Annual power generation maintenance plan arrangement method for high-proportion new energy power system

Abstract

The invention belongs to the technical field of intelligent scheduling of a power system, and relates to a high-proportion new energy power system annual power generation maintenance plan arrangement method, which comprises the steps of 1, data collection, cleaning and preparation; 2, taking an augmented Lagrangian penalty function for minimizing a maintenance plan connection variable as a target to construct and form a maintenance coordination problem cluster; establishing and forming a maintenance formulating problem cluster by taking minimization of a maintenance plan connection variable as a target; alternately solving the two clusters in parallel, and increasing the Canagrangian penalty term coefficient and the connection variable value according to an ATC method until the maintenance connection variable value obtained by the two problem clusters meets a convergence condition; according to the method, the system maintenance requirements under various operation conditions can be comprehensively considered, and the calculation cost can be effectively reduced, so that the maintenance plan can be made more flexibly and accurately, and the stable operation of the power system is ensured.

Description

Technical Field

[0001] This invention belongs to the field of intelligent dispatching technology for power systems, specifically relating to a method for arranging annual power generation and maintenance plans for a high-proportion renewable energy power system. Background Technology

[0002] Annual maintenance schedules are a crucial boundary for further balancing analysis, determining the upper limits of the power system's dispatch capacity and regulation capabilities at various times throughout the year, and significantly impacting the power system's absorption and supply capacity. In current new power systems, the proportion of renewable energy generation capacity is continuously increasing, and the inaccuracy of wind and solar power generation forecasts has an increasingly significant impact on the scheduling of power generation maintenance plans: traditional methods, based on a few specific typical scenarios, often fail to cope with the ever-changing actual operating scenarios. Meanwhile, the uncertainty of current renewable energy output is often modeled using probabilistic models, and many practical models and methods already exist.

[0003] However, due to limitations in computational complexity, current algorithmic foundations, and computer hardware, probabilistic maintenance scheduling optimization models for large-scale real-world systems are still immature. Their focus is on simultaneously considering load supply demand, renewable energy regulation demand, and computational complexity to quickly obtain a set of highly generalizable maintenance schedules. This results in existing technologies exhibiting insufficient generality and high computational costs in providing annual generator unit maintenance schedules. Summary of the Invention

[0004] This invention aims to provide a method for rapidly obtaining a set of feasible, universal, and robust annual power system maintenance plans while considering numerous new energy power generation scenarios. This addresses the problems of insufficient universality and high computational costs in existing technologies for annual generator maintenance planning. By introducing stochastic optimization (SO) methods and objective cascade analysis (ATC) techniques, the method can more comprehensively consider the system maintenance space under various operating conditions, significantly improving the universality and reliability of maintenance plans. This provides strong support for the load supply tasks and new energy consumption targets of new power systems.

[0005] This invention provides the following technical solution: a method for scheduling annual power generation and maintenance plans for a high-proportion renewable energy power system, comprising the following steps: Step 1: Data collection, cleaning and preparation; acquire historical power system operation data, preprocess the data to generate diverse power system renewable energy output scenarios.

[0006] Step 2: Establish and solve the maintenance plan optimization model; with the objective of minimizing the augmented Lagrange penalty function of the maintenance plan connection variables, construct a 0-1 maintenance plan coordination problem for each maintenance requirement, forming a maintenance coordination problem cluster; with the objective of minimizing the augmented Lagrange penalty function of the maintenance plan connection variables, renewable energy curtailment, and unit congestion costs, and maximizing the system's power supply capacity and absorption capacity, construct a maintenance plan formulation sub-model for each renewable energy scenario, forming a maintenance formulation problem cluster; alternately and in parallel solve the maintenance coordination problem cluster and the maintenance formulation problem cluster, and update the augmented Lagrange penalty term coefficients and connection variable values ​​according to the ATC method until the maintenance connection variable values ​​obtained from the two problem clusters satisfy the convergence condition.

[0007] Preferably, in step 1, the historical power system operation data includes: historical weather data, wind power output curves, photovoltaic power output curves, power system grid parameters, generator parameters, load demand, and generator maintenance demand data; preprocessing includes: data cleaning, noise reduction, normalization, and verification and error correction of power system parameters.

[0008] Preferably, in step 2, the augmented Lagrange penalty function of the maintenance plan connection variable is:

[0009] in, For maintenance plan i of d Daily state variables; For maintenance plan i of d Daily maintenance status communication variables; For the connection variable The Lagrange penalty coefficient; For the connection variable The augmented Lagrange penalty coefficient; Allow start / end time for maintenance plan i; A collection of generator set maintenance plans; A collection of new energy scenarios; For daily indexing.

[0010] More preferably, the constraints of the augmented Lagrange penalty function of the maintenance plan connection variables include: the start time and duration constraints of the maintenance requirements.

[0011] Preferably, in step 2, the sub-model for constructing and formulating maintenance plans for new energy scenarios is as follows:

[0012]

[0013]

[0014]

[0015]

[0016]

[0017]

[0018]

[0019]

[0020] in, To augment the Lagrange penalty function, Penalties for curtailment of renewable energy sources For generator set capacity blockage penalty, To achieve the goal of surplus power supply capacity around the clock, For the minimum power supply capacity bus surplus target, For the minimum power supply capacity period surplus target, To achieve the target of surplus capacity for new energy consumption, For the minimum absorption capacity busbar profit target, The profit target is set for the period with the lowest absorption capacity. For maintenance plan i of d The 0-1 status on the day, from Given, it is a constant; For maintenance plan i of d The daily maintenance status is a communication variable, with a value of 0 to 1, which is a continuous quantity. For the connection variable The Lagrange penalty coefficient; For the connection variable The augmented Lagrange penalty coefficient; For the scene s In the context of renewable energy units, the amount of power curtailed during peak / valley load periods is denoted as ; For the scene s Lower busbar k exist d The daily net load peak hour power supply capacity is in surplus; For the scene s Dispatchable units i exist d Daily net load peak time blocking capacity; For the scene s New energy units i exist dThe daily net load trough can absorb the space surplus; busbar k The next batch of new energy generating units; A collection of new energy generating units for the power system; A set of dispatchable generating units in a power system; This refers to the collection of busbars in a power system.

[0021] More preferably, the constraints of the sub-model for constructing maintenance plans in the new energy scenario include: coupling constraints between availability and maintenance, constraints on fixed start-up mode of generating units, logical constraints on start-up mode, power constraints on peak net load, power constraints on valley net load, power constraints on peak load verification, and constraints on dispatchable power plants.

[0022] More preferably, the constraints of the sub-model for constructing maintenance plans for the new energy scenario include constraints on new energy power plants: net load peak-time curtailment constraints, net load valley-time curtailment constraints, and net load valley-time absorption constraints.

[0023] More preferably, the constraints of the sub-model for constructing maintenance plans for the new energy scenario include generator set maintenance constraints, maintenance start time constraints, maintenance duration constraints, power system bus constraints, net load peak-time power supply capacity surplus constraints, net load peak-time standby constraints, net load peak-time branch power flow constraints, net load valley-time branch power flow constraints, net load peak-time profit and loss balance constraints, net load valley-time power balance constraints, and start-up mode verification constraints.

[0024] Preferably, in step 2, the alternating parallel solution specifically includes: First, parallel solution of the unaugmented Lagrange penalty term. Maintenance schedule arrangement problem cluster Obtain the initial values ​​of the maintenance status communication variables. .

[0025] Subsequently, the maintenance planning coordination problem cluster was solved in parallel. Obtain new values ​​for maintenance status communication variables. Parallel solution considering augmented Lagrange penalty terms Maintenance schedule arrangement problem cluster Obtain new values ​​for maintenance status communication variables. Update maintenance plan coordination issues cluster The coefficients of the augmented Lagrange penalty function.

[0026]

[0027] in, k This represents the number of iterations. To augment the Lagrange penalty increment coefficient.

[0028] Solve the maintenance planning coordination problem cluster in parallel again. Obtain new values ​​for maintenance status communication variables. Calculate the numerical convergence criterion or the 0-1 convergence criterion. If the criterion is met, the iteration ends; otherwise, the maintenance plan scheduling problem cluster is updated. The coefficients of the augmented Lagrange penalty function:

[0029] Repeat the above alternating solution process until the algorithm converges to obtain the annual generator set maintenance plan.

[0030] The beneficial effects of this invention are: 1. This invention significantly improves the versatility and computational efficiency of annual power generation maintenance plans by introducing stochastic optimization (SO) methods and objective cascade analysis (ATC) techniques. This invention not only comprehensively considers system maintenance needs under various operating conditions but also effectively reduces computational costs. Compared with existing technologies, this method allows for more flexible and accurate maintenance planning, ensuring the stable operation of the power system. Furthermore, by preprocessing historical data and generating scenarios, this invention improves the quality and consistency of input data, providing more accurate data support for maintenance plans. This enables the system to maximize its power supply capacity and renewable energy absorption capacity while minimizing maintenance costs, thereby enhancing the reliability and flexibility of the entire power system.

[0031] 2. This invention can significantly enhance the ability of new power systems to cope with uncertainties from a planning and operational perspective. First, it improves the reliability of the power system by reducing sudden power outages caused by equipment failures through more scientific and reasonable maintenance arrangements, thus ensuring the security of power supply. Second, it enhances the system's ability to adapt to fluctuations in renewable energy sources by optimizing maintenance plans, enabling renewable energy generating units to operate at their optimal state, thereby promoting the effective utilization and consumption of renewable energy. Third, this invention helps reduce the operating costs of the power system by refining maintenance time and cycles, avoiding unnecessary downtime losses and maintenance costs. Finally, the optimized maintenance plan based on this invention can also provide important reference for the future expansion and upgrading of the power system, helping decision-makers better plan the allocation and development direction of power resources. Therefore, this invention has significant practical implications for achieving intelligent and efficient management of power systems. Attached Figure Description

[0032] Figure 1 This invention presents a comparison chart of the gap between the iterative solution and the integrated direct solution of an annual power generation maintenance plan arrangement method for a high-proportion new energy power system. Figure 2This is a comparison chart of the time required for the iterative solution and the integrated direct solution of this invention; Figure 3 This is a comparison chart of the number of iterations for different convergence criteria of the present invention; Figure 4 The diagram shows the iterative convergence process of the algorithm under different scenarios of this invention; Figure 5 This is a comparison chart of the results obtained by the iterative method and the integrated direct solution of the present invention; Figure 6 This is a diagram illustrating the iterative convergence process of a provincial-level practical system example of the present invention. Detailed Implementation

[0033] The related technologies of the present invention will be clearly and completely described below with reference to the accompanying drawings of the embodiments of the present invention. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those of ordinary skill in the art without creative effort are within the scope of protection of the present invention.

[0034] like Figures 1-6 As shown, the annual power generation maintenance plan arrangement method for a high-proportion renewable energy power system in this embodiment specifically includes the following steps: Step 1: Data Collection, Cleaning, and Preparation. Relevant data is obtained from relevant departments, including historical weather data, wind power output curves, photovoltaic power output curves, power system grid parameters, generator parameters, load demand, and generator maintenance demand data. Historical power system operation data is preprocessed. This step includes scene data cleaning, noise reduction, and normalization, as well as power system parameter verification and error correction to ensure the quality and consistency of the input data. Based on the preprocessed data, diverse new energy power output scenarios for the power system are generated.

[0035] Step 2: Establish and solve the maintenance plan optimization model. With the objective of minimizing the augmented Lagrange penalty function of the maintenance plan's connection variables, considering constraints such as 0-1 variable constraints, maintenance start time, and duration, a 0-1 maintenance plan coordination problem is constructed for each maintenance requirement, forming a maintenance coordination problem cluster. With the objective of minimizing the augmented Lagrange penalty function of the maintenance plan's connection variables, renewable energy curtailment, and unit congestion costs, and maximizing the system's power supply capacity and absorption capacity, constraints such as adjustable generating units, power plants, renewable energy units, buses, system power, and power flow are considered. The 0-1 variable constraints are relaxed, and a maintenance plan formulation sub-model is constructed for each renewable energy scenario, forming a maintenance formulation problem cluster. The two problem clusters are solved alternately and in parallel, and the augmented Lagrange penalty term coefficients and connection variable values ​​are updated according to the ATC method until the maintenance connection variable values ​​obtained from the two problem clusters satisfy the convergence condition.

[0036] The model of this invention obtains relevant data from relevant departments, including historical weather data, wind power output curves, photovoltaic power output curves, power system grid parameters, generator parameters, load demand, and generator maintenance demand data.

[0037] After obtaining the above information, the following steps are taken to implement the probabilistic annual power generation maintenance plan optimization and acceleration method based on the improved ATC algorithm proposed in this invention.

[0038] Step 2-1: Construct a maintenance coordination problem cluster. With the objective of minimizing the augmented Lagrange penalty function of the maintenance plan connection variables, and considering constraints such as 0-1 variable constraints, maintenance start time, and duration, a 0-1 maintenance plan coordination problem is constructed for each maintenance requirement, forming a maintenance coordination problem cluster.

[0039] The 0-1 maintenance plan coordination problem proposes a unified maintenance scheme that satisfies physical and temporal constraints, based on the maintenance plans provided by all sub-problems in the scenarios. This is based on the fact that the annual generator maintenance plan scheduling model still faces problems such as large computational load and excessively long solution time when the number of scenarios is large. Therefore, an acceleration method is considered to improve the solution efficiency for smaller problems while avoiding unsolvable or unacceptably time-consuming problems for larger problems.

[0040] The original model for optimizing the probabilistic annual generator set maintenance plan can be expressed as:

[0041] Its essence is a mixed integer programming problem (MILP), therefore slack variables are introduced. Rewrite it in matrix form as a standard linear programming problem:

[0042] in, x For all decision variables, M This is the constraint matrix. In the above model, since the various scenarios are only connected through maintenance state variables, the coupling matrix blocks between scenarios are all 0. M It is an extremely sparse matrix. Through row and column transformations, and by changing the matrix... M Based on the scenario, the blocks are as follows:

[0043] in, Constraints on the maintenance plan for dispatchable units. For the scene s The coupling constraints between the operating status and maintenance plan of the dispatchable units. For the scene s Internal constraints. Except for the blocks mentioned above, all other parts are equal to 0; and simultaneously, It is also a highly sparse matrix, which has only one non-zero element in the row containing the scheduleable unit maintenance plan constraint, and all other elements are 0.

[0044] Based on the above characteristics, the Objective Cascade Analysis (ATC) method is used to decompose the complete model into a 0-1 maintenance plan model. 、 Scene sub-model .

[0045] At the same time, due to There is no coupling relationship between the different maintenance variables in the problem in terms of objective function and constraints. Therefore, the above large 0-1 maintenance problem can be decomposed into several 0-1 maintenance subproblems according to maintenance requirements. Each maintenance requirement establishes an independent Integer Quadratic Programming (MIQP) coordination problem. While reducing the problem size, each maintenance subproblem can also be solved in parallel, further improving computational efficiency.

[0046] The above The problem aims to define the augmented Lagrange penalty function of the maintenance plan connection variables:

[0047] in, For maintenance plan i of d Daily status variable, 0 means not under maintenance, 1 means under maintenance; for The maintenance plan provided by the model i of d The daily maintenance status communication variable is a continuous value between 0 and 1. The value in the middle is constant; For the connection variable The Lagrange penalty coefficient; For the connection variable The augmented Lagrange penalty coefficient; Allow start / end time for maintenance plan i; A collection of generator set maintenance plans; A collection of new energy scenarios; For daily indexing.

[0048] At the same time, the start time and duration constraints of maintenance needs should be considered.

[0049]

[0050] in, The variable is 0-1, representing the start of maintenance plan i. The duration of maintenance plan i; The duration of the maintenance coordination model.

[0051] Step 2-2: Construct a maintenance planning problem cluster. With the goal of minimizing the augmented Lagrange penalty function of maintenance plan connection variables, renewable energy curtailment, and unit congestion costs, and maximizing system power supply and absorption capacity, constraints such as adjustable generating units, power plants, renewable energy units, buses, system power, and power flow are considered. 0-1 variable constraints are relaxed, and a maintenance planning sub-model is constructed for each renewable energy scenario, forming a maintenance planning problem cluster.

[0052] Precise annual power generation and maintenance scheduling model for single-scenario dispatchable units Generally, it can be modeled as follows:

[0053] in, For continuous variables, For 0-1 variables, F To optimize the objective function. Specifically, integer variables. x It can be further broken down into maintenance variables. and unit state variables Further consideration of scene numbering This can be expressed as:

[0054] After considering the augmented Lagrange penalty, although the single-scenario maintenance planning model... Medium maintenance variables Even though the problem has been relaxed to a continuous variable, it is still a MIQP problem. When the system is large and there are many scenarios, it is usually difficult to solve or the time cost is high.

[0055] To address the aforementioned problems, this invention introduces statistics from a probabilistic analysis perspective: Scenario s Lower unit i exist d Daily work location expectations Considering the typical multi-scenario issues of unit combinations, strictly speaking, the unit... i The mathematical expectation of the working position is: unit under multiple scenarios i power The average value is calculated, but it is usually difficult to obtain due to the large amount of computation. Therefore, this invention uses the unit state variables. Relaxation as a continuous variable The power obtained after solving Approximate the result of the rigorous solution, i.e., assume:

[0056] in, For the scene The generator power vector of the next unit. Simultaneously, in the constraints of the above multi-scenario power generation planning model, the logical constraints between maintenance variables and unit state variables are as follows: It can be converted into: .

[0057] when At that time, it was obvious that ,Right now .because Since there is no continuous state, the following approach is adopted:

[0058] By approximating the logical constraints of the strict maintenance variables and unit state variables mentioned above, a strict maintenance plan can still be obtained under the above framework. Furthermore, due to the consideration of multiple scenarios, the robustness and adaptability of this maintenance plan will be significantly enhanced.

[0059] Scenario-based maintenance plan development model Augmented Lagrange penalty function of maintenance plan connection variables Penalties for curtailment of renewable energy Generator set capacity blockage penalty All-day power supply capacity surplus target Minimum power supply capacity bus surplus target Minimum power supply capacity surplus target Target for surplus capacity of new energy consumption Minimum absorption capacity busbar profit target Minimum absorption capacity period profit target The specific details are as follows:

[0060]

[0061]

[0062]

[0063]

[0064]

[0065]

[0066]

[0067]

[0068] in, For maintenance plan i of d The 0-1 status on the day, from Given, it is a constant; For maintenance plan i of d The daily maintenance status is a communication variable, with a value of 0 to 1, which is a continuous quantity. For the connection variable The Lagrange penalty coefficient; For the connection variable The augmented Lagrange penalty coefficient; For the scene s In the context of renewable energy units, the amount of power curtailed during peak / valley load periods is denoted as ; For the scene s Lower busbar k exist d The daily net load peak hour power supply capacity is in surplus; For the scene s Dispatchable units i exist d Daily net load peak time blocking capacity; For the scene s New energy units i exist d The daily net load trough can absorb the space surplus; busbar k The next batch of new energy generating units; A collection of new energy generating units for the power system; A set of dispatchable generating units in a power system; This refers to the collection of busbars in a power system.

[0069] The coefficients are: renewable energy curtailment penalty coefficient, generator capacity congestion penalty coefficient, all-time power supply capacity surplus target coefficient, minimum power supply capacity bus surplus target coefficient, minimum power supply capacity time period surplus target coefficient, renewable energy absorption capacity surplus target coefficient, minimum absorption capacity bus surplus target coefficient, and minimum absorption capacity time period surplus target coefficient.

[0070] This also considers constraints related to dispatchable generating units (reservoir-capacity hydropower and thermal power units), dispatchable power plants (reservoir-capacity hydropower and thermal power plants), new energy power plants (wind power, photovoltaic, and run-of-river hydropower), generator unit maintenance constraints, power system bus constraints, and power system power constraints. Specific constraints are as follows: Constraints on dispatchable generating units (reservoir capacity hydropower and thermal power units): Coupling constraints between availability and maintenance:

[0071] Unit fixed start-up mode constraints:

[0072] Power-on method logic constraints:

[0073] Net load peak-time blocking power constraints:

[0074] Net load valley power constraints:

[0075] Net load peak time power constraint verification:

[0076] in, For maintenance needs i exist d The execution state variable for each day is relaxed to a continuous variable; For maintenance needs i exist d The start-of-day flag variable is relaxed to a continuous variable; As a dispatchable unit i exist d The available binary state quantities of a day are relaxed to continuous variables. For the scene s Dispatchable units i exist d The daily power-on method binary state quantity is relaxed to a continuous variable; For the scene s Dispatchable units i exist d Daily net load off-peak power; For the scene s Dispatchable units i exist d Verify the power at the daily net load peak load time; As a dispatchable unit i exist d Available generating capacity and minimum permissible output at peak daily net load; As a dispatchable unit i exist d Available generating capacity and minimum permissible output during daily net load off-peak hours.

[0077] Constraints of dispatchable power plants (reservoir capacity hydropower plants, thermal power plants):

[0078] in, For power plantsp of d Minimum number of machines to be started per day; For power plants i The following is a collection of units; A collection of power plants in a power system; Constraints on new energy power plants (wind power, photovoltaic, run-of-river hydropower): Net load peak-hour power curtailment constraints:

[0079] Net load off-peak curtailment constraints:

[0080] Net load valley absorption constraints:

[0081] in, The available resource power of new energy unit i during the peak and valley times of net load on day d under scenario s; For new energy units i The installed capacity.

[0082] Generator set maintenance constraints: Maintenance start time constraints:

[0083] Maintenance duration constraints:

[0084] Power system bus constraints: Net load peak-hour power supply capacity surplus constraint:

[0085] Net load peak time standby constraints:

[0086] in, busbar k exist d Daily net peak and valley load electricity demand; busbar k exist d Daily electricity demand ceiling; busbar k exist d Daily fixed reserve power requirements.

[0087] Power system power constraints: Net load peak time branch power flow constraints:

[0088] Net load valley time branch power flow constraints:

[0089] Break-even constraint at peak net load:

[0090] Net load valley power balance constraints:

[0091] Power-on method verification constraints:

[0092] in, branch road l The maximum active transmission limit.

[0093] Steps 2-3: Solve the two problem clusters using the ATC method. Solve the two problem clusters alternately and in parallel, and update the augmented Lagrange penalty coefficients and connection variable values ​​according to the ATC method until the maintenance connection variable values ​​obtained for the two problem clusters satisfy the convergence condition.

[0094] First, using commercial or open-source solvers, solve for the unaugmented Lagrange penalty term in parallel. Maintenance schedule arrangement problem cluster Obtain the initial values ​​of the maintenance status communication variables. .

[0095] Subsequently, using commercial or open-source solvers, the maintenance planning coordination problem cluster is solved in parallel. Obtain new values ​​for maintenance status communication variables. Parallel solutions can be obtained using commercial or open-source solvers, taking into account the augmented Lagrange penalty term. Maintenance schedule arrangement problem cluster Obtain new values ​​for maintenance status communication variables. Update maintenance plan coordination issues cluster The coefficients of the augmented Lagrange penalty function:

[0096] in, k This represents the number of iterations. To augment the Lagrange penalty increment coefficient.

[0097] Solve the maintenance planning coordination problem cluster in parallel again. Obtain new values ​​for maintenance status communication variables. Calculate the numerical convergence criterion or the 0-1 convergence criterion. If the criterion is met, the iteration ends; otherwise, the maintenance plan scheduling problem cluster is updated. The coefficients of the augmented Lagrange penalty function:

[0098] Repeat the above alternating solution process until the algorithm converges to obtain a set of general annual generator set maintenance schedules.

[0099] Specifically, the numerical convergence criterion is: The 0-1 convergence criterion is: The 0-1 convergence criterion does not require maintenance scheduling issues. The connection variables converge to the coordination model The given maintenance status values ​​only require coordination of the model. The maintenance schedule obtained from two consecutive iterations remains unchanged. In practical applications, in most cases, the 0-1 convergence criterion can significantly reduce the number of algorithm iterations, and during the period from 0-1 convergence to numerical convergence, the coordination model... Since the established maintenance status usually does not change, it is actually recommended to use it as the convergence criterion for the "0-1 convergence criterion" algorithm.

[0100] Example To verify the effectiveness of this method, the JPES-ROTS testing system was selected for analysis in this embodiment. The programming language was C++, the commercial solver used was Gurobi v11.0.0, and the computer hardware configuration was a 64-core AMD Ryzen Threadripper PRO 3995WX with 256GB of RAM.

[0101] This implementation sets 36 maintenance requirements, with a unified maintenance optimization window of six months. Maintenance requirement durations range from 2 to 3 weeks. Detailed maintenance requirements are shown in Table 1. The JPES-ROTS testing system comprises 44 buses, 77 branches, 24 thermal power units, 10 reservoir-capacity hydropower units, and 11 new energy units. The example maintenance requirement data is shown in Table 1.

[0102] To test the superiority of the present invention, this embodiment sets up a total of 5 comparative cases under different number of scenarios to analyze the performance of the algorithm under different number of scenarios. The number of scenarios for different cases is shown in Table 2.

[0103]

[0104] Figure 1The diagram shows a line graph illustrating the solution time of the integrated iterative solution method and the direct solution model proposed in this embodiment. In terms of time, the ATC iterative solution significantly reduces the solution time, achieving a speedup of approximately 5.42 times for 50 scenarios, reducing the time from approximately 7 minutes to about 1 minute. Furthermore, the solution time of the iterative method does not significantly increase with the number of scenarios; the resulting speedup gain becomes increasingly significant with the number of scenarios, demonstrating a significant advantage in maintenance and optimization problems involving massive number of scenarios. This invention considers a large number of scenarios simultaneously during the planning stage (rather than focusing on one or a few specific scenarios), ensuring that the plan is usable in most cases. This ensures that even if the actual scenarios change compared to the calculations, the plan still has relatively good applicability. The model also incorporates targeted algorithms for the large number of scenarios, splitting the originally large problem into two alternating solution clusters.

[0105] Figure 2 The results show histograms of the convergence gap when the two solution methods converge. The results indicate that the iterative algorithm is not significantly inferior to the integrated solution in terms of optimality of the solution. It can solve for a smaller gap in a shorter time, and both gaps are less than 1%. This means that the optimality of the algorithm is not sensitive to changes in the number of scenarios and can be effectively applied to maintenance and optimization problems with massive scenarios, meeting general engineering needs.

[0106] Figure 3 A bar chart comparing the number of iterations for "numerical convergence" and "0-1" convergence under different numbers of scenarios is presented. The iterative algorithm maintains a stable convergence speed with increasing number of scenarios, meaning the number of convergence iterations does not change significantly with the number of scenarios. Furthermore, setting the convergence condition to 0-1 convergence—that is, stopping iteration once the maintenance plan is determined—further reduces the number of iterations when the model without a maintenance plan meets numerical convergence requirements. This results in an average reduction of approximately 30-40% in iterations / computation time, demonstrating significant efficiency.

[0107] Figure 4 Demonstrates the convergence process of the iterative method under different scenarios. The changes in the average value, maximum value, and objective function are analyzed. The results show that the algorithm's iterative convergence stability is significantly enhanced as the number of scenarios increases. Furthermore, when the number of scenarios is large, the 0-1 convergence value typically does not change further.

[0108] Figure 5 The paper presents the system power generation capacity balance at the net load peak load time under 50 scenarios, using the iterative method and the integrated solution. Both solution methods schedule maintenance results to the spring and autumn low load periods, and make slight local adjustments based on the renewable energy power generation situation, which illustrates the rationality of the solution results of this algorithm.

[0109] To verify the practical application value of this invention, further verification was conducted using real wind and solar power data from a provincial-level system in my country. The example included 80 maintenance requirements and 50 new energy scenarios, encompassing 611 nodes, 709 branches, 26 thermal power units, 248 hydropower units, and several new energy power plants. The algorithm proposed in this embodiment converged after 13 iterations with a threshold of 0.00199, achieving a solution time of 8315 seconds (2.3 hours). When using the 0-1 criterion, the algorithm converged on the 9th iteration, significantly reducing the solution time to 5439 seconds (1.51 hours).

[0110] Figure 6 The iterative process of solving the annual power generation maintenance plan optimization problem of the above system using the method proposed in this invention is demonstrated. Results show that in practical applications, the number of iterations using the method of this invention does not significantly increase compared to the small test case, and the convergence process is also very stable.

[0111] This invention focuses on scheduling the power generation and maintenance of large dispatchable generating units based on future annual forecasts / typical data. Using an annual cycle and daily time units, it captures peak and off-peak net load moments each day to maximize system power surplus and absorption capacity. Simultaneously considering network constraints on supply and regulation capabilities, a stochastic optimization model for annual dispatchable generating unit maintenance plans considering multiple renewable energy scenarios is constructed. Through in-depth analysis and decomposition of the model's mathematical structure, the ATC method is used to break down the complete model into two major problem clusters that can be solved in parallel. By alternately solving the two problem clusters, a set of annual power generation and maintenance plans applicable to all renewable energy scenarios within the model can be obtained after a finite number of iterations.

[0112] The method designed in this invention has a convergence speed that is insensitive to system size and good iterative stability. At the same time, it can significantly reduce computational costs and save computation time. It also effectively improves the rationality, versatility and robustness of the annual maintenance plan for large dispatchable units such as thermal power and hydropower units, and provides effective technical support for the load supply tasks and new energy consumption targets of the new power system.

[0113] In summary, the proposed method for scheduling annual power generation maintenance plans in high-proportion renewable energy power systems has significant theoretical and practical application value. At the theoretical level, the method constructs a stochastic optimization model for annual dispatchable unit power generation maintenance plans that considers multiple renewable energy scenarios. This model deeply integrates the characteristics of renewable energy power generation with the actual needs of system operation, comprehensively considering various constraints. It provides new ideas and model frameworks for academic research in related fields, enriching the theoretical system of power system maintenance scheduling.

[0114] It should be emphasized that the above are merely preferred embodiments of the present invention and are not intended to limit the present invention in any way. Any simple modifications, equivalent changes and alterations made to the above embodiments based on the technical essence of the present invention shall still fall within the scope of the technical solution of the present invention.

Claims

1. A method for scheduling annual power generation and maintenance in a high-proportion renewable energy power system, characterized in that, Includes the following steps: Step 1: Data collection, cleaning and preparation; acquire historical power system operation data, preprocess the data to generate diverse power system renewable energy output scenarios; Step 2: Establish and solve the maintenance plan optimization model; With the goal of minimizing the augmented Lagrange penalty function of the maintenance plan linkage variables, a 0-1 maintenance plan coordination problem is constructed for each maintenance requirement, forming a maintenance coordination problem cluster. With the goal of minimizing the augmented Lagrange penalty function of the maintenance plan linkage variables, renewable energy curtailment, and unit congestion costs, and maximizing the system's power supply capacity and absorption capacity, a maintenance plan formulation sub-model is constructed for each renewable energy scenario, forming a maintenance formulation problem cluster. The maintenance coordination problem cluster and the maintenance formulation problem cluster are solved alternately and in parallel, and the augmented Lagrange penalty term coefficients and linkage variable values ​​are updated according to the ATC method until the maintenance linkage variable values ​​obtained from the two problem clusters satisfy the convergence condition.

2. The method for arranging annual power generation and maintenance plans for a high-proportion renewable energy power system according to claim 1, characterized in that, In step 1, the historical power system operation data includes: historical weather data, wind power output curves, photovoltaic power output curves, power system grid parameters, generator parameters, load demand, and generator maintenance demand data; the preprocessing includes: data cleaning, noise reduction, normalization, and verification and error correction of power system parameters.

3. The method for arranging annual power generation and maintenance plans for a high-proportion renewable energy power system according to claim 1, characterized in that, In step 2, the augmented Lagrange penalty function of the maintenance plan connection variable is: in, For maintenance plan i of d Daily state variables; For maintenance plan i of d Daily maintenance status communication variables; For the connection variable The Lagrange penalty coefficient; For the connection variable The augmented Lagrange penalty coefficient; Allow start / end time for maintenance plan i; A collection of generator set maintenance plans; A collection of new energy scenarios; For daily indexing.

4. The method for arranging annual power generation and maintenance plans for a high-proportion new energy power system according to claim 3, characterized in that, The constraints of the augmented Lagrange penalty function of the maintenance plan connection variables include: the start time and duration constraints of the maintenance requirements.

5. The method for arranging annual power generation and maintenance plans for a high-proportion new energy power system according to claim 1, characterized in that, In step 2, the sub-model for constructing and formulating maintenance plans for new energy scenarios is as follows: in, To augment the Lagrange penalty function, Penalties for curtailment of renewable energy sources For generator set capacity blockage penalty, To achieve the goal of surplus power supply capacity around the clock, For the minimum power supply capacity bus surplus target, For the minimum power supply capacity period surplus target, To achieve the target of surplus capacity for new energy consumption, For the minimum absorption capacity busbar profit target, The profit target is set for the period with the lowest absorption capacity. For maintenance plan i of d The 0-1 status on the day, from Given as a constant; For maintenance plan i of d The daily maintenance status is a communication variable, with a value of 0 to 1, which is a continuous quantity. For the connection variable The Lagrange penalty coefficient; For the connection variable The augmented Lagrange penalty coefficient; For the scene s In the context of renewable energy units, the amount of power curtailed during peak / valley load periods is denoted as ; For the scene s Lower busbar k exist d The daily net load peak hour power supply capacity is in surplus; For the scene s Dispatchable units i exist d Daily net load peak time blocking capacity; For the scene s New energy units i exist d The daily net load trough can absorb the space surplus; busbar k The next batch of new energy generating units; A collection of new energy generating units for the power system; A set of dispatchable generating units in a power system; This refers to the collection of busbars in a power system.

6. The method for arranging annual power generation and maintenance plans for a high-proportion new energy power system according to claim 5, characterized in that, The constraints of the sub-model for constructing maintenance plans in the new energy scenario include: coupling constraints between availability and maintenance, constraints on fixed unit start-up mode, logical constraints on start-up mode, power constraints during peak net load time, power constraints during valley load time, power constraints during peak net load time verification, and constraints on dispatchable power plants.

7. The method for arranging annual power generation and maintenance plans for a high-proportion new energy power system according to claim 5, characterized in that, The constraints of the sub-model for constructing maintenance plans for the new energy scenario include constraints on new energy power plants: net load peak-hour curtailment constraints, net load valley-hour curtailment constraints, and net load valley-hour absorption constraints.

8. The method for arranging annual power generation and maintenance plans for a high-proportion new energy power system according to claim 5, characterized in that, The constraints of the sub-model for constructing maintenance plans in the new energy scenario include: generator set maintenance constraints, maintenance start time constraints, maintenance duration constraints, net load peak power supply capacity surplus constraints, net load peak standby constraints, net load peak branch power flow constraints, net load valley branch power flow constraints, net load peak profit and loss balance constraints, net load valley power balance constraints, and start-up mode verification constraints.

9. The method for arranging annual power generation and maintenance plans for a high-proportion renewable energy power system according to claim 1, characterized in that, In step 2, the alternating parallel solution specifically includes: First, parallel solution of the unaugmented Lagrange penalty term. Maintenance schedule arrangement problem cluster Obtain the initial values ​​of the maintenance status communication variables. ; Subsequently, the maintenance planning coordination problem cluster was solved in parallel. Obtain new values ​​for maintenance status communication variables. Parallel solution considering augmented Lagrange penalty terms Maintenance schedule arrangement problem cluster Obtain new values ​​for maintenance status communication variables. Update maintenance plan coordination issues cluster The coefficients of the augmented Lagrange penalty function: in, k This represents the number of iterations. To augment the Lagrange penalty increment coefficient; Solve the maintenance planning coordination problem cluster in parallel again. Obtain new values ​​for maintenance status communication variables. Calculate the numerical convergence criterion or the 0-1 convergence criterion. If the criterion is met, the iteration ends; otherwise, the maintenance plan scheduling problem cluster is updated. The coefficients of the augmented Lagrange penalty function: Repeat the above alternating solution process until the algorithm converges to obtain the annual generator set maintenance plan.

Images