continuous integration method to build the risk model, assuming that the prediction error follows a normal distribution. Figure 5 As can be seen, Model 2 has a load shedding risk of 0. However, the conservatism of the lower boundary of the uncertainty set leads to an increase in reserve requirement, increasing operating and reserve costs. Model 4 uses a data-driven risk model, effectively assessing the operational risks caused by wind power forecasting errors. Therefore, the lower boundary of the uncertainty set in Model 4 basically falls on the boundary of the forecasting error band, reasonably reducing the load shedding risk. At the same time, during periods of high wind power output, a certain degree of wind curtailment reduces reserve requirement, ensuring the economic efficiency of system operation.

[0169] Analysis of the backup regulation capacity of energy storage under rapid changes in operating conditions:

[0170] ESS (Emergency Serviceable Start-up) resources can provide backup services by operating under varying conditions. Figure 6 The standby scheduling plans for conventional units and ESS were compared in Models 3 and 4. From... Figure 6 As can be seen, compared to constant operating conditions, ESS provides 22.8% more reserve capacity through variable operating conditions, especially by reducing reserve. This effectively solves the problem of increased wind curtailment risk caused by insufficient peak-shaving capacity of traditional units.

[0171] Furthermore, the backup capacity provided by the ESS is limited by its maximum capacity. As the discharge duration increases, the backup capacity provided by the ESS increases. This indicates that the constructed two-stage robust optimization model can effectively guarantee the deliverability of ESS backup in actual scheduling. Moreover, as the energy storage discharge duration increases, the system operation risk is further reduced. Therefore, by rationally configuring the energy storage capacity, the system's flexible adjustment capability can be effectively guaranteed, and the system operation risk can be reduced.

[0172] Analysis of the advantages of the improved C&CG algorithm:

[0173] This section applies the IC&CG algorithm to the IEEE-RTS 24-bus system and compares it with the nested C&CG algorithm. Table 3 shows the number of iterations and solution time for the two algorithms. For the nested C&CG algorithm, the inner C&CG problem becomes increasingly difficult to converge as the number of iterations of the outer C&CG increases. When the number of iterations of the inner C&CG algorithm reaches 6 and the solution result of the lower-level model no longer changes, the iteration is stopped and the outer C&CG algorithm is entered. As can be seen from Table 4, the number of outer iterations is the same for both algorithms, but the inner loop C&CG problem of the nested C&CG algorithm contains multiple iterations. In the IEEE-118-bus system, the IC&CG algorithm reduces the solution time by 37.5%. It can be observed that the performance of the IC&CG algorithm decreases as the power system size increases. This is because the solution time of the main problem increases with the increase in the number of returned C&CG cut planes. Therefore, further exploration of methods to improve the solution efficiency of the main problem is needed in the next step.

[0174] Table 3

[0175]

[0176] Based on the above analysis, we can conclude that the improved C&CG algorithm can effectively solve the two-stage robust model where the subproblem is a MILP problem, and can effectively reduce the complexity and time of problem solving.

[0177] To address the operational risks associated with a high proportion of VRE access, traditional robust scheduling models are insufficient. This embodiment establishes a risk-driven robust energy and reserve scheduling model and explores the operational flexibility of the ESS through a variable operating condition reserve model. Case studies demonstrate that:

[0178] (1) The proposed risk-driven robust energy-reserve scheduling model solves the problem of how to measure potential losses when the realized value of VRE uncertainty exceeds the specified uncertainty set. By optimizing the boundary of the uncertainty set, it achieves a balance between flexibility adjustment reserve and operation risk, effectively reducing the operation risk of high-penetration VRE power systems.

[0179] (2) The proposed data-driven risk modeling method solves the problem that existing risk models cannot account for VRE prediction errors under different output levels, and accurately assesses the wind curtailment and load shedding risks of the power system. Compared with model-driven risk models, the established model reduces the conservatism of the uncertainty set, and reduces operating and reserve costs by 9.5% and 24.5%, respectively.

[0180] (3) A flexible standby model for ESS operation under varying operating conditions was established, which effectively improved the standby capability of ESS, reduced the generation and standby costs of conventional units, and improved the VRE acceptability of the system. In addition, based on a two-stage robust optimization model, the capacity limitation on ESS standby was considered, ensuring the deliverability of ESS standby services.

[0181] (4) The proposed IC&CG algorithm can effectively solve the two-stage robust optimization model where the subproblem is a MILP problem, and incorporates the flexible adjustment capability of ESS under varying operating conditions into the RRER model. Compared with the nested C&CG algorithm, the IC&CG algorithm omits the inner C&CG iteration process. In the standard example, the running time is improved by 6 times.

[0182] Example 2:

[0183] This invention also proposes a risk-driven high-proportion renewable energy power system energy reserve decision-making system 200, comprising:

[0184] Learning unit 201 is used to define the operational risk of the VRE acceptance domain for the power system. After definition, a conservative sparse neural network (CSNN) is used to learn the mapping relationship between the boundary of the uncertainty set and the operational risk for the power system.

[0185] Modeling unit 202 is used to transform a pre-trained CSNN risk model into a linear model using the Big-M method based on the mapping relationship, and to construct a risk-driven adaptive uncertainty set.

[0186] Framework unit 203 is used to establish a risk-forward robust scheduling framework based on the linearized model and the adaptive uncertainty set through offline training and online mapping.

[0187] Output unit 204 is used to identify the operational risks of the power system through online mapping using the aforementioned risk-forward robust scheduling framework, generate a current power reserve decision based on the operational risks, and execute the power reserve decision.

[0188] The expressions for wind curtailment risk (WCR) and load shedding risk (LSR) defined in the operational risk definition are as follows:

[0189]

[0190] Among them, R w,t For system operation risks, t is the time index and w is the wind farm index; Contribute to wind power forecasting, As the upper bound of the acceptable domain, As the lower bound of the acceptable domain, e represents the wind power prediction error.w,t This is to predict the probability distribution of the error.

[0191] The expression for the mapping relationship is as follows:

[0192]

[0193] in, To mitigate the risk of wind power curtailment in renewable energy sources, For an uncertain upper boundary set, To mitigate the risk of load shedding, This represents the lower boundary of an uncertain set.

[0194] The expression for the linear optimization model is as follows:

[0195]

[0196] in, This represents the output of neuron i in the l-th layer of the neural network. The weights between neurons i and j M is the bias of neuron i in layer l. U and M L These are the positive and negative large values ​​used to relax the ReLU function.

[0197] The expression for the uncertain set is as follows:

[0198]

[0199] in, For wind power with uncertain output, Contribute to wind power forecasting, and These are binary auxiliary variables, For budgets with uncertain timeframes, Budget for spatial uncertainty.

[0200] The risk-proactive robust scheduling framework includes:

[0201] Energy storage backup model for variable operating conditions, risk-driven energy backup collaborative scheduling model, and ESS backup model for variable operating conditions.

[0202] The objective function of the risk-driven energy reserve coordinated scheduling model is as follows:

[0203]

[0204] in, and The operating and standby costs of conventional generating units are separate. κ represents the backup cost of the ESS, and κ is the risk coefficient. and VCR and LSR, respectively. i.k P represents the incremental cost of the consumption characteristics of unit i in the k-th stage. i,k,t For the segmented output of conventional units, and These are the start-up and shutdown costs for conventional units. These represent adjustments to the standby cost coefficient for conventional generating units, one upward and one downward. and These represent the upward / downward adjustment of the backup cost coefficient for energy storage.

[0205] The risk-driven energy reserve coordinated scheduling model includes constraints in two phases:

[0206] The pre-scheduling constraints in the first phase and the rescheduling constraints in the second phase.

[0207] Optionally, using the aforementioned risk-prone robust scheduling framework, operational risks of the power system are identified through online mapping, and based on these operational risks, a current power reserve decision is generated, including:

[0208] Using the aforementioned risk-prone robust scheduling framework, the inner-layer problem is transformed into a single-layer problem by performing a dual transformation, resulting in the worst-case scenario for VRE processing to generate C&CG cut constraints. Based on the C&CG cut constraints, the operational risks of the power system are identified through online mapping, and based on the operational risks, the current power reserve decision is generated.

[0209] The core of this invention is to characterize the mapping relationship between the boundary of the uncertain set and operational risk using a conservative sparse neural network, and to couple Big-M linearization with a scheduling model to form an uncertain set that adapts to operating conditions. This data-driven framework can quickly and accurately assess operational risks and achieve a synergistic trade-off between reserve configuration and risk level during the optimization process, avoiding over-conservatism. Furthermore, an executable linearized reserve model is constructed to fully leverage the rapid start-up and shutdown and bidirectional adjustment capabilities of energy storage (ESS), improving reserve availability and thus significantly enhancing the grid integration and consumption level of renewable energy (VRE) and system resilience. This method and system possess good scalability and engineering feasibility, and can be directly integrated into existing scheduling platforms.

[0210] Example 3:

[0211] Based on the same inventive concept, this invention also provides a computer device, which includes a processor and a memory. The memory stores a computer program, which includes program instructions. The processor executes the program instructions stored in the computer storage medium. The processor may be a Central Processing Unit (CPU), or other general-purpose processors, digital signal processors (DSPs), application-specific integrated circuits (ASICs), field-programmable gate arrays (FPGAs), or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components, etc. It is the computing and control core of the terminal, suitable for implementing one or more instructions, specifically suitable for loading and executing one or more instructions in the computer storage medium to implement corresponding method flows or corresponding functions, thereby implementing the steps of the methods in the above embodiments.

[0212] Example 4:

[0213] Based on the same inventive concept, this invention also provides a storage medium, specifically a computer-readable storage medium (Memory), which is a memory device in a computer device used to store programs and data. It is understood that the computer-readable storage medium here can include both the built-in storage medium in the computer device and extended storage media supported by the computer device. The computer-readable storage medium provides storage space that stores the terminal's operating system. Furthermore, this storage space also stores one or more instructions suitable for loading and execution by a processor. These instructions can be one or more computer programs (including program code). It should be noted that the computer-readable storage medium here can be high-speed RAM or non-volatile memory, such as at least one disk storage device. The processor can load and execute one or more instructions stored in the computer-readable storage medium to implement the steps of the method in the above embodiments.

[0214] Those skilled in the art will understand that embodiments of the present invention can be provided as methods, systems, or computer program products. Therefore, the present invention can take the form of a completely hardware embodiment, a completely software embodiment, or an embodiment combining software and hardware aspects. Furthermore, the present invention can take the form of a computer program product implemented on one or more computer-usable storage media (including but not limited to disk storage, CD-ROM, optical storage, etc.) containing computer-usable program code. The solutions in the embodiments of the present invention can be implemented using various computer languages, such as the object-oriented programming language Java and the interpreted scripting language JavaScript.

[0215] This invention is described with reference to flowchart illustrations and / or block diagrams of methods, apparatus (systems), and computer program products according to embodiments of the invention. It will be understood that each block of the flowchart illustrations and / or block diagrams, and combinations of blocks in the flowchart illustrations and / or block diagrams, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, special-purpose computer, embedded processor, or other programmable data processing apparatus to produce a machine, such that the instructions, which execute via the processor of the computer or other programmable data processing apparatus, generate instructions for implementing the flowchart illustrations and / or block diagrams. Figure 1 One or more processes and / or boxes Figure 1 A device that provides the functions specified in one or more boxes.

[0216] These computer program instructions may also be stored in a computer-readable storage medium that can direct a computer or other programmable data processing device to function in a particular manner, such that the instructions stored in the computer-readable storage medium produce an article of manufacture including instruction means, which are implemented in a process Figure 1 One or more processes and / or boxes Figure 1 The function specified in one or more boxes.

[0217] These computer program instructions may also be loaded onto a computer or other programmable data processing equipment to cause a series of operational steps to be performed on the computer or other programmable equipment to produce a computer-implemented process, thereby providing instructions that execute on the computer or other programmable equipment for implementing the process. Figure 1 One or more processes and / or boxes Figure 1 The steps of the function specified in one or more boxes.

[0218] Although preferred embodiments of the invention have been described, those skilled in the art, upon learning the basic inventive concept, can make other changes and modifications to these embodiments. Therefore, the appended claims are intended to be interpreted as including both the preferred embodiments and all changes and modifications falling within the scope of the invention.

[0219] Obviously, those skilled in the art can make various modifications and variations to this invention without departing from its spirit and scope. Therefore, if these modifications and variations fall within the scope of the claims of this invention and their equivalents, this invention also intends to include these modifications and variations.

Claims

1. A risk-driven high-proportion new energy power system power reserve decision method, characterized in that, The application relates to a method for generating an energy reserve decision for a power system. The method comprises the following steps: Defining an operation risk of a VRE receiving area of the power system, and then learning a mapping relationship between an uncertain set boundary and the operation risk of the power system by using a conservative sparse neural network (CSNN) for the power system; Based on the mapping relationship, converting a pre-trained CSNN risk model into a linear model by using a Big-M method, and constructing a risk-driven adaptive uncertain set; Based on the linear model and the adaptive uncertain set, establishing a risk-anticipating robust scheduling framework by means of offline training and online mapping; 2. The high-proportion new energy power system power reserve decision method according to claim 1, characterized in that, Using the risk-anticipating robust scheduling framework to identify an operation risk of the power system by means of online mapping, and generating a current energy reserve decision based on the operation risk and executing the energy reserve decision. where R w,t is the system operation risk, t is the time index, and w is the wind farm index; is the wind power forecast output, is the upper bound of the admissible region, is the lower bound of the admissible region, is the wind power forecast error, e w,t is the forecast error distribution probability.

3. The high-proportion new energy power system power reserve decision method according to claim 1, characterized in that, Expressions of wind curtailment risk (WCR) and load shedding risk (LSR) defined by the operation risk are as follows: wherein, is a new energy curtailment wind risk, is an upper bound on the uncertainty set, is a cut load risk, is a lower bound on the uncertainty set.

4. The high-proportion new energy power system power reserve decision method according to claim 1, characterized in that, An expression of the mapping relationship is as follows: where, is the output of the neuron i in the l-th layer of the neural network, is the weight between neurons i and j, is the bias of the neuron i in the l-th layer, M U and M L are the positive and negative large values for the relaxed ReLU function.

5. The high-proportion new energy power system power reserve decision method according to claim 1, characterized in that, An expression of the linear optimization model is as follows: wherein, is the wind power uncertainty output, is the wind power forecast output, and are binary auxiliary variables, is the temporal uncertainty budget, is the spatial uncertainty budget.

6. The high-proportion new energy power system power reserve decision method according to claim 1, characterized in that, An expression of the uncertain set is as follows: The risk-anticipating robust scheduling framework comprises:

7. The high-proportion new energy power system power reserve decision method according to claim 6, characterized in that, A variable working condition operation energy storage reserve model, a risk-driven energy reserve collaborative scheduling model and a variable working condition operation ESS reserve model. wherein, and respectively the operating and standby cost of the conventional units, is the standby cost of the ESS, and K is the risk coefficient, and respectively the VCR and LSR, and c i.k is the incremental cost of the kth segment of unit i, and P i,k,t is the segmented output of the conventional unit, and respectively the start-up and shut-down cost of the conventional unit, respectively the up / down reserve cost coefficient of the conventional unit, and respectively the up / down reserve cost coefficient of the energy storage.

8. The high-proportion new energy power system power reserve decision method according to claim 7, characterized in that, An objective function of the risk-driven energy reserve collaborative scheduling model is as follows: The risk-driven energy reserve collaborative scheduling model comprises: 9.The high-proportion new energy power system electricity reserve decision method according to claim 1, characterized in that, A first-stage pre-scheduling constraint and a second-stage rescheduling constraint. Using the risk-anticipating robust scheduling framework to identify an operation risk of the power system by means of online mapping, and generating a current energy reserve decision based on the operation risk, comprises:

10. A risk-driven high-proportion new energy power system power reserve decision system, characterized in that, Using the risk-anticipating robust scheduling framework to convert an inner problem into a single-layer problem for solving after dual transformation, to obtain a worst scenario of VRE processing to generate a C&CG cut constraint, identifying an operation risk of the power system by means of online mapping based on the C&CG cut constraint, and generating a current energy reserve decision based on the operation risk. The application relates to a method for generating an energy reserve decision for a power system. The method comprises the following steps: Defining an operation risk of a VRE receiving area of the power system, and then learning a mapping relationship between an uncertain set boundary and the operation risk of the power system by using a conservative sparse neural network (CSNN) for the power system; Based on the mapping relationship, converting a pre-trained CSNN risk model into a linear model by using a Big-M method, and constructing a risk-driven adaptive uncertain set; 11. The high-proportion new-energy power system electricity reserve decision system according to claim 10, characterized in that, Based on the linear model and the adaptive uncertain set, establishing a risk-anticipating robust scheduling framework by means of offline training and online mapping; where R w,t is the system operation risk, t is the time index, and w is the wind farm index; is the wind power forecast output, is the upper bound of the admissible region, is the lower bound of the admissible region, is the wind power forecast error, e w,t is the forecast error distribution probability.

12. The high-proportion new-energy power system electricity reserve decision system according to claim 10, characterized in that, Using the risk-anticipating robust scheduling framework to identify an operation risk of the power system by means of online mapping, and generating a current energy reserve decision based on the operation risk and executing the energy reserve decision. wherein, is a new energy curtailment wind risk, is an upper bound of an uncertainty set, is a cut load risk, is a lower bound of an uncertainty set.

13. The high proportion of new energy power system electric quantity reserve decision system according to claim 10, characterized in that, Expressions of wind curtailment risk (WCR) and load shedding risk (LSR) defined by the operation risk are as follows: where, is the output of the neuron i in the l-th layer of the neural network, is the weight between neurons i and j, is the bias of the neuron i in the l-th layer, M U and M L are the positive and negative large values for the relaxed ReLU function.

14. The high proportion of new energy power system electric quantity reserve decision system according to claim 10, characterized in that, An expression of the mapping relationship is as follows: An expression of the linear optimization model is as follows: An expression of the uncertain set is as follows: The risk-anticipating robust scheduling framework comprises: A variable working condition operation energy storage reserve model, a risk-driven energy reserve collaborative scheduling model and a variable working condition operation ESS reserve model. An objective function of the risk-driven energy reserve collaborative scheduling model is as follows: The risk-driven energy reserve collaborative scheduling model comprises: A first-stage pre-scheduling constraint and a second-stage rescheduling constraint. Using the risk-anticipating robust scheduling framework to identify an operation risk of the power system by means of online mapping, and generating a current energy reserve decision based on the operation risk, comprises: Using the risk-anticipating robust scheduling framework to convert an inner problem into a single-layer problem for solving after dual transformation, to obtain a worst scenario of VRE processing to generate a C&CG cut constraint, identifying an operation risk of the power system by means of online mapping based on the C&CG cut constraint, and generating a current energy reserve decision based on the operation risk. wherein, is the wind power uncertainty output, is the wind power forecast output, and are binary auxiliary variables, is the temporal uncertainty budget, is the spatial uncertainty budget.

15. The high proportion of new energy power system electric quantity reserve decision system according to claim 10, characterized in that, The risk-anticipating robust scheduling framework comprises: The energy reserve model of variable operating conditions, the risk-driven energy reserve collaborative scheduling model and the ESS reserve model of variable operating conditions.

16. The high proportion of new energy power system electric quantity backup decision system according to claim 15, characterized in that, The objective function of the risk-driven energy reserve collaborative scheduling model is as follows: wherein, and respectively the operating and standby cost of the conventional units, is the standby cost of the ESS, and K is the risk coefficient, and respectively the VCR and LSR, and c i.k is the incremental cost of the kth segment of unit i, and P i,k,t is the segmented output of the conventional unit, and respectively the start-up and shut-down cost of the conventional unit, respectively the up / down reserve cost coefficient of the conventional unit, and respectively the up / down reserve cost coefficient of the ESS.

17. The high proportion of new energy power system electric quantity backup decision system according to claim 16, characterized in that, The risk-driven energy reserve collaborative scheduling model comprises two-stage constraints: The pre-scheduling constraints of the first stage and the rescheduling constraints of the second stage.

18. The high proportion of new energy power system electric quantity backup decision system according to claim 10, characterized in that, The risk-anticipating robust scheduling framework is used to identify the operation risk of the power system through online mapping, and based on the operation risk, a current electricity reserve decision is generated, comprising: The risk-anticipating robust scheduling framework is used to identify the operation risk of the power system through online mapping, and based on the operation risk, a current electricity reserve decision is generated, comprising:

19. A computer device, comprising: Comprise: One or more processors; Processors for executing one or more programs; When the one or more programs are executed by the one or more processors, the method as claimed in any one of claims 1-9 is implemented.

20. A computer-readable storage medium, characterized in that, The computer program is stored thereon, and when the computer program is executed, the method as claimed in any one of claims 1-9 is implemented.

Images