A method for assessing the continuous-time reliability of power systems considering the time-varying trajectory uncertainties of new energy sources

By constructing a continuous-time source-load trajectory and a finite-dimensional mixed-integer optimization model, the problem of identifying short-term power supply shortage risks caused by rapid changes in renewable energy output and load in the power system is solved, realizing quantitative assessment of power system operation risks and dispatch decision support.

CN123121484APending Publication Date: 2026-11-17GUANGXI UNIV
View PDF 0 Cites 0 Cited by

Patent Information

Application Number
CN202611182851.1
Authority / Receiving Office
CN · China
Patent Type
Applications(China)
Current Assignee / Owner
Filing Date
2026-08-05
Publication Date
2026-11-17

AI Technical Summary

Technical Problem

Existing power system reliability assessment methods are insufficient to fully characterize the continuous state evolution of renewable energy output and load over short time scales, which may lead to discrepancies in system safety by discrete time-point models and failure to identify hidden risks within a given period.

Method used

By constructing continuous-time source-load trajectories, a finite-dimensional mixed-integer optimization model based on Bernstein basis functions is established to solve for power system operating characteristic parameters, optimal control coefficients, reconstruct load reduction trajectories, and quantify load loss risk.

Benefits of technology

It enables the identification of short-term power supply shortage risks caused by rapid changes in renewable energy output and load, provides a quantitative assessment of the operational risks of high-proportion renewable energy power systems and a basis for dispatching decisions, and avoids model size expansion caused by high-density discrete sampling.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN123121484A_ABST
    Figure CN123121484A_ABST
Patent Text Reader

Abstract

This invention discloses a method for continuous-time operation reliability assessment of power systems considering the uncertainty of time-varying trajectories of new energy sources, comprising the following steps: Step 1) Obtaining several complete source-load time-series scenarios with probabilities for assessment in the time domain, and converting the complete source-load time-series scenarios into continuous-time source-load trajectories; Step 2) Establishing continuous-time operation response models corresponding to each source-load time-series scenario; Step 3) Defining operation reliability indices based on continuous load reduction trajectories for each scenario; Step 4) Representing the continuous-time operation response models corresponding to each source-load time-series scenario in a sub-Bernstein basis function and converting them into finite-dimensional mixed-integer optimization models; solving the finite-dimensional mixed-integer optimization models to obtain the optimal control coefficients of the power system operation characteristic parameters; Step 5) Reconstructing the continuous load reduction trajectory based on the optimal control coefficients of the power system operation characteristic parameters, and calculating the operation reliability indices.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] This invention relates to the field of power system operation risk analysis and reliability quantification technology, specifically a method for continuous-time operation reliability assessment of power systems considering the uncertainty of time-varying trajectories of new energy sources. Background Technology

[0002] With the large-scale integration of new energy sources, the output of new energy sources and the system load exhibit stronger volatility over short time scales. Existing operational reliability assessments typically use predicted values, average values, or discrete sampled values ​​at fixed time intervals as input, and only check power balance, unit output, and network security constraints at each discrete time point. While these methods are computationally simple, they struggle to fully characterize the continuous state evolution between adjacent sampling time points.

[0003] When all endpoints of a time interval meet supply and demand balance, but a rapid increase in net load, a sharp drop in renewable energy output, or insufficient ramp-up capability of generating units occurs within the interval, the discrete-time model may determine that the system is safe, while a short-term load loss has actually occurred during continuous operation. While shortening the sampling interval can improve time resolution, the model size increases significantly with the number of sampling points, and limited sampling still cannot fundamentally eliminate missed detections of intra-segment risks. The intra-segment implicit risk referred to in this invention refers to a short-term power supply insufficiency risk that occurs between adjacent sampling times, caused by rapid changes in renewable energy output, load fluctuations, insufficient ramp-up capability of generating units, or limited line transmission, even though the discrete sampling time or time segment endpoints all meet system operating constraints and are not identified by discrete-time verification. Summary of the Invention

[0004] The purpose of this invention is to provide a method for continuous-time operational reliability assessment of power systems that considers the uncertainty of time-varying trajectories of new energy sources, comprising the following steps:

[0005] Step 1) Obtain several complete source-load time-series scenes with probabilities in the time domain for evaluation, and convert the complete source-load time-series scenes into continuous-time source-load trajectories;

[0006] Step 2) Establish continuous-time response models for each source load time-series scenario;

[0007] Step 3) Define operational reliability metrics based on continuous load reduction trajectories for each scenario;

[0008] Step 4) Represent the continuous-time runtime response model corresponding to each source load time-series scenario in... The Bernstein basis functions are transformed into a finite-dimensional mixed-integer optimization model; the finite-dimensional mixed-integer optimization model is solved to obtain the optimal control coefficients of the power system operating characteristic parameters;

[0009] Step 5) Based on the optimal control coefficients of the power system operating characteristic parameters, reconstruct the continuous load reduction trajectory and calculate the evaluation results after reconstructing the continuous load reduction trajectory.

[0010] Furthermore, in step 1), the steps for constructing the continuous-time source load trajectory are as follows:

[0011] Step 1.1) Denote the source-load scene set as New energy power stations are integrated into The set of nodes is The indexes for scenarios, new energy power stations, and nodes are respectively and ;

[0012] For the scene Corresponding probability Normalization is performed as follows:

[0013] (1)

[0014] Step 1.2) Evaluate the time domain Divided into The nth adjacent time segment, the th The time segments are shown below:

[0015] (2)

[0016] For any time The physical time is mapped to a locally normalized time interval, as shown below:

[0017] (3)

[0018] In the formula, The segment length is... For locally normalized time variables;

[0019] The segment lengths are shown below:

[0020] (4)

[0021] Step 1.3) Define within the locally normalized time interval The second Bernstein basis functions are shown below:

[0022] (5)

[0023] (6)

[0024] In the formula, ;

[0025] Step 1.4) Define the Bernstein basis function column vector as follows:

[0026] (7)

[0027] Step 1.5) Constructing the Scene China New Energy's contribution in the first The continuous time trajectory within each time segment is shown below:

[0028] (8)

[0029] In the formula, For new energy power stations Bernstein control coefficient vector;

[0030] Scenario building The load of the middle node is at the The continuous time trajectory within each time segment is shown below:

[0031] (9)

[0032] In the formula, For nodes The load control coefficient vector.

[0033] Furthermore, in step 2), the objective function of the continuous-time response model corresponding to each source-load time-series scenario is as follows:

[0034] (10)

[0035] In the formula, , and These are respectively a set of conventional generating units, a set of new energy power stations, and a set of nodes; For the scene The operating costs within the next evaluation time domain; , The units In the Start-up and shutdown action variables at the initial boundary of each time segment; , These are the continuous active power output of generating units, the power reduction of new energy sources, and the power reduction of nodal loads; , These are the unit start-up cost coefficient, shutdown cost coefficient, continuous power generation cost coefficient, renewable energy reduction cost coefficient, and load reduction cost coefficient, respectively. Number of time segments;

[0036] The constraints of the continuous-time operation response model for each source-load scenario include unit operating status and start-up / shutdown action constraints, minimum continuous start-up time constraints that should be met after unit start-up, minimum continuous shutdown time constraints that should be met after unit shutdown, the upper and lower limits of the unit's continuous active power output under the operating status constraints, the upward / downward ramp rate constraints that should be met at any time for the unit's continuous active power output trajectory except for the unit start-up / shutdown boundaries, constraints on the power reduction of new energy sources and the power reduction of nodal loads, power balance constraints, and line power flow / line transmission capacity and reference node phase angle constraints.

[0037] Furthermore, the constraints on the unit's operating status and start-up / shutdown actions are as follows:

[0038] (11)

[0039] (12)

[0040] (13)

[0041] (14)

[0042] In the formula, , The units The operational status within adjacent time segments; To evaluate the unit before the start of the time domain The running status;

[0043] The minimum continuous start-up time constraints that the unit must meet after startup are as follows:

[0044] (15)

[0045] The minimum continuous downtime constraints that the unit must meet after shutdown are as follows:

[0046] (16)

[0047] In the formula, , The units The minimum continuous start-up time and minimum continuous shutdown time; ξ is the auxiliary time variable in the integral;

[0048] The continuous active power output of the unit should meet the upper and lower limits of the output under the constraints of the operating conditions, as shown below:

[0049] (17)

[0050] In the formula, The units Minimum allowable output and maximum allowable output;

[0051] Except for the unit start-up and shutdown boundaries, the continuous active power output trajectory of the unit should meet the up / down ramp rate constraints at any time as shown below:

[0052] (18)

[0053] In the formula, , The units Maximum uphill and maximum downhill speeds;

[0054] The power reduction constraints for renewable energy sources and nodal loads are as follows:

[0055] (19)

[0056] (20)

[0057] In the formula, , These are power reductions from new energy sources and power reductions from the load;

[0058] The power balance constraints are as follows:

[0059] (twenty one)

[0060] In the formula, and Access nodes A collection of conventional generating units and a collection of new energy power stations; For nodes The set of connected adjacent nodes;

[0061] The line power flow / line transmission capacity and reference node phase angle constraints are shown below:

[0062] (twenty two)

[0063] (twenty three)

[0064] (twenty four)

[0065] In the formula, for The susceptance; They are nodes and nodes The continuous-time phase angle trajectory; For the line The transmission capacity; This is a reference node.

[0066] Furthermore, in step 3), the operational reliability indicators include the probability of load failure at any given time, the time-domain average probability of load failure, the expected duration of load failure, and the expected value of insufficient power.

[0067] The probability of load loss at any given time is shown below:

[0068] (25)

[0069] In the formula, To reduce power consumption for the total system load, It is an indicator function; This is a non-negative decision threshold used to determine whether load reduction has occurred in the system;

[0070] The time-domain average failure probability is shown below:

[0071] (26)

[0072] In the formula, T represents the evaluation time domain. The total duration;

[0073] The expected duration of the loss of load is as follows:

[0074] (27)

[0075] In the formula, This refers to the cumulative duration of load loss.

[0076] The expected value for low battery is as follows:

[0077] (28)

[0078] In the formula, This refers to the power shortage in this scenario.

[0079] Furthermore, the total system load reduction power is shown below:

[0080] (29)

[0081] The cumulative unload duration is shown below:

[0082] (30)

[0083] In the formula, Reduce power consumption to reduce the total system load.

[0084] Furthermore, in step 4), the objective function of the finite-dimensional mixed-integer optimization model is as follows:

[0085] (31)

[0086] In the formula, It is a column vector of all 1s whose dimension is equal to the exponent plus 1; For the scene The corresponding finite-dimensional operating cost; and These are the Bernstein coefficient vectors corresponding to unit output, renewable energy reduction, and load reduction, respectively. The variables are for segmented start-up and shutdown; the equality constraints of the finite-dimensional mixed integer optimization model include node power balance constraints and line power flow and reference node phase angle constraints.

[0087] The inequality constraints of the finite-dimensional mixed-integer optimization model include upper and lower limits of unit output, boundary constraints of renewable energy reduction and load reduction, line transmission capacity constraints, and unit continuous ramping constraints.

[0088] The unit start-up and shutdown logic and minimum start-up and shutdown time constraints of the finite-dimensional mixed-integer optimization model are equivalently transformed into state summation constraints on the corresponding time segments, as shown below:

[0089] (32)

[0090] In the formula, Auxiliary segment index used for summation; The duration of the equal-length start / stop decision segments; Indicates rounding up;

[0091] The trajectory connection constraints at the unit start-up and shutdown boundaries of the finite-dimensional mixed-integer optimization model are shown below:

[0092]

[0093] (33)

[0094] In the formula, , The units Maximum uphill speed and maximum downhill speed.

[0095] Furthermore, the node power balance constraints are as follows:

[0096] (34)

[0097] In the formula, and These are known coefficient vectors for available renewable energy sources and nodal loads, respectively; This represents the vector of line power flow coefficients;

[0098] The line power flow and reference node phase angle constraints are shown below:

[0099] (35)

[0100] In the formula, The vector of node phase angle coefficients; It is a K+1 dimensional column vector of all zeros;

[0101] The upper and lower limits of the unit's output are constrained as follows:

[0102] (36)

[0103] The boundary constraints for renewable energy reduction and load reduction are as follows:

[0104] (37)

[0105] The line transmission capacity constraints are as follows:

[0106] (38)

[0107] The continuous ramp-up constraints for the unit are as follows:

[0108] (39)

[0109] (40)

[0110] In the formula, It is a K-dimensional column vector of all 1s; Let be the first-order difference matrix of the Kth trajectory.

[0111] Furthermore, the power system operation characteristic parameters include unit output, renewable energy reduction, load reduction, line power flow, and node phase angle trajectory.

[0112] Furthermore, in step 5), the scene The following is the set of times when load shedding occurs:

[0113] (41)

[0114] The evaluation results after reconstructing the continuous load reduction trajectory include operational reliability indicators and expected operating costs; among them, the operational reliability indicators include the expected value of insufficient power, the probability of load loss at any time, the time-domain average probability of load loss, and the expected duration of load loss, and the expected operating costs are economic statistics results;

[0115] The expected value for low battery is as follows:

[0116] (42)

[0117] In the formula, For the scene The power shortage within the next evaluation time domain;

[0118] Scene The power shortage in the next evaluation time domain is shown below:

[0119] (43)

[0120] In the formula, This is the total load reduction control factor for the system;

[0121] The total system load reduction control coefficients are shown below:

[0122] (44)

[0123] (45)

[0124] (46)

[0125] In the formula, The continuous time trajectory for reducing node load during reconstruction; The total system load reduction for the reconstructed system is achieved through continuous time trajectory reduction. Optimal control coefficients for load reduction at nodes in each scenario; A column vector of K-th degree Bernstein basis functions;

[0126] The probability of load loss at any given time is shown below:

[0127] (47)

[0128] The time-domain average failure probability is shown below:

[0129] (48)

[0130] The expected operating costs are as follows:

[0131] (49)

[0132] In the formula, For the scene The optimal objective function value of the corresponding finite-dimensional mixed-integer optimization model;

[0133] The expected duration of the loss of load is as follows:

[0134] (50)

[0135] In the formula, This represents the time-domain average probability of load failure.

[0136] The technical effects of this invention are as follows:

[0137] This invention uses a continuous-time trajectory to uniformly represent renewable energy output, load demand, and system dispatch response, identifying short-term power shortage risks within dispatch intervals caused by rapid changes in source and load, limited unit ramp-up, and network transmission constraints. Based on the solution space transformation of Bernstein polynomials, the infinite-dimensional operational reliability assessment model in the continuous-time domain is transformed into a finite-dimensional mixed-integer model that can be processed by existing optimization solvers, maintaining continuous-time representation capabilities while avoiding model size expansion caused by high-density discrete sampling. This method can quantify load loss risk from three dimensions: probability of occurrence, temporal location, and power consequences, providing a technical basis for operational risk identification and dispatch decisions in high-proportion renewable energy power systems. Attached Figure Description

[0138] Figure 1 This is a topology diagram of the modified IEEE 30-node system according to the present invention;

[0139] Figure 2 This is a comparison chart of discrete-time and continuous-time inputs of wind power output and net load in the representative scenario of this invention;

[0140] Figure 3 The diagram shows the moment-time load loss probability and probability difference between the discrete-time model and the continuous-time model of this invention.

[0141] Figure 4 This is a magnified view of the load reduction trajectory and a local portion of the discrete-time model and the continuous-time model of this invention in the representative scenario of this invention. Detailed Implementation

[0142] The present invention will be further described below with reference to embodiments, but it should not be construed that the scope of the present invention is limited to the following embodiments. Various substitutions and modifications made based on ordinary technical knowledge and common practices in the art without departing from the above-described technical concept of the present invention should be included within the scope of protection of the present invention.

[0143] Example 1:

[0144] See Figures 1 to 4 A method for assessing the continuous-time operational reliability of power systems considering the time-varying trajectory uncertainties of new energy sources includes the following steps:

[0145] Step 1) Obtain several complete source-load time-series scenes with probabilities in the time domain for evaluation, and convert the complete source-load time-series scenes into continuous-time source-load trajectories;

[0146] Step 2) Establish continuous-time response models for each source load time-series scenario;

[0147] Step 3) Define operational reliability metrics based on continuous load reduction trajectories for each scenario;

[0148] Step 4) Represent the continuous-time runtime response model corresponding to each source load time-series scenario in... The Bernstein basis functions are transformed into a finite-dimensional mixed-integer optimization model; the finite-dimensional mixed-integer optimization model is solved to obtain the optimal control coefficients of the power system operating characteristic parameters;

[0149] Solve finite-dimensional mixed-integer optimization models using existing integer optimization solvers;

[0150] Step 5) Based on the optimal control coefficients of the power system operating characteristic parameters, reconstruct the continuous load reduction trajectory and calculate the evaluation results after reconstructing the continuous load reduction trajectory.

[0151] Example 2:

[0152] The main structure of this embodiment is the same as that of embodiment 1. Further, in step 1), the steps for constructing the continuous-time source load trajectory are as follows:

[0153] Step 1.1) Denote the source-load scene set as New energy power stations are integrated into The set of nodes is The indexes for scenarios, new energy power stations, and nodes are respectively and ;

[0154] For the scene Corresponding probability Normalization is performed as follows:

[0155] (1)

[0156] Step 1.2) Evaluate the time domain Divided into The nth adjacent time segment, the th The time segments are shown below:

[0157] (2)

[0158] For any time The physical time is mapped to a locally normalized time interval, as shown below:

[0159] (3)

[0160] In the formula, The segment length is... For locally normalized time variables;

[0161] The segment lengths are shown below:

[0162] (4)

[0163] Step 1.3) Define within the locally normalized time interval The second Bernstein basis functions are shown below:

[0164] (5)

[0165] (6)

[0166] In the formula, ;

[0167] Step 1.4) Define the Bernstein basis function column vector as follows:

[0168] (7)

[0169] Step 1.5) Constructing the Scene China New Energy's contribution in the first The continuous time trajectory within each time segment is shown below:

[0170] (8)

[0171] In the formula, For new energy power stations Bernstein control coefficient vector;

[0172] Scenario building The load of the middle node is at the The continuous time trajectory within each time segment is shown below:

[0173] (9)

[0174] In the formula, For nodes The load control coefficient vector; the new energy output trajectory and nodal load trajectory at the common endpoint of adjacent time segments maintain numerical continuity, satisfying... .

[0175] Example 3:

[0176] The main structure of this embodiment is the same as any one of embodiments 1-2. Further, in step 2), the objective function of the continuous-time running response model corresponding to each source load time-series scenario is as follows:

[0177] (10)

[0178] In the formula, , and These are respectively a set of conventional generating units, a set of new energy power stations, and a set of nodes; For the scene The operating costs within the next evaluation time domain; , The units In the Start-up and shutdown action variables at the initial boundary of each time segment; , These are the continuous active power output of generating units, the power reduction of new energy sources, and the power reduction of nodal loads; , These are the unit start-up cost coefficient, shutdown cost coefficient, continuous power generation cost coefficient, renewable energy reduction cost coefficient, and load reduction cost coefficient, respectively. Number of time segments;

[0179] The constraints of the continuous-time operation response model for each source-load scenario include unit operating status and start-up / shutdown action constraints, minimum continuous start-up time constraints that should be met after unit start-up, minimum continuous shutdown time constraints that should be met after unit shutdown, the upper and lower limits of the unit's continuous active power output under the operating status constraints, the upward / downward ramp rate constraints that should be met at any time for the unit's continuous active power output trajectory except for the unit start-up / shutdown boundaries, constraints on the power reduction of new energy sources and the power reduction of nodal loads, power balance constraints, and line power flow / line transmission capacity and reference node phase angle constraints.

[0180] Example 4:

[0181] The main structure of this embodiment is the same as any one of embodiments 1-3. Furthermore, the constraints on the unit operating status and start-up / shutdown actions are as follows:

[0182] (11)

[0183] (12)

[0184] (13)

[0185] (14)

[0186] In the formula, , The units The operational status within adjacent time segments; To evaluate the unit before the start of the time domain The running status;

[0187] The minimum continuous start-up time constraints that the unit must meet after startup are as follows:

[0188] (15)

[0189] The minimum continuous downtime constraints that the unit must meet after shutdown are as follows:

[0190] (16)

[0191] In the formula, , The units The minimum continuous start-up time and minimum continuous shutdown time; ξ is the auxiliary time variable in the integral;

[0192] The continuous active power output of the unit should meet the upper and lower limits of the output under the constraints of the operating conditions, as shown below:

[0193] (17)

[0194] In the formula, The units Minimum allowable output and maximum allowable output;

[0195] Except for the unit start-up and shutdown boundaries, the continuous active power output trajectory of the unit should meet the up / down ramp rate constraints at any time as shown below:

[0196] (18)

[0197] In the formula, , The units Maximum uphill and maximum downhill speeds;

[0198] The power reduction constraints for renewable energy sources and nodal loads are as follows:

[0199] (19)

[0200] (20)

[0201] In the formula, , These are power reductions from new energy sources and power reductions from the load;

[0202] The power balance constraints are as follows:

[0203] (twenty one)

[0204] In the formula, and Access nodes A collection of conventional generating units and a collection of new energy power stations; For nodes The set of connected adjacent nodes;

[0205] The line power flow / line transmission capacity and reference node phase angle constraints are shown below:

[0206] (twenty two)

[0207] (twenty three)

[0208] (twenty four)

[0209] In the formula, for The susceptance; They are nodes and nodes The continuous-time phase angle trajectory; For the line The transmission capacity; This is a reference node.

[0210] Example 5:

[0211] The main structure of this embodiment is the same as any one of embodiments 1-4. Further, in step 3), the operational reliability indicators include the momentary load failure probability, the time-domain average load failure probability, the expected load failure duration, and the expected power shortage value.

[0212] The probability of load loss at any given time is shown below:

[0213] (25)

[0214] In the formula, To reduce power consumption for the total system load, It is an indicator function; This is a non-negative decision threshold used to determine whether load reduction has occurred in the system;

[0215] The time-domain average failure probability is shown below:

[0216] (26)

[0217] In the formula, T represents the evaluation time domain. Total duration, units and time variables Consistent, ;

[0218] The expected duration of the loss of load is as follows:

[0219] (27)

[0220] In the formula, This refers to the cumulative duration of load loss.

[0221] The expected value for low battery is as follows:

[0222] (28)

[0223] In the formula, This refers to the power shortage in this scenario.

[0224] Example 6:

[0225] The main structure of this embodiment is the same as any one of embodiments 1-5. Furthermore, the total system load reduction power is as follows:

[0226] (29)

[0227] The cumulative unload duration is shown below:

[0228] (30)

[0229] In the formula, Reduce power consumption to reduce the total system load.

[0230] Example 7:

[0231] The main structure of this embodiment is the same as any one of embodiments 1-6. Further, in step 4), the objective function of the finite-dimensional mixed-integer optimization model is as follows:

[0232] (31)

[0233] In the formula, It is a column vector of all 1s whose dimension is equal to the exponent plus 1; For the scene The corresponding finite-dimensional operating cost; and These are the Bernstein coefficient vectors corresponding to unit output, renewable energy reduction, and load reduction, respectively. The variables are for segmented start-up and shutdown; the equality constraints of the finite-dimensional mixed integer optimization model include node power balance constraints and line power flow and reference node phase angle constraints.

[0234] The inequality constraints of the finite-dimensional mixed-integer optimization model include upper and lower limits of unit output, boundary constraints of renewable energy reduction and load reduction, line transmission capacity constraints, and unit continuous ramping constraints.

[0235] The unit start-up and shutdown logic and minimum start-up and shutdown time constraints of the finite-dimensional mixed-integer optimization model are equivalently transformed into state summation constraints on the corresponding time segments, as shown below:

[0236] (32)

[0237] In the formula, Auxiliary segment index used for summation; The duration of the equal-length start / stop decision segments; Indicates rounding up;

[0238] The trajectory connection constraints at the unit start-up and shutdown boundaries of the finite-dimensional mixed-integer optimization model are shown below:

[0239]

[0240] (33)

[0241] In the formula, , The units The maximum uphill and downhill rates; when there is no start-stop transition, the trajectory connection constraints at the start-stop boundary of the unit in the finite-dimensional mixed integer optimization model ensure the continuity of the endpoint output and the first derivative of the endpoint in adjacent time segments; when a start-stop transition occurs, the trajectory connection constraints at the start-stop boundary of the unit in the finite-dimensional mixed integer optimization model relax the corresponding connection constraints through the start action variable or the stop action variable.

[0242] Example 8:

[0243] The main structure of this embodiment is the same as any one of embodiments 1-7. Furthermore, the node power balance constraints are as follows:

[0244] (34)

[0245] In the formula, and These are known coefficient vectors for available renewable energy sources and nodal loads, respectively; This represents the vector of line power flow coefficients;

[0246] The line power flow and reference node phase angle constraints are shown below:

[0247] (35)

[0248] In the formula, The vector of node phase angle coefficients; It is a K+1 dimensional column vector of all zeros;

[0249] The upper and lower limits of the unit's output are constrained as follows:

[0250] (36)

[0251] The boundary constraints for renewable energy reduction and load reduction are as follows:

[0252] (37)

[0253] The line transmission capacity constraints are as follows:

[0254] (38)

[0255] The continuous ramp-up constraints for the unit are as follows:

[0256] (39)

[0257] (40)

[0258] In the formula, It is a K-dimensional column vector of all 1s; Let be the first-order difference matrix of the Kth trajectory.

[0259] Example 9:

[0260] The main structure of this embodiment is the same as any one of embodiments 1-8. Furthermore, the power system operation characteristic parameters include unit output, new energy reduction, load reduction, line power flow, and node phase angle trajectory.

[0261] Example 10:

[0262] The main structure of this embodiment is the same as any one of embodiments 1-9. Further, in step 5), the scenario... The following is the set of times when load shedding occurs:

[0263] (41)

[0264] Determination threshold Used to distinguish between actual load reduction and minimal non-zero load reduction caused by the feasibility tolerance of the optimization solver and floating-point operation errors; in the theoretical definition, it can be taken as... In numerical calculations, a small positive number with dimensions consistent with the model power can be used to eliminate numerical calculation errors; in this embodiment, a value is taken as... When the total load reduction power of the system exceeds this threshold, the system is deemed to have insufficient power supply.

[0265] The evaluation results after reconstructing the continuous load reduction trajectory include operational reliability indicators and expected operating costs; among them, the operational reliability indicators include the expected value of insufficient power, the probability of load loss at any time, the time-domain average probability of load loss, and the expected duration of load loss, and the expected operating costs are economic statistics results;

[0266] The expected value for low battery is as follows:

[0267] (42)

[0268] In the formula, For the scene The power shortage within the next evaluation time domain;

[0269] Scene The power shortage in the next evaluation time domain is shown below:

[0270] (43)

[0271] In the formula, This is the total load reduction control factor for the system;

[0272] The total system load reduction control coefficients are shown below:

[0273] (44)

[0274] (45)

[0275] (46)

[0276] In the formula, The continuous time trajectory for reducing node load during reconstruction; The total system load reduction for the reconstructed system is achieved through continuous time trajectory reduction. Optimal control coefficients for load reduction at nodes in each scenario; It is a column vector of K-order Bernstein basis functions, consisting of K+1 Bernstein basis functions;

[0277] The probability of load loss at any given time is shown below:

[0278] (47)

[0279] The time-domain average failure probability is shown below:

[0280] (48)

[0281] The expected operating costs are as follows:

[0282] (49)

[0283] In the formula, For the scene The optimal objective function value of the corresponding finite-dimensional mixed-integer optimization model, i.e., the scenario-optimal operating cost obtained under the optimal unit operating state, start-up and shutdown actions, and optimal continuous operation response control coefficients; superscript " " indicates the optimal value;

[0284] The expected duration of the loss of load is as follows:

[0285] (50)

[0286] In the formula, This represents the time-domain average probability of load failure.

[0287] Example 11:

[0288] The main structure of this embodiment is the same as any one of embodiments 1-10. Furthermore, the specific generation methods of the complete source-load time series scene include three main categories: statistical model-based, deep learning-based, and hybrid methods that consider spatiotemporal correlation.

[0289] 1. Generative methods based on statistical models

[0290] The core of this type of method is to extract statistical patterns from historical data and then generate new scenarios through sampling.

[0291] Probabilistic model sampling: First, the historical probability distributions of power output and load from new energy sources such as wind power and photovoltaics are fitted, such as normal distributions and Beta distributions. Then, a large number of scenarios are generated by sampling from these distributions. An improved approach is to dynamically adjust the sampling probability interval of the prediction error based on the historical fluctuation characteristics of new energy sources, and then combine it with methods such as Latin hypercube sampling to make the generated scenarios more closely match actual fluctuations.

[0292] Time series modeling focuses on the temporal dependencies of data. For example, Markov chains are used to describe the temporal characteristics of prediction errors and generate scenarios. More sophisticated methods, such as Markov chain Monte Carlo simulations, can generate a large number of source-load scenarios, followed by scenario reduction to obtain representative typical scenarios.

[0293] 2. Deep Learning-Based Generative Methods

[0294] Deep learning methods excel at automatically learning complex nonlinear patterns in data without requiring manual setting of probability distributions.

[0295] Generative Adversarial Networks (GANs) generate scenarios that are highly similar to historical data distributions through a game between a generator and a discriminator. For example, a temporal GAN ​​can generate high-quality load scenarios, while a temporal GAN ​​combining a temporal convolutional network and a multi-head attention mechanism can generate source load scenarios at multiple time scales such as daily, weekly, and monthly.

[0296] Variational autoencoders: generate new samples by learning the latent feature space of data, and can be used to describe the correlation of multi-source load scenarios.

[0297] Diffusion model: Generates data through a stepwise denoising process, effectively handling high-dimensional data. It can be embedded into long short-term memory networks to improve the learning ability for complex temporal features.

[0298] 3. Hybrid approach considering spatiotemporal correlation

[0299] This type of approach emphasizes joint modeling of the source and load as a whole to preserve their physical coupling characteristics.

[0300] Joint probability distribution modeling: directly model and sample the multidimensional joint probability distribution of wind, solar, and load to ensure the relevance of the generated scene from the source.

[0301] Applications of Copula theory: Copula functions can model the marginal distributions of variables separately from their correlation structures, making them a powerful tool for handling complex correlations. For example, the Copula-Markov chain model can be used to handle temporal correlations, while the R-Vine Copula model can be used to handle spatial correlations.

[0302] Typical coupling pattern matching: First, define typical source-load coupling patterns based on meteorological and other conditions. Then, use generative networks or clustering methods to generate wind-solar and load scenarios respectively. Finally, match and combine them according to the patterns.

[0303] Multi-timescale decomposition: For medium- and long-term planning, the problem can be decomposed into different time scales, such as year, solar term, and hour. For example, the annual electricity consumption can be generated first through a progressive diffusion model, and then gradually refined to solar term and 15-day electricity consumption, and finally generated hourly power scenarios.

[0304] Example 12:

[0305] This invention aims to provide a continuous-time operation reliability assessment method for power systems that addresses implicit risks within scheduling intervals, thereby compensating for risk omissions caused by discrete-time point verification, and forming a finite-dimensional model that can be processed by existing optimization solvers while retaining continuous-time characterization capabilities.

[0306] The technical solution to achieve the objective of this invention is as follows: Source-load uncertainty is characterized by probabilistic renewable energy output and load time-series scenarios, and a continuous source-load trajectory is constructed. For each source-load scenario, a continuous-time operation response model is established, taking into account unit start-up and shutdown, ramp-up capability, node power balance, and network security constraints. Furthermore, Bernstein polynomials are used to perform a finite-dimensional transformation of the solution space, converting the continuous-time operation response model into a finite-dimensional mixed-integer optimization model. After solving, the continuous load reduction trajectory for each scenario is reconstructed, and operational reliability indicators such as the probability of load loss at any given time, the time-domain average probability of load loss, the expected duration of load loss, and the expected value of insufficient power are calculated. These indicators are then statistically summarized according to scenario probabilities, thereby completing the continuous-time operation reliability assessment of the power system.

[0307] The specific technical solution is as follows:

[0308] Step S1: Construct continuous-time source-load trajectories. Several complete source-load time-series scenarios with probabilities within the evaluation time domain are used as input. Each scenario includes the corresponding new energy output sequence and node load sequence at each prediction time. The complete source-load time-series scenarios can be provided by an external probability prediction platform, an operational data platform, or a preset scenario library; this invention does not limit the specific scenario generation method.

[0309] Assuming the source load scene set is New energy power stations are integrated into The set of nodes is The indexes for scenarios, new energy power stations, and nodes are respectively and Scene The corresponding probability is The probabilities for each scenario are all non-negative and, after normalization, satisfy the following: .

[0310] Evaluation time domain Divided into The nth adjacent time segment, the th Each time segment is recorded as follows:

[0311]

[0312] Its segment length is For any given time Physical time is mapped to a locally normalized time interval using the following formula. :

[0313]

[0314] In equation (2), For locally normalized time variables; defined within the locally normalized time interval. Second Bernstein basis functions:

[0315]

[0316]

[0317] Define the Bernstein basis function column vector as follows:

[0318]

[0319] For the scenario The output of new energy sources or the load at nodes in the first phase The Bernstein polynomial representation is used within each time segment:

[0320]

[0321] In the formula, Representing a scene China's new energy power output or node load is at the first Continuous time trajectory within a time segment; This represents the corresponding Bernstein control coefficient vector. Specifically, the scenario... New Energy Power Station The continuous time can be represented by the output trajectory as follows:

[0322]

[0323] in, For new energy power stations Bernstein control coefficient vector.

[0324] Scene Next node The continuous-time load demand trajectory is represented as:

[0325]

[0326] in, For nodes The load control coefficient vector.

[0327] The aforementioned new energy output and node load control coefficients are predetermined based on source-load time-series scenario data and can be obtained through methods such as endpoint interpolation, endpoint value and endpoint change rate constraints, or trajectory fitting.

[0328] To ensure the continuity of trajectory values ​​at the common endpoints of adjacent time segments, the control coefficients of adjacent segments must satisfy:

[0329]

[0330] When the source load trajectory needs to satisfy first-order smoothness, the first-order derivative at the common endpoint of adjacent time segments further satisfies:

[0331]

[0332] The left side of equation (10) is the first The first derivative at the right endpoint of the nth time segment, with the right side being the nth time segment. The first derivative at the left endpoint of each time segment. Equation (9) ensures the continuity of the trajectory values ​​at the common endpoint of adjacent time segments; Equation (10) ensures the continuity of the first derivative of the trajectory at the common endpoint.

[0333] By using Bernstein polynomial representation within time segments and trajectory connection conditions between adjacent time segments, discrete renewable energy output and nodal load time-series data are transformed into continuous-time source-load trajectories covering the entire evaluation time domain. The continuous-time source-load trajectories and their Bernstein control coefficients obtained in step S1 serve as known input parameters for subsequent continuous-time operational response models and are not used as optimization decision variables.

[0334] Step S2: Establish a continuous-time operational response model. Based on the continuous source-load trajectory obtained in Step S1, establish a continuous-time operational response model for each source-load scenario to describe the system's operational status and scheduling response under given renewable energy output and node load trajectories. The model uses unit operating status, startup actions, shutdown actions, unit output, renewable energy reduction, load reduction, line power flow, and node phase angle as operational variables, and satisfies unit operation and network security constraints throughout the entire evaluation time domain.

[0335] For arbitrary source load scenarios The objective is to minimize the sum of unit start-up and shutdown costs and continuous operation costs within the evaluation time domain.

[0336]

[0337] In the formula, , and These are respectively a set of conventional generating units, a set of new energy power stations, and a set of nodes; For the scene The operating costs within the next evaluation time domain; and The units In the Start-up and shutdown action variables at the initial boundary of each time segment; and These are the continuous active power output of generating units, the power reduction of new energy sources, and the power reduction of nodal loads; and These are the unit start-up cost coefficient, shutdown cost coefficient, continuous power generation cost coefficient, renewable energy reduction cost coefficient, and load reduction cost coefficient, respectively.

[0338] The following conditions must be met between the unit's operating status and start-up / shutdown actions:

[0339]

[0340] In the formula, and The units The operational status within adjacent time segments; Indicates the unit before the start of the evaluation time domain. The running status.

[0341] The unit must meet the minimum continuous start-up time constraint after startup:

[0342]

[0343] The unit must meet the minimum continuous downtime constraint after shutdown:

[0344]

[0345] In the formula, and The units The minimum continuous start-up time and minimum continuous shutdown time are denoted by ξ; ξ is the auxiliary time variable in the integral.

[0346] The continuous active power output of the unit should meet the upper and lower limits of output under operating condition constraints:

[0347]

[0348] In the formula, and The units Minimum allowable output and maximum allowable output.

[0349] Except for the unit start-up and shutdown boundaries, the continuous active power output trajectory of the unit should meet the uphill and downhill ramp rate constraints at any time:

[0350]

[0351] In the formula, and The units Maximum uphill speed and maximum downhill speed.

[0352] The power reduction from renewable energy sources and the power reduction from nodal loads should respectively meet the following requirements:

[0353]

[0354]

[0355] Equations (17) and (18) ensure that the power reduction of new energy sources and the power reduction of load are non-negative, and limit that they shall not exceed the available new energy output and node load demand given in step S1.

[0356] The power balance constraint must be satisfied at any time and at any node:

[0357]

[0358] In the formula, and Access nodes A collection of conventional generating units and a collection of new energy power stations; For nodes The set of connected adjacent nodes; For the scene Downline At any moment By node Flow to Node The positive trend.

[0359] The line power flow, line transmission capacity, and reference node phase angle should meet the following requirements:

[0360]

[0361] In the formula, for The susceptance; They are nodes and nodes The continuous-time phase angle trajectory; For the line The transmission capacity; This is a reference node.

[0362] By solving the above model, we can obtain the continuous-time operational response trajectories of unit output, renewable energy reduction, load reduction, line power flow, and node phase angle under various source-load scenarios. This provides a foundation for the subsequent finite-dimensional transformation of the operational response model, reconstruction of the continuous load reduction trajectory, and calculation of operational reliability indicators.

[0363] Step S3 defines the operational reliability index based on the continuous load reduction trajectory for each scenario. This step only provides the mathematical definition of the operational reliability index; the specific values ​​are calculated after the finite-dimensional transformation and solution of the operational response model in step S4, and after reconstructing the continuous load reduction trajectory in step S5.

[0364] First, regarding the scenario The total load reduction power of the system is obtained by summing the load reduction power of each node at any given time.

[0365]

[0366] when When it exceeds the judgment threshold, it indicates that the scene The system at time A power supply shortage occurs. Based on this, the probability of load failure at any given time is defined as:

[0367]

[0368] In the formula, This is an indicator function; it takes the value 1 if the condition inside the parentheses is true, and 0 otherwise. This is a non-negative threshold used to determine whether a load reduction has occurred in the system; it is theoretically defined as zero, but can be taken as a very small positive tolerance in numerical calculations.

[0369] Integrating the probability of load loss at time point in the evaluation time domain and taking the time average, we obtain the time-domain average probability of load loss:

[0370]

[0371] For the scenario Define the cumulative unload duration:

[0372]

[0373] The expected duration of load loss is obtained by weighting the load loss duration for each scenario according to the scenario probability:

[0374]

[0375] For the scenario Integrating the total load reduction power over the evaluation time domain yields the power shortage for this scenario:

[0376]

[0377] Further weighting according to scenario probability yields the expected value for insufficient battery power:

[0378]

[0379] In summary, the operational reliability metrics defined in this step include the probability of load failure at any given time. Time-domain average load loss probability Expected duration of load loss and insufficient battery life .

[0380] Step S4: Represent the continuous-time operational response models of each scenario in Step S2 on the K-th order Bernstein basis defined in Step S1, and convert them into finite-dimensional mixed-integer optimization models. Based on the local time mapping, Bernstein basis functions, and continuous-time trajectory expression given in Step S1, represent the continuous operational response variables such as unit output, renewable energy reduction, load reduction, line power flow, and node phase angle as corresponding Bernstein control coefficient vectors. Among these, the renewable energy output control coefficient and node load control coefficient in Step S1 are used as known parameters, and the control coefficients of the operational response variables are used as optimization decision variables.

[0381] To complete the aforementioned finite-dimensional transformation, the integral, derivative, and interval boundaries of the continuous trajectory adopt the following basic relationships:

[0382] The integral of the continuous trajectory over a single time segment can be accurately calculated using all control coefficients:

[0383]

[0384] In the formula, Let be a column vector of all 1s with dimension equal to the degree plus 1; this formula shows that the interval integral can be accurately calculated by the sum of all control coefficients.

[0385] The time derivative of a continuous trajectory can be expressed as the difference between adjacent control coefficients:

[0386]

[0387] The above formula shows that the coefficient of the trajectory derivative is determined by the difference between adjacent control coefficients, and its degree is one order lower than that of the original trajectory.

[0388] If a continuous trajectory is constant at zero throughout the entire time segment, then its control coefficient vector satisfies:

[0389]

[0390] The above equation shows that the necessary and sufficient condition for a continuous trajectory to be zero throughout the entire time segment is that all Bernstein control coefficients are zero.

[0391] For continuous trajectories that need to satisfy uniform upper and lower bounds throughout the entire time segment, apply corresponding boundaries to all control coefficients:

[0392]

[0393] The above formula constructs the interval boundary conditions for a continuous trajectory by applying uniform upper and lower bounds to all control coefficients.

[0394] Based on the convex hull property of Bernstein polynomials, the boundary of the control coefficients guarantees that the continuous trajectory satisfies:

[0395]

[0396] The above formula represents the continuous trajectory interval constraint derived from the coefficient boundary.

[0397] (1) Reconstruction of the objective function

[0398] Based on the above interval integration relationship, the scenario in step S2... The continuous-time objective function is transformed into a finite-dimensional objective function:

[0399]

[0400] In the formula, For the scene The corresponding finite-dimensional operating cost; and These are the Bernstein coefficient vectors corresponding to unit output, renewable energy reduction, and load reduction, respectively. These are variables for segmented start-up and shutdown; the cost coefficients are consistent with the definitions in the continuous-time objective function of step S2.

[0401] (2) Reconstruction of equality constraints

[0402] In continuous-time models, nodal power balance, line power flow definition, reference node phase angle, and the continuity condition of common endpoints unaffected by unit start-up and shutdown are all equality constraints. Utilizing the linear independence of Bernstein basis functions, these constraints are transformed into control coefficient-by-control equations.

[0403] The node power balance constraints are reconstructed as follows:

[0404]

[0405] In the formula, and These are known coefficient vectors for available renewable energy sources and nodal loads, respectively; Let be the line power flow coefficient vector; this formula requires that power balance be true simultaneously for all control coefficients.

[0406] The line power flow and reference node phase angle constraints are reconstructed as follows:

[0407]

[0408] In the formula, The vector of node phase angle coefficients; It is a K+1 dimensional column vector of all zeros;

[0409] For continuous operation response trajectories that are not directly affected by changes in the start-up and shutdown status of the generating units, the common endpoint of adjacent time segments satisfies the conditions shown in Equation (9). This ensures that the trajectories of renewable energy reduction, load reduction, line power flow, and node phase angle remain numerically continuous at the common endpoint of adjacent time segments.

[0410] (3) Reconstruction of inequality constraints

[0411] Continuous-time inequalities are reconstructed using control coefficient boundaries or derivative coefficient boundaries; simultaneously, the unit start-up and shutdown logic, minimum start-up and shutdown times, and trajectory connection conditions at start-up and shutdown boundaries are converted into finite-dimensional discrete constraints or large M constraints.

[0412] The upper and lower limits of unit output constraints are reconstructed as follows:

[0413]

[0414] The boundary constraints for renewable energy reduction and load reduction are reconstructed as follows:

[0415]

[0416] The line transmission capacity constraint is reconstructed as follows:

[0417]

[0418] To represent the derivative of a continuous trajectory, construct the first-order difference matrix of the Kth-order trajectory:

[0419]

[0420] In the formula, the left side is a first-order difference matrix, the number of rows of which is equal to the polynomial degree and the number of columns of which is equal to the number of control coefficients; each row contains only -1 and 1 in two adjacent columns, which are used to calculate the difference between adjacent control coefficients; the right side gives the real matrix space to which the matrix belongs.

[0421] Based on the first-order difference matrix, the unit's continuous ramping constraint is reconstructed as follows:

[0422]

[0423] In the formula, the first-order difference matrix forms the difference between adjacent control coefficients and applies the unit's uphill and downhill ramp limits to all derivative coefficients.

[0424] (4) Unit start-up and shutdown logic and minimum start-up and shutdown time constraints

[0425] The unit operating state and start-up / shutdown actions in step S2 are both finite-dimensional discrete variables, requiring no Bernstein transformation; their state transition relationships remain as follows:

[0426]

[0427] In the finite-dimensional model, the unit start-up and shutdown decisions are made using time segments of equal length, with each time segment having a length of 1. Therefore, the minimum continuous power-on time and minimum continuous power-off time expressed in physical time in step S2 can be equivalently converted into state summation constraints on the corresponding time segments:

[0428]

[0429] In the formula, Auxiliary segment index used for summation; The duration of the equal-length start / stop decision segments; This indicates rounding up to the nearest integer.

[0430] (5) Trajectory connection constraints at the start-up and shutdown boundaries of the unit

[0431] When the unit's start-up and shutdown states are taken into account, the unit's terminal output connection conditions are as follows:

[0432]

[0433] In the formula, when there is no start-stop transition, the output of the end points of adjacent units is forced to be equal. When a start-stop transition occurs, the end point connection is relaxed according to the maximum output of the unit.

[0434] The condition for the connection of the unit endpoint derivatives in adjacent time segments is:

[0435]

[0436] In the formula, the first derivatives of the endpoints of adjacent segments are equal when there is no start-stop transition, and the derivatives are relaxed and connected according to the climbing ability when a start-stop transition occurs.

[0437] The aforementioned objective function reconstruction, equality constraint reconstruction, and inequality constraint reconstruction together constitute a mixed-integer optimization model that can be processed by existing optimization solvers. Solving the model for each source-load scenario yields the optimal control coefficients for unit output, renewable energy reduction, load reduction, line power flow, and node phase angle trajectory.

[0438] Step S5: Reconstruct the continuous load reduction trajectory and calculate operational reliability indicators. Based on the optimal control coefficients for load reduction at each scenario node obtained in Step S4, reconstruct the continuous load reduction trajectory at each node and system level, and calculate the momentary load loss probability, time-domain average load loss probability, expected load loss duration, and expected power shortage value according to the definition in Step S3.

[0439] The optimal control coefficients for load reduction at each scenario node obtained in step S4 are denoted as follows:

[0440]

[0441] The continuous time trajectory of node load reduction is reconstructed as follows:

[0442]

[0443] Then, the load reduction control coefficients of each node are summarized to obtain the scenario. Next Vector of total system load reduction control coefficients within each time segment:

[0444]

[0445] Using the integral relationship from step S4, the scenario The power shortage in the next evaluation time domain can be directly calculated from the total system load reduction control factor:

[0446]

[0447] Based on the given scenario probabilities By applying probability weights to the power shortage amounts in each scenario, the expected value of insufficient power is obtained:

[0448]

[0449] According to the judgment threshold set in step S3 Define the scenario The following is the set of times when load shedding occurs:

[0450]

[0451] Within each time segment, the following equation can be solved: Real roots within the range:

[0452]

[0453] By sorting the real roots obtained from the above formula together with the interval endpoints 0 and 1, and determining whether the trajectory values ​​between adjacent roots are greater than the judgment threshold, the local time interval in which load reduction occurs within that time segment can be determined. This is achieved through physical time mapping: The local normalized time interval can be mapped to the actual load reduction time interval.

[0454] Based on the definition in step S3, the probability of load loss at any given time is calculated using the reconstructed total system load reduction trajectory:

[0455]

[0456] Integrating the probability of load loss at time point 1 over the entire evaluation time domain and taking the time average, we obtain the time-domain average probability of load loss:

[0457]

[0458] In addition to reliability metrics, the optimal operating costs for each scenario obtained in step S4 can be weighted by probability to obtain the expected operating cost:

[0459]

[0460] According to the definition in step S3, the cumulative unload duration for each scenario is probability-weighted to obtain the expected unload duration:

[0461]

[0462] Steps S1 to S5 sequentially complete the characterization of source load uncertainty and continuous trajectory, continuous-time operation response modeling, operation reliability index definition, finite-dimensional transformation and solution of operation response model, as well as continuous load reduction trajectory reconstruction and probability weighted statistics, thus forming a complete power system continuous-time operation reliability assessment process.

[0463] Example 13:

[0464] Based on the aforementioned technical solution, this embodiment sets the Bernstein polynomial degree to 3. Except for the newly added cubic Bernstein control coefficients, cubic trajectory difference matrix, and related cubic model parameters, the meanings of all other sets, indices, variables, parameters, and reliability indicators are consistent with the aforementioned technical solution.

[0465] This embodiment uses wind power as a new energy source. Therefore, the wind power output, wind power reduction, and wind farm mentioned below correspond to the new energy output, new energy reduction, and new energy power station in the aforementioned technical solution, respectively. The physical meaning, dimensions, and value range of the other symbols are consistent with those in the aforementioned technical solution.

[0466] Step S1: Construct the cubic Bernstein continuous-time source-charge trajectory.

[0467] The aforementioned technical solution uses the same scene set, scene probability, evaluation time domain, and time segmentation. For the h-th time segment, the segment length and locally normalized time are defined using the following formula:

[0468]

[0469] In the locally normalized time, the cubic Bernstein basis functions specialize as follows:

[0470]

[0471] In equation (56), .

[0472] The corresponding cubic Bernstein basis function column vector is:

[0473]

[0474] The continuous-time available wind power output trajectory of wind farm w in scenario s, and the continuous-time load demand trajectory of node n, are respectively represented as follows:

[0475]

[0476] Both four-dimensional control coefficient vectors in the formula are predetermined by the input scenario. To simultaneously match the left and right endpoint values ​​and the endpoint change rate of the segments, the following control coefficient relationship is constructed for any of the above source load trajectories:

[0477]

[0478] The rate of change at the endpoints can be given directly from external prediction results, or it can be obtained from data at adjacent prediction times using one-sided differencing or central differencing. The continuity conditions for trajectory values ​​and the continuity conditions for the first derivative at the common endpoints of adjacent time segments are as follows:

[0479]

[0480] This yields the continuous-time source-load trajectory expressed using piecewise cubic Bernstein polynomials. The wind power output control coefficients and nodal load control coefficients are known parameters for subsequent models and are not optimization decision variables.

[0481] Step S2: Establish a continuous-time running response model under three consecutive source load inputs.

[0482] The continuous-time wind farm output trajectory and nodal continuous-time load demand trajectory constructed in step S1 can be directly used as inputs for this step, with their cubic Bernstein control coefficients remaining known parameters. For any positive probability scenario, a single-scenario operational response model is established and solved, with the scenario probability not included in the single-scenario optimization objective.

[0483] The objective for a single scenario consists of discrete start-up and shutdown costs and continuous operation costs:

[0484]

[0485] In the formula, start-up cost and shutdown cost are outside the time integral and are calculated once for each discrete start-up and shutdown action; continuous power generation cost, wind power reduction cost and load reduction cost are within the time integral.

[0486] The unit's operating state remains unchanged across all time segments. The state transition relationships, mutual exclusion relationships, and variable value ranges are as follows:

[0487]

[0488] In the formula, Indicates the unit before the start of the evaluation time domain. The running status.

[0489] The minimum continuous start-up time constraints after unit startup and the minimum continuous shutdown time constraints after unit shutdown are as follows:

[0490]

[0491] The upper and lower limits of the unit's active power output, as well as the constraints on the uphill and downhill ramp rates at any time except for the unit's start-up and shutdown boundaries, are as follows:

[0492]

[0493] The upper bounds for wind power reduction and load reduction are derived from the known continuous wind power trajectory and continuous load trajectory given in step S1, respectively:

[0494]

[0495] The node power balance, line DC power flow relationship, line transmission capacity limitation, and reference node phase angle constraint are as follows:

[0496]

[0497] This model obtains the continuous-time operational response trajectories of unit output, wind power reduction, load reduction, line power flow, and node phase angle under various scenarios.

[0498] Step S3: Define the operational reliability index based on the continuous load reduction trajectory.

[0499] This step only defines operational reliability metrics and does not perform numerical calculations. The system total load reduction power, time-domain load failure probability, and time-domain average load failure probability are defined as follows:

[0500]

[0501] The cumulative load loss duration and expected load loss duration for a single scenario are defined as follows:

[0502]

[0503] The expected values ​​for power shortage and insufficient power are defined as follows:

[0504]

[0505] The probabilities of the above scenarios are given in step S1. The indicative function takes the value 1 when the condition within the parentheses is met, and 0 otherwise. The load reduction threshold can theoretically be 0. However, in numerical calculations, to exclude the minimal non-zero load reduction caused by the feasibility tolerance of the optimization solver and floating-point operation errors, this embodiment takes... MW. The specific value of the indicator is calculated after solving in step S4 and reconstructing the continuous load reduction trajectory in step S5.

[0506] Step S4: Convert the continuous-time response model into a cubic Bernstein finite-dimensional mixed integer optimization model.

[0507] The unit output, wind power reduction, load reduction, line power flow, and node phase angle are each represented by a four-dimensional control coefficient vector:

[0508]

[0509] The control coefficients of the operating response variable in the formula are the optimization decision variables for this step; the control coefficients of available wind power output and nodal load in step S1 are still known parameters. The integral and first derivative of the cubic trajectory are related as follows:

[0510]

[0511] The relationships between the four-dimensional all-1 vector, the four-dimensional zero vector, the three-row, four-column first-order difference matrix, the constant-zero trajectory, the boundary of the control coefficient, and the boundary of the continuous trajectory interval are as follows:

[0512]

[0513] Using the exact integral relationship of the formula, the single-scene finite-dimensional objective function can be written as:

[0514]

[0515] The formula is only a finite-dimensional objective for scenario s; startup and shutdown costs are still calculated as discrete actions outside the integral term.

[0516] The equation for the control coefficients of node power balance is:

[0517]

[0518] The line power flow definition, the phase angle of the reference node, and the numerical continuity condition of the common endpoints of trajectories not directly affected by start-stop are as follows:

[0519]

[0520] The continuous-time inequality is transformed into the following control coefficient boundary and derivative control coefficient boundary:

[0521]

[0522] In the formula, It is a three-dimensional column vector of all 1s; Let be the first-order difference matrix corresponding to the three trajectories.

[0523] The segmented start-up and stop state transitions, minimum start-up time, and minimum stop-up time are maintained in the finite-dimensional model as follows:

[0524]

[0525]

[0526] In this embodiment Furthermore, the minimum start-up time and minimum shutdown time of the unit are both integer hours, therefore the number of the above segments is respectively equal to and .

[0527] At the common boundary of adjacent time segments, the start and stop actions of the next segment are used to characterize the boundary start-stop events; when there is no start-stop transition, the output of the endpoints and the first derivative of the endpoints are forced to be continuous; when a start-stop transition occurs, the big M constraint is relaxed.

[0528]

[0529] The aforementioned single-scenario finite-dimensional objective function, equality constraints, inequality constraints, start-stop logic, and trajectory connection constraints at the start-stop boundary collectively constitute the cubic Bernstein finite-dimensional mixed-integer optimization model corresponding to scenario s. Solving the model for each positive probability source-load scenario yields the optimal control coefficients for the unit's operating state, start-stop actions, and each continuous operating response trajectory.

[0530] Step S5: Reconstruct the continuous load reduction trajectory and calculate the operational reliability index.

[0531] First, the continuous trajectory of optimal load reduction at the node level is reconstructed; then, the optimal control coefficients of each node are summarized to obtain the vector of total system load reduction control coefficients. Using the four system load reduction control coefficients within each time segment, the power shortage in a single scenario is accurately calculated, and the expected value of the power shortage is obtained by weighting the values ​​according to scenario probabilities.

[0532]

[0533] For each scenario and time segment, convex hull prediction is first performed using four control coefficients for the total system load reduction: if all four control coefficients do not exceed the judgment threshold, there is no power shortage in that segment; if all four control coefficients exceed the judgment threshold, power shortage occurs in that segment throughout the entire time period. For segments that cannot be directly determined, the intersection points of cubic trajectories and the judgment threshold are calculated, and physical-time mapping and probability statistics are completed.

[0534]

[0535] In the formula, This represents the interval length of the corresponding time set. For time segments where the power shortage state cannot be directly determined through the above convex hull prediction, Bernstein subdivision and root isolation can be used, or it can be converted into a power-basis cubic polynomial for root finding. All real roots within the locally normalized interval are retained, and the real roots are sorted together with the interval endpoints 0 and 1, and sign judgment is performed in adjacent intervals. Repeated roots, contact thresholds, and multiple power shortage intervals within the same segment are processed separately and then mapped back to physical time.

[0536] As defined in step S3, the reconstructed trajectory calculates the duration of insufficient power supply, the moment-to-moment load failure probability curve, the time-domain average load failure probability, the expected duration of load failure, and the expected value of insufficient power for each scenario. The expected operating cost is an economic statistical result and not an operational reliability indicator. This embodiment ultimately outputs the continuous trajectory of load reduction at each scenario node, the continuous trajectory of total system load reduction for each scenario, the time interval of insufficient power supply for each scenario, the cumulative duration of insufficient power supply for each scenario, the moment-to-moment load failure probability curve, the time-domain average load failure probability, the expected duration of load failure (LOLE), the expected value of insufficient power (EENS), the amount of power shortage (ENS) for each scenario, and the expected operating cost.

[0537] Simulation Results and Analysis

[0538] To verify the ability of this invention to identify the risk of insufficient power supply within the scheduling interval, simulations were conducted using a modified IEEE 30-node system. The evaluation time domain in this embodiment is 24 hours. System topology as follows Figure 1 As shown. The system consists of 6 conventional turbines with a base load of 335 MW; two wind farms are connected to nodes 17 and 24 respectively, each with an installed capacity of 72 MW, for a total installed capacity of 144 MW; the ramp-up capability of the conventional turbines is set to 0.40 pu / h of their maximum output; the continuous trajectory is represented by a cubic Bernstein polynomial.

[0539] The simulation selects 10 complete source-load scenarios obtained through scenario reduction, and uses the probability corresponding to each reduced scenario as a weight, with the sum of scenario probabilities being 1. The discrete-time model (DT) and the continuous-time model (CT) of this invention use the same source-load scenarios, scenario probabilities, equipment parameters, network constraints, cost coefficients, and load reduction penalty coefficients to eliminate the interference of factors other than the time representation method on the comparison results. DT uses a piecewise constant form to describe the source-load input and operational response; CT uses a continuous trajectory to represent the evolution within the segments and checks the operational constraints throughout the entire time segment.

[0540] To compare the two time representation methods, a representative scenario 1 is selected, whose wind power output and net load input are as follows: Figure 2 As shown.

[0541] Figure 2 This indicates that DT represents source-load changes as segmented discrete quantities, while CT retains the peaks and valleys, rates of change, and local transitions within each segment, especially capable of characterizing rapid wind power changes and local peak net load around 12 hours. Therefore, steep changes in source-load within a segment can be incorporated into subsequent operational response calculations, providing a more complete input for identifying short-term power supply risks caused by insufficient unit ramping capability.

[0542]

[0543] Table 1 shows that the expected energy deficit (EENS) of the CT model is 8.822 MWh, an increase of 0.618 MWh compared to 8.204 MWh of the DT model, representing an increase of approximately 7.53%. The time-domain average load shedding probability increased from 0.155 to 0.240, and the expected load shedding duration (LOLE) increased from 3.724 h to 5.755 h. These results indicate that some source-load scenarios only experience power shortages within the scheduling segment and are not fully accounted for by the DT model.

[0544] The expected operating cost of the CT model is RMB 283,731.47, an increase of approximately 8.30% compared to the RMB 261,993.28 of the DT model. This cost difference stems from the CT model's continuous verification of segmented internal power balance, unit ramping, and network operation constraints. The results show that the continuous-time model can identify segmented power shortage risks that the discrete-time model fails to fully capture, and correspondingly increases the operating cost required to meet continuous operation constraints. The average solution time for a single scenario of the CT model is 14.03 s, indicating that the finite-dimensional mixed-integer model obtained through Bernstein solution space transformation can be effectively processed by existing optimization solvers.

[0545] To compare the two models' results in identifying the timing of risk occurrence, Figure 3 The probability of load loss at any time and the difference between them are given within the 24-hour evaluation time domain.

[0546] Figure 3 This indicates that the CT model identifies non-zero load loss probabilities that the DT model does not reflect within multiple scheduling segments. In this embodiment, the entire time interval where the load loss probability of the CT model is greater than the numerical tolerance, while the load loss probability of the DT model is less than the numerical tolerance, is defined as the risk period identified by CT but missed by DT. Its cumulative duration is the sum of the lengths of the above time intervals. In this embodiment, the cumulative duration of the risk identified by CT but missed by DT is 3.493 h; the maximum positive probability difference between the two models is 0.568, and the maximum load loss probabilities of the CT model and the DT model are 1.000 and 0.776, respectively. Figure 3 (b) Further, it is shown that the temporal representation method changes the distribution of risk in the assessment time domain. Although the failure probability of the CT model is lower than that of the DT model at a few moments, its temporal average is still 0.085 higher, indicating that the risk identified by CT but missed by DT is dominant overall.

[0547] To analyze the formation mechanism of the above probability differences, Figure 4 The load reduction trajectory and two local time windows representing scenario 1 are given.

[0548] In representative scenario 1, the power shortage, load shedding duration, and maximum load reduction power corresponding to the DT model are 18.079 MWh, 6.000 h, and 6.677 MW, respectively; while the corresponding indices for the CT model are 19.478 MWh, 7.485 h, and 13.802 MW, respectively. Locally magnified results show that the CT model can capture short-term load reduction peaks around 12 h and 17 h, some of which are underestimated or completely omitted in the DT results.

[0549] In summary, under the condition that the source-load scenario, system parameters, and network constraints remain consistent, this invention can reconstruct the source-load changes and load reduction trajectories within the scheduling segment, and quantify the short-term power shortage risk that the DT model cannot fully reflect from three dimensions: occurrence probability, duration, and power shortage amount. Simulation results verify the effectiveness of this invention in identifying implicit operational risks within the scheduling interval, as well as the solvability of the finite-dimensional model.

Claims

1. A method for continuous-time operation reliability assessment of power systems considering time-varying trajectory uncertainty of new energy, characterized in that, Includes the following steps: Step 1) Obtain several complete source-load time-series scenes with probabilities in the time domain for evaluation, and convert the complete source-load time-series scenes into continuous-time source-load trajectories; Step 2) Establish continuous-time response models for each source-load time-series scenario; Step 3) Define operational reliability metrics based on continuous load reduction trajectories for each scenario; Step 4) representing the continuous-time operating response model corresponding to each source-load time sequence scenario as secondary Bernstein basis functions, and converting into a finite-dimensional mixed-integer optimization model; solving the finite-dimensional mixed-integer optimization model to obtain optimal control coefficients of the power system operating characteristic parameters; Step 5) Based on the optimal control coefficients of the power system operating characteristic parameters, reconstruct the continuous load reduction trajectory and calculate the evaluation results after reconstructing the continuous load reduction trajectory.

2. The method for power system continuous-time operation reliability evaluation considering new energy time-varying trajectory uncertainty according to claim 1, characterized in that, In step 1), the steps for constructing the continuous-time source load trajectory are as follows: Step 1.1) record the source load scene set as , the new energy station set as , the node set as , the indexes of the scene, the new energy station and the node are respectively and ; to the scene corresponding probability normalization is performed as follows: (1) Step 1.2) divide the evaluation time domain into adjacent time segments, the first time segment being as follows:​ (2) For any instant of time The physical time is mapped to a local normalized time interval as follows: (3) wherein is the segment length, is the local normalized time variable; The segment lengths are shown below: (4) Step 1.3) Defining within the local normalized time interval The next Bernstein basis function is defined as follows: (5) (6) In the formulae, ; Step 1.4) Define the Bernstein basis function column vector as follows: (7) Step 1.5) Constructing the scenario The continuous time trajectory of the new energy output in the first time segment is as follows: (8) wherein is the Bernstein control coefficient vector for the new energy field station ​ Constructing a scenario The continuous time trajectory of the load of the intermediate node in the first time segment is as follows: (9) In the formula, is the load control coefficient vector of the node .

3. The method for power system continuous-time operation reliability evaluation considering new energy time-varying trajectory uncertainty according to claim 2, characterized in that, In step 2), the objective function of the continuous-time response model corresponding to each source-load time-series scenario is as follows: (10) In the formula, , and These are respectively a set of conventional generating units, a set of new energy power stations, and a set of nodes; For the scene The operating costs within the next evaluation time domain; , The units In the Start-up and shutdown action variables at the initial boundary of each time segment; , These are the continuous active power output of generating units, the power reduction of new energy sources, and the power reduction of nodal loads; , These are the unit start-up cost coefficient, shutdown cost coefficient, continuous power generation cost coefficient, renewable energy reduction cost coefficient, and load reduction cost coefficient, respectively. Number of time segments; The constraints of the continuous-time operation response model for each source-load scenario include unit operating status and start-up / shutdown action constraints, minimum continuous start-up time constraints that should be met after unit start-up, minimum continuous shutdown time constraints that should be met after unit shutdown, the upper and lower limits of the unit's continuous active power output under the operating status constraints, the upward / downward ramp rate constraints that should be met at any time for the unit's continuous active power output trajectory except for the unit start-up / shutdown boundaries, constraints on the power reduction of new energy sources and the power reduction of nodal loads, power balance constraints, and line power flow / line transmission capacity and reference node phase angle constraints.

4. The method for power system continuous-time operation reliability evaluation considering new energy time-varying trajectory uncertainty according to claim 3, characterized in that, The unit operating status and start-up / shutdown constraints are as follows: (11) (12) (13) (14) In the formula, , The units The operational status within adjacent time segments; To evaluate the unit before the start of the time domain The running status; The minimum continuous start-up time constraints that the unit must meet after startup are as follows: (15) The minimum continuous downtime constraints that the unit must meet after shutdown are as follows: (16) wherein , are the minimum continuous on-time and minimum continuous off-time of the unit respectively; ξ is an auxiliary time variable in the integral. The continuous active power output of the unit should meet the upper and lower limits of the output under the constraints of the operating conditions, as shown below: (17) wherein respectively the unit minimum and maximum allowed output power Except for the unit start-up and shutdown boundaries, the continuous active power output trajectory of the unit should meet the up / down ramp rate constraints at any time as shown below: (18) In the formula, , The units Maximum uphill and maximum downhill speeds; The power reduction constraints for renewable energy sources and nodal loads are as follows: (19) (20) In the formula, , These are power reductions from new energy sources and power reductions from the load; The power balance constraints are as follows: (21) In the formula, and Access nodes A collection of conventional generating units and a collection of new energy power stations; For nodes The set of connected adjacent nodes; The line power flow / line transmission capacity and reference node phase angle constraints are shown below: (22) (23) (24) wherein is the susceptance; are the continuous-time phase angle trajectories of nodes and node respectively; is the transmission capacity of line ; and is the reference node.

5. The method for power system continuous-time operation reliability evaluation considering new energy time-varying trajectory uncertainty according to claim 4, characterized in that, In step 3), the operational reliability indicators include the probability of load failure at any given time, the time-domain average probability of load failure, the expected duration of load failure, and the expected value of insufficient power. The probability of load loss at any given time is shown below: (25) In the formula, Reduce power consumption to reduce the overall system load; It is an indicator function; This is a non-negative decision threshold used to determine whether load reduction has occurred in the system; The time-domain average failure probability is shown below: (26) In the formula, T is the total duration of the time domain under evaluation ​ The expected duration of the loss of load is as follows: (27) In the formula, is the cumulative off-load duration; The expected value for low battery is as follows: (28) In the formula, is the lack of power supply for this scenario.

6. The method for power system continuous-time operation reliability evaluation considering new energy time-varying trajectory uncertainty according to claim 5, characterized in that, The total system load reduction power is shown below: (29) The cumulative unload duration is shown below: (30) In the formula, Reduce power consumption to reduce the total system load.

7. The method for power system continuous-time operation reliability evaluation considering new energy time-varying trajectory uncertainty according to claim 5, characterized in that, In step 4), the objective function of the finite-dimensional mixed-integer optimization model is as follows: (31) In the formula, It is a column vector of all 1s whose dimension is equal to the exponent plus 1; For the scene The corresponding finite-dimensional operating cost; and These are the Bernstein coefficient vectors corresponding to unit output, renewable energy reduction, and load reduction, respectively. The variables are for segmented start-up and shutdown; the equality constraints of the finite-dimensional mixed integer optimization model include node power balance constraints and line power flow and reference node phase angle constraints. The inequality constraints of the finite-dimensional mixed-integer optimization model include upper and lower limits of unit output, boundary constraints of renewable energy reduction and load reduction, line transmission capacity constraints, and unit continuous ramping constraints. The unit start-up and shutdown logic and minimum start-up and shutdown time constraints of the finite-dimensional mixed-integer optimization model are equivalently transformed into state summation constraints on the corresponding time segments, as shown below: (32) wherein is the auxiliary segment index used for summation; is the duration of the equal length start-stop decision segment; denotes the ceiling function; The trajectory connection constraints at the unit start-up and shutdown boundaries of the finite-dimensional mixed-integer optimization model are shown below: (33) In the formula, , The units Maximum uphill speed and maximum downhill speed.

8. The method for assessing the continuous-time operational reliability of a power system considering the uncertainty of time-varying trajectories of new energy sources, as described in claim 7, is characterized in that... The node power balance constraints are as follows: (34) In the formula, and These are known coefficient vectors for available renewable energy sources and nodal loads, respectively. This represents the vector of line power flow coefficients; The line power flow and reference node phase angle constraints are shown below: (35) In the formula, The vector of node phase angle coefficients; It is a K+1 dimensional column vector of all zeros; The upper and lower limits of the unit's output are constrained as follows: (36) The boundary constraints for renewable energy reduction and load reduction are as follows: (37) The line transmission capacity constraints are as follows: (38) The continuous ramp-up constraints for the unit are as follows: (39) (40) In the formula, It is a K-dimensional column vector of all 1s; Let be the first-order difference matrix of the Kth trajectory.

9. The method for assessing the continuous-time operational reliability of a power system considering the uncertainty of time-varying trajectories of new energy sources, as described in claim 1, is characterized in that... Power system operating characteristic parameters include unit output, renewable energy reduction, load reduction, line power flow, and node phase angle trajectory.

10. The method for assessing the continuous-time operational reliability of a power system considering the uncertainty of time-varying trajectories of new energy sources, as described in claim 8, is characterized in that... In step 5), the scene The following is the set of times when load shedding occurs: (41) The evaluation results after reconstructing the continuous load reduction trajectory include operational reliability indicators and expected operating costs; among them, the operational reliability indicators include the expected value of insufficient power, the probability of load loss at any time, the time-domain average probability of load loss, and the expected duration of load loss, and the expected operating costs are economic statistics results; The expected value for low battery is as follows: (42) In the formula, For the scene The power shortage within the next evaluation time domain; Scene The power shortage in the next evaluation time domain is shown below: (43) In the formula, This is the total load reduction control factor for the system; The total system load reduction control coefficients are shown below: (44) (45) (46) In the formula, The continuous time trajectory for reducing node load during reconstruction; The total system load reduction for the reconstructed system is achieved by cutting the continuous time trajectory. Optimal control coefficients for load reduction at nodes in each scenario; A column vector of K-th degree Bernstein basis functions; The probability of load loss at any given time is shown below: (47) The time-domain average failure probability is shown below: (48) The expected operating costs are as follows: (49) In the formula, For the scene The optimal objective function value of the corresponding finite-dimensional mixed-integer optimization model; The expected duration of the loss of load is as follows: (50) In the formula, This represents the time-domain average probability of load failure.