Machine learning-based accelerated optimization method and system for unit commitment by eliminating redundant constraints

Abstract

The present invention discloses a method and system for accelerating the optimization of unit commitment by eliminating redundant constraints based on machine learning, including: transforming the original unit commitment problem into a problem without line power flow constraints, using 0-1 variables to control the establishment of line power flow constraints, thereby eliminating dense and complex security constraints. And by using the machine learning method to learn the range of the number of changes of each unit in all time periods, a recognition rule for redundant minimum continuous start-stop time constraints is formed, thereby eliminating redundant minimum continuous start-stop time constraints, reducing the number of problem constraints and the problem scale, and improving the efficiency of unit commitment optimization.

Description

Technical Field

[0001] The present invention belongs to the field of machine learning, and relates to a method and system for accelerating the optimization of unit commitment based on machine learning to eliminate redundant constraints. Background Art

[0002] Complex and dense security constraints and minimum continuous start-stop time constraints are one of the key factors affecting the solution efficiency of large-scale unit commitment. In fact, some of these constraints are invalid or redundant, and removing them from the original model will not affect the overall solution domain, but adding these constraints makes the problem difficult to solve; reducing the security power flow constraints and minimum continuous start-stop time constraints can greatly improve the solution efficiency of the overall model. However, how to accurately identify the characteristics of these constraints and eliminate redundant constraints is a difficult point. Summary of the Invention

[0003] The purpose of the present invention is to solve the problems in the prior art that the characteristics of the security power flow constraints and the minimum continuous start-stop time constraints cannot be accurately identified and redundant constraints cannot be eliminated, and to provide a method and system for accelerating the optimization of unit commitment based on machine learning to eliminate redundant constraints.

[0004] To achieve the above object, the present invention adopts the following technical solutions:

[0005] A method for accelerating the optimization of unit commitment based on machine learning to eliminate redundant constraints, comprising:

[0006] Obtaining the semi-continuous constraints of the upper and lower limits of the unit output of the SCUC model and the line power flow constraints;

[0007] If both the semi-continuous constraints of the upper and lower limits of the unit output in the SCUC model and the line power flow constraints hold, the upper and lower limits of the two constraints are very close, and then a sub-problem containing only 0-1 variables is introduced and the sub-problem containing only 0-1 variables is solved;

[0008] Adding the solution of the sub-problem containing only 0-1 variables to the objective function of the original unit commitment problem based on a penalty method, so that the solution of the sub-problem is as close as possible to the optimal solution of the original unit commitment problem;

[0009] Based on the solution of the sub-problem and the optimal solution of the original unit commitment problem, obtaining the offset of the unit state variable of each unit at each time period between the two; changing the characteristic requirement D of the original unit commitment problem, obtaining multiple optimal solutions, obtaining the relationship between the requirement D and the offset, and learning it into a predictive regression model;

[0010] Given a new unit commitment problem, predicting the offset of the unit state variable of each unit at each time period of the new unit commitment problem to be solved based on the learned predictive model, identifying and removing the redundant constraints in the minimum continuous start-stop time constraints, so as to achieve the accelerated optimization of unit commitment.

[0011] A further improvement of the present invention lies in:

[0012] Furthermore, the semi - continuous constraint on the upper and lower limits of the output of the SCUC model unit is specifically:

[0013]

[0014] where I i,t is the start - stop status variable of the unit, and P i L and P i U are the minimum and maximum outputs of the unit respectively.

[0015] Furthermore, the line power flow constraint is specifically:

[0016]

[0017] where GSF i,l is the transfer distribution factor, representing the transfer ratio of power from unit i to line l.

[0018] Furthermore, introduce a sub - problem containing only 0 - 1 variables and solve the sub - problem containing only 0 - 1 variables. Specifically: if the line power flow constraint holds, the upper and lower limits of the line power flow constraint are very close to the upper and lower limits of the transformed form of the semi - continuous constraint; construct a sub - problem containing only 0 - 1 variables to solve the values of the corresponding 0 - 1 variables. The objective of the sub - problem containing 0 - 1 variables is to minimize the distance between the upper and lower limits of the semi - continuous constraint and the upper and lower limits of the line power flow constraint, and the closer the upper and lower limits of the semi - continuous constraint are, the better.

[0019] Convert the semi - continuous constraint into the same form as the line power flow constraint, as shown in formula (3):

[0020]

[0021] Construct a sub - problem containing only 0 - 1 variables:

[0022]

[0023] where the constraint term is the constraint H(I)≥0 of the unit commitment problem containing only I i,t , W i,t and Y i,t ; W i,t and Y i,t are the decision variables of whether the unit is on or off respectively.

[0024] Convert the above objective function into a linear expression L(I i,t containing only I i,t) Solve such sub - problems to obtain the solution S of the 0 - 1 variables;

[0025]

[0026] Furthermore, in a penalty - based manner, add the solution of the sub - problem with only 0 - 1 variables to the objective function of the original unit commitment problem, so that the solution of the sub - problem is as close as possible to the optimal solution of the unit commitment problem. Specifically:

[0027]

[0028] Among them, the constraint terms are G(I, P, W)≥0, H(I)≥0. Express the unit commitment problem in terms of the unit start - stop decision variable I, the unit output variable P, and other variables w; where G(I, P, W) is the constraint set without line power flow constraints; H(I)≥0 is the constraint of the unit commitment problem with only I i,t , W i,t , Y i,t constraints of 0 - 1 variables.

[0029] Furthermore, the minimum continuous start - stop time constraint is specifically:

[0030]

[0031]

[0032] Furthermore, based on the solution of the sub - problem and the optimal solution of the original unit commitment problem, obtain the offset of the unit status variable of each unit at each time period between the two; change the characteristic demand D of the original unit commitment problem, obtain multiple optimal solutions, get the relationship between the demand D and the offset, and learn it into a prediction regression model. Specifically: change the demand D, solve the unit commitment problem, and obtain several optimal solutions I * , calculate the distance between each I * and the solution S of the sub - problem with only 0 - 1 variables; count the offset degree of each unit at each time period respectively, form the relationship between the demand D and the offset vector v; for the SCUC model determined by the demand d∈D, there is a functional relationship such that

[0033]

[0034] Among them, v j is each component of the offset degree v of each unit at each time period respectively.

[0035] Further, redundant constraints in the identified minimum continuous start-stop time constraints are recognized and removed, specifically: Given a new unit commitment problem, a learning-based prediction model predicts the offset of the unit status variable of each unit in each time period of the new unit commitment problem to be solved. For the

[0036]

[0037] or

[0038]

[0039] make a judgment. g(v i ) represents the functional expression of the offset. If formulas (10) and (11) still hold, the original constraint is redundant, and then the constraint is removed.

[0040] An accelerated unit commitment optimization system for eliminating redundant constraints based on machine learning includes:

[0041] A first acquisition module, which acquires the semi-continuous constraints of the upper and lower limits of unit output and the line power flow constraints of the SCUC model;

[0042] An introduction module, if the semi-continuous constraints of the upper and lower limits of unit output and the line power flow constraints in the SCUC model both hold, the upper and lower limits of the two constraints are very close, and then a sub-problem containing only 0-1 variables is introduced and the sub-problem containing only 0-1 variables is solved;

[0043] A second acquisition module, which adds the solution of the sub-problem containing only 0-1 variables to the objective function of the original unit commitment problem based on a penalty method, so that the solution of the sub-problem is as close as possible to the optimal solution of the original unit commitment problem;

[0044] An offset data acquisition module, which acquires the offset of the unit status variable of each unit between the two based on the solution of the sub-problem and the optimal solution of the original unit commitment problem; changes the characteristic requirement D of the original unit commitment problem, obtains multiple optimal solutions, gets the relationship between the requirement D and the offset, and learns it into a prediction regression model;

[0045] A judgment module, which, given a new unit commitment problem, predicts the offset of the unit status variable of each unit in each time period of the new unit commitment problem to be solved based on a learning-based prediction model, recognizes and removes redundant constraints in the minimum continuous start-stop time constraints to achieve accelerated unit commitment optimization.

[0046] A terminal device includes a memory, a processor, and a computer program stored in the memory and executable on the processor. When the processor executes the computer program, the steps of the above method are implemented.

[0047] A computer-readable storage medium stores a computer program. When the computer program is executed by a processor, the steps of the above method are implemented.

[0048] Compared with the prior art, the present invention has the following beneficial effects:

[0049] The present invention transforms the unit commitment problem into a problem without line power flow constraints, uses 0-1 variables to control the establishment of line power flow constraints, thereby eliminating dense and complex security constraints. And by using the machine learning method to learn the range of change times of each unit in all time periods, a redundant minimum continuous start-stop time constraint recognition rule is formed, thereby eliminating the redundant minimum continuous start-stop time constraint, reducing the number of problem constraints and the problem scale, and improving the efficiency of unit commitment optimization. BRIEF DESCRIPTION OF THE DRAWINGS

[0050] In order to more clearly illustrate the technical solutions of the embodiments of the present invention, the drawings required for use in the embodiments will be briefly introduced below. It should be understood that the following drawings only show some embodiments of the present invention, and therefore should not be regarded as limiting the scope. For those of ordinary skill in the art, other related drawings can be obtained based on these drawings without creative efforts.

[0051] Figure 1 It is a schematic flow chart of the unit commitment accelerated optimization method based on machine learning to eliminate redundant constraints of the present invention;

[0052] Figure 2 It is a schematic structural diagram of the unit commitment accelerated optimization system based on machine learning to eliminate redundant constraints of the present invention;

[0053] Figure 3 It is a schematic diagram of the unit commitment accelerated optimization algorithm based on machine learning to eliminate redundant constraints of the present invention. DETAILED DESCRIPTION OF THE EMBODIMENTS

[0054] To make the objectives, technical solutions, and advantages of the embodiments of the present invention clearer, the technical solutions in the embodiments of the present invention will be clearly and completely described below in conjunction with the drawings in the embodiments of the present invention. Obviously, the described embodiments are some, but not all, of the embodiments of the present invention. Usually, the components of the embodiments of the present invention described and shown in the drawings here can be arranged and designed in various different configurations.

[0055] Accordingly, the following detailed description of the embodiments of the present invention provided in the accompanying drawings is not intended to limit the scope of the claimed invention, but merely represents selected embodiments of the present invention. All other embodiments obtained by those of ordinary skill in the art based on the embodiments of the present invention without creative efforts shall fall within the scope of protection of the present invention.

[0056] It should be noted that like reference numerals and letters denote like items in the following drawings. Therefore, once an item is defined in one drawing, it does not need to be further defined and explained in subsequent drawings.

[0057] In the description of the embodiments of the present invention, it should be noted that if terms such as "upper", "lower", "horizontal", "inner", etc. are used to indicate the orientation or positional relationship, it is based on the orientation or positional relationship shown in the drawings, or the orientation or positional relationship in which the product of the invention is usually placed during use. This is only for the convenience of describing the present invention and simplifying the description, rather than indicating or implying that the device or element referred to must have a specific orientation, be constructed and operated in a specific orientation, and thus should not be construed as a limitation of the present invention. In addition, terms such as "first", "second", etc. are only used for distinguishing descriptions and cannot be understood as indicating or implying relative importance.

[0058] In addition, if the term "horizontal" appears, it does not mean that the component is required to be absolutely horizontal, but it can be slightly inclined. For example, "horizontal" only means that its direction is more horizontal relative to "vertical", and does not mean that the structure must be completely horizontal, but it can be slightly inclined.

[0059] In the description of the embodiments of the present invention, it should also be noted that unless otherwise clearly specified and limited, if terms such as "set", "installed", "connected", "coupled" are used, they should be understood in a broad sense. For example, it can be a fixed connection, a detachable connection, or an integral connection; it can be a mechanical connection or an electrical connection; it can be directly connected or indirectly connected through an intermediate medium, and it can be the communication inside two elements. For those of ordinary skill in the art, the specific meanings of the above terms in the present invention can be understood according to specific circumstances.

[0060] The present invention will be further described in detail below with reference to the accompanying drawings:

[0061] See Figure 1 , the present invention discloses a method for accelerating the optimal search of unit commitment based on machine learning to eliminate redundant constraints, including:

[0062] S101, obtaining the semi - continuous constraints of the upper and lower limits of unit output of the SCUC model and the line power flow constraints;

[0063] The semi - continuous constraints of the upper and lower limits of unit output of the SCUC model are specifically:

[0064]

[0065] Among them, I i,t is the start-stop state variable of the unit, and P i L and P i U are the minimum and maximum power outputs of the unit respectively.

[0066] Line power flow constraints, specifically:

[0067]

[0068] Among them, GSF i,l is the transfer distribution factor, indicating the transfer ratio of power from unit i to line l.

[0069] S102, if both the semi - continuous constraints on the upper and lower limits of the unit output in the SCUC model and the line power flow constraints hold, then the upper and lower limits of the two constraints are very close. Furthermore, a sub - problem containing only 0 - 1 variables is introduced and the sub - problem containing only 0 - 1 variables is solved;

[0070] If the line power flow constraint holds, then the distance between the upper and lower limits of the line power flow constraint is very close to the upper and lower limits of the transformed form of the semi - continuous constraint; construct a sub - problem containing only 0 - 1 variables to solve for the values of the corresponding 0 - 1 variables. The objective of the sub - problem containing 0 - 1 variables is to minimize the distance between the upper and lower limits of the semi - continuous constraint and the upper and lower limits of the line power flow constraint, and the closer the upper and lower limits of the semi - continuous constraint are, the better;

[0071] Transform the semi - continuous constraint into the same form as the line power flow constraint, as shown in formula (3):

[0072]

[0073] Construct a sub - problem containing only 0 - 1 variables:

[0074]

[0075] Among them, the constraint term is the constraint H(I)≥0 for the unit commitment problem containing only I i,t , W i,t and Y i,t ; W i,t and Y i,t are the decision variables for whether the unit is on or off respectively;

[0076] Transform the above - mentioned objective function into a linear expression L(I i,t ) containing only I i,t ), solve such a sub - problem to obtain the solution S of the 0 - 1 variables;

[0077]

[0078] S103. Add the solution of the sub-problem with only 0-1 variables to the objective function of the original unit commitment problem in a penalty-based manner, so that the solution of the sub-problem is as close as possible to the optimal solution of the original unit commitment problem. Specifically:

[0079]

[0080] Among them, the constraint terms are G(I, P, W) ≥ 0 and H(I) ≥ 0. Express the unit commitment problem in terms of the unit start-stop decision variable I, the unit output variable P, and other variables w. Among them, G(I, P, W) is the constraint set without line power flow constraints, and H(I) ≥ 0 is the constraint of the unit commitment problem with only I i,t , W i,t , Y i,t constraints of 0-1 variables.

[0081] S104. Based on the solution of the sub-problem and the optimal solution of the original unit commitment problem, obtain the offset of the unit state variable of each unit at each time period between the two. Change the characteristic demand D of the original unit commitment problem, obtain multiple optimal solutions, get the relationship between the demand D and the offset, and learn it into a prediction regression model;

[0082] Change the demand D, solve the unit commitment problem, and obtain several optimal solutions I * , calculate the distance between each I * and the solution S with only 0-1 variables; count the offset degree of each unit at each time period respectively, and form the relationship between the demand D and the offset vector v; for the SCUC model determined by the demand d ∈ D, there is a functional relationship such that

[0083]

[0084] Among them, v j is each component of the offset degree v of each unit at each time period respectively.

[0085] S105. Given a new unit commitment problem, predict the offset of the unit state variable of each unit at each time period of the new unit commitment problem to be solved based on the learned prediction model, identify and remove the redundant constraints in the minimum continuous start-stop time constraint to achieve accelerated optimization of the unit commitment.

[0086] The minimum continuous start-stop time constraint is specifically as follows:

[0087]

[0088]

[0089] Determine whether it is the minimum continuous start-stop time constraint. Specifically: for the

[0090]

[0091] or

[0092]

[0093] make a judgment. g(v i ) represents the functional expression of the offset. If formulas (10) and (11) still hold, the original constraint is redundant, and then this constraint is removed.

[0094] See Figure 2 , the present invention discloses a unit commitment accelerated optimization system for eliminating redundant constraints based on machine learning, including:

[0095] A first acquisition module, which acquires the semi-continuous constraints of the upper and lower limits of the unit output of the SCUC model and the line power flow constraints;

[0096] An introduction module. If the semi-continuous constraints of the upper and lower limits of the unit output and the line power flow constraints in the SCUC model both hold, the upper and lower limits of the two constraints are very close, and then a sub-problem containing only 0-1 variables is introduced and the sub-problem containing only 0-1 variables is solved;

[0097] A second acquisition module, which adds the solution of the sub-problem containing only 0-1 variables to the objective function of the original unit commitment problem in a penalty-based manner, so that the solution of the sub-problem is as close as possible to the optimal solution of the original unit commitment problem;

[0098] An offset data acquisition module, which acquires the unit state variable offset of each unit between the two at each time period based on the solution of the sub-problem and the optimal solution of the original unit commitment problem; changes the characteristic requirement D of the original unit commitment problem, obtains multiple optimal solutions, gets the relationship between the requirement D and the offset, and learns it into a prediction regression model;

[0099] A judgment module, which gives a new unit commitment problem, predicts the unit state variable offset of each unit of the new unit commitment problem to be solved based on the learned prediction model, identifies and removes the redundant constraints in the minimum continuous start-stop time constraint, so as to achieve accelerated optimization of unit commitment.

[0100] Embodiment:

[0101] See Figure 3, the present invention discloses a unit commitment accelerated optimization algorithm based on machine learning to eliminate redundant constraints, including removing line power flow constraints and redundant minimum continuous start-stop time constraints.

[0102] The first part is to represent the line power flow constraint by 0-1 variables, thereby converting the information of the line power flow constraint into the values of 0-1 variables, and adding them to the objective function of the original problem in the form of penalties, so that it can be restored to the original optimal solution and the 0-1 solution will not deviate too far. The second part takes the solution S of the first part as a benchmark, learns the offset of the values of the variables of each unit in S from the corresponding optimal solution at 96 time moments through machine learning methods, and uses this offset to judge which minimum continuous start-stop time constraints are redundant and do not play an important boundary role in the solution domain. After identification, such constraints can be removed, which can further reduce the problem scale and play an accelerating role.

[0103] According to the connection between the security constraints in the thermal-hydropower problem and the upper and lower limits of unit output constraints, in the first stage, the unit commitment problem is transformed into a problem without line power flow constraints through traditional optimization methods, and 0-1 variables are used to control the establishment of line power flow constraints, thereby eliminating dense and complex security constraints. On this basis, in the second stage, machine learning methods are used to learn the range of change times of each unit in all time periods, form identification rules for redundant minimum continuous start-stop time constraints, and thus eliminate redundant minimum continuous start-stop time constraints.

[0104] (1) Reconstruct the line power flow constraint using 0-1 variables

[0105] Since the SCUC model has semi-continuous constraints such as the upper and lower limits of unit output, as shown in formula (1):

[0106]

[0107] where, I i,t is the start-stop state variable of the unit, P i L and P i U are the minimum and maximum outputs of the unit respectively.

[0108] The line power flow constraint is as shown in formula (2):

[0109]

[0110] where, GSF i,l is the transfer distribution factor, indicating the transfer ratio of power from unit i to line l.

[0111] If the semi-continuous constraint holds, convert the semi-continuous constraint into the same form as the line power flow constraint, as shown in formula (3):

[0112]

[0113] If the line flow constraint holds, then the upper and lower limits of the line flow constraint are very close to the upper and lower limits of the semi - continuous constraint. Then, for the overall problem, there is no need to add dense and complex line flow constraints. Instead, a sub - problem containing only 0 - 1 variables is constructed to solve for the values of the corresponding 0 - 1 variables. The objective of this sub - problem is to minimize the distance between the upper and lower limits of the semi - continuous constraint and the upper and lower limits of the line flow constraint, and the closer the upper and lower limits of the semi - continuous constraint are, the better. That is, the line flow constraint is made to hold through the values of these 0 - 1 variables. Further, the solutions of these 0 - 1 variables can be added to the objective function of the original problem in a penalty manner. This makes the solution of the sub - problem as close as possible to the true optimal solution, thus playing the role of fixing variables in the solution process, accelerating the solution efficiency, and at the same time making the line flow constraint hold.

[0114] The construction steps of the sub - problem are as follows:

[0115] Step 1 (Construct a small - scale sub - problem):

[0116] The objective function of the unit commitment problem:

[0117]

[0118] Constraints: The unit commitment problem only contains the constraints of I i,t , W i,t , Y i,t such that H(I)≥0; where, W i,t and Y i,t are the decision variables for whether the unit is on or off, respectively.

[0119] Transform the above - mentioned objective function into a linear expression L(I i,t ) that only contains I i,t ), solve such a sub - problem, and obtain the solution S of the 0 - 1 variables;

[0120]

[0121] Transform it into a linear expression L(I i,t ) that only contains I i,t ), solve such a sub - problem, and obtain the solution S of the 0 - 1 variables.

[0122] Step 2 (Solve the original problem with penalty terms)

[0123] Based on a penalty approach, add the solution of the 0 - 1 variable sub - problem to the objective function of the unit commitment problem. Specifically:

[0124]

[0125] Constraints: G(I, P, W) ≥ 0, H(I) ≥ 0, and other constraints

[0126] Express the unit commitment problem in terms of the unit start-stop decision variable I, the unit output variable P, and other variables w. Among them, G(I, P, W) is the set of constraints without line power flow constraints. H(I) ≥ 0 is the constraint of the original problem that only contains I i,t , W i,t , Y i,t The distance between the solution of the 0-1 variable and the solution of the sub-problem is penalized, which is equivalent to repairing the solution of the sub-problem to a feasible solution while retaining the possibility of the line power flow constraint being satisfied. This is because both the sub-problem and the original problem contain a large number of constraints related to 0-1 variables, and the difference between the feasible solutions of the original problem and the sub-problem is not significant. Solving the transformed problem above can achieve accelerated solution.

[0127] (2) Use machine learning methods to predict the offset of the minimum continuous start-stop time

[0128] Through the method of constructing and solving sub-problems in stages above, the problem that the dense and complex line power flow constraints affect the solution efficiency is solved. There are many forms of expression for the minimum continuous start-stop time constraint. According to this solution, its form is as shown in formulas (7) and (8)

[0129]

[0130]

[0131] By obtaining the solution S of the 0-1 variable formed by some constraints in the unit commitment problem. This solution S is still some distance from the true optimal solution, and this solution S is used to measure the change of the true optimal solution. That is, observe the offset between the true optimal solution and the solution S. And only need to observe the offset of each unit in 96 time periods in this offset, and predict the offset of each unit state variable in 96 time periods by means of the deviation between the information of the true solution and this solution S. And the solution S is unchanged and can be solved offline as a measurement benchmark. When the offset of each unit state variable in 96 time periods is predicted, subtract or add the offset from the value of the constraint expression of the minimum continuous start-stop time constraint of the 0-1 variable sub-problem to see if it exceeds the right-hand side of the formula. If it does not exceed, it means that this formula does not affect the overall optimality and is a redundant constraint, which can be removed. The steps of this algorithm are as follows:

[0132] Step 1: Construct a sub-problem formed by the constraints that only contain 0-1 variables in the original problem. As shown in (1), the transformed form of the line power flow constraint is used as the objective function, and the solution S is solved.

[0133] Step 2: Change the demand D, and obtain multiple optimal solutions I by solving the original problem * , and calculate each I * 's distance from S. For example, if I * = (0, 1, 1, 0) and S = (0, 0, 0, 0), the distance s is calculated as 2. The dimension of the distance vector here is the number of units, and the deviation degree (distance) of each unit at its own 96 moments is respectively counted to form the relationship between D and the deviation vector v, and a prediction model is learned. The prediction model can be predicted using a regression model.

[0134] For the SCUC model determined by the parameter d ∈ D of demand, there exists such a functional relationship such that

[0135]

[0136] where v j is each component of the deviation degree v of each unit at its own 96 moments.

[0137] After predicting each deviation amount,

[0138] For the

[0139]

[0140] or

[0141]

[0142] in the minimum continuous start-stop time constraint, make a judgment. g(v i ) represents the functional expression of the deviation amount. If formulas (10) and (11) still hold, the original constraint is redundant, and then this constraint is removed, so as to achieve the effect of reducing the problem scale and improving the solution efficiency.

[0143] Performance test:

[0144] Adopt the adjusted examples constructed based on the standard examples (IEEE 30, IEEE 118, WP 2383), and these examples contain more complex power flow constraints.

[0145] Table 1 Comparison of problem scales

[0146]

[0147] Table 2 Comparison of SCIP solution efficiency

[0148]

[0149] From the above result analysis, it can be seen that by reducing the dense, complex, and numerous line flow constraints and the minimum continuous start-stop time constraints, the present invention can greatly reduce the scale of the problem, thereby accelerating the solution of the unit commitment problem. The solution efficiency is improved significantly, especially more prominent in large-scale models. The solution efficiency is basically improved by about 1.7 to 2 times. The number of violations of the line flow constraints is 0, indicating the effectiveness and reliability of the present invention, and it does not significantly change the solution domain. Thus, the innovation and effectiveness of the present invention can also achieve a good acceleration effect on the relatively slow SCIP.

[0150] A terminal device provided by an embodiment of the present invention. The terminal device of this embodiment includes: a processor, a memory, and a computer program stored in the memory and executable on the processor. When the processor executes the computer program, the steps in the above-mentioned various method embodiments are implemented. Or, when the processor executes the computer program, the functions of each module / unit in the above-mentioned various device embodiments are implemented.

[0151] The computer program can be divided into one or more modules / units. The one or more modules / units are stored in the memory and executed by the processor to complete the present invention.

[0152] The terminal device can be a computing device such as a desktop computer, a notebook, a palm computer, and a cloud server. The terminal device may include, but is not limited to, a processor and a memory.

[0153] The processor may be a central processing unit (CPU), or may also be other general-purpose processors, digital signal processors (DSPs), application-specific integrated circuits (ASICs), field-programmable gate arrays (FPGAs), or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components, etc.

[0154] The memory can be used to store the computer program and / or module. The processor realizes various functions of the terminal device by running or executing the computer program and / or module stored in the memory, and by calling the data stored in the memory.

[0155] If the modules / units integrated in the terminal device are implemented in the form of software function units and sold or used as independent products, they can be stored in a computer-readable storage medium. Based on such an understanding, to implement all or part of the processes in the above-described embodiment methods of the present invention, it can also be completed by a computer program instructing relevant hardware. The computer program can be stored in a computer-readable storage medium. When the computer program is executed by a processor, the steps of the above-described various method embodiments can be implemented. Among them, the computer program includes computer program code, and the computer program code can be in the form of source code, object code, executable file, or some intermediate form, etc. The computer-readable medium can include: any entity or device capable of carrying the computer program code, recording medium, USB flash drive, mobile hard disk, magnetic disk, optical disc, computer memory, read-only memory (ROM, Read-Only Memory), random access memory (RAM, Random Access Memory), electrical carrier signal, telecommunication signal, and software distribution medium, etc. It should be noted that the content included in the computer-readable medium can be appropriately increased or decreased according to the requirements of legislation and patent practice in the jurisdiction. For example, in some jurisdictions, according to legislation and patent practice, the computer-readable medium does not include electrical carrier signals and telecommunication signals.

[0156] The above are only the preferred embodiments of the present invention and are not intended to limit the present invention. For those skilled in the art, the present invention can have various changes and modifications. Any modification, equivalent replacement, improvement, etc. made within the spirit and principle of the present invention shall be included within the protection scope of the present invention.

Claims

1. A unit commitment optimization method based on machine learning to eliminate redundant constraints, characterized in that: include: Obtain the semi-continuous constraints of the upper and lower limits of the SCUC model unit output, specifically: Among them, I i,t is the start and stop state variable of the unit, and are the minimum and maximum output of the unit respectively; Get the line flow constraints, specifically: Among them, GSF i,l is the transfer distribution factor, which indicates the proportion of power transferred from unit i to line l; If the semi-continuous constraints of the upper and lower limits of the unit output and the line power flow constraints in the SCUC model are both established, then a sub-problem containing only 0-1 variables is introduced and solved to obtain the solution S of the 0-1 variables; The solution of the sub-problem containing only 0-1 variables is added to the objective function of the original unit combination problem in a penalty-based manner, so that the solution of the sub-problem is close to the optimal solution of the original unit combination problem, specifically: Among them, the constraints are G(I,P,W)≥0, H(I)≥0, I is the unit start and stop decision variable, P is the unit output variable, and w is other variables; G(I,P,W) is a constraint set without line flow constraints; H(I)≥0 means that the unit combination problem only contains I i,t , W i,t , Y i,t , constraints on 0-1 variables; Based on the solution of the sub-problem and the optimal solution of the original unit combination problem, the unit state variable offset of each unit in each time period is obtained; the characteristic demand D of the original unit combination problem is changed to obtain multiple optimal solutions, the relationship between the demand D and the offset is obtained, and a prediction regression model is learned; given a new unit combination problem, the unit state variable offset of each unit in the new unit combination problem to be solved in each time period is predicted based on the learned prediction model, and for the minimum continuous start-stop time constraint or Among them, W i,t and Y i,t are the decision variables for whether to start or shut down the unit; Make a judgment, g(v i ) represents the functional expression of the offset. If formula (10) and formula (11) are still valid, the original minimum continuous start-stop time constraint is redundant, and this constraint is removed to achieve accelerated optimization of unit combination.

2. The unit commitment accelerated optimization method based on machine learning to eliminate redundant constraints according to claim 1 is characterized in that: The subproblem containing only 0-1 variables is introduced and solved, specifically: if the line power flow constraint is established, the upper and lower limits of the line power flow constraint are very close to the upper and lower limits of the semi-continuous constraint transformation form; a subproblem containing only 0-1 variables is constructed to solve the value of the corresponding 0-1 variable, and the goal of the subproblem containing only 0-1 variables is to minimize the distance between the upper and lower limits of the semi-continuous constraint and the upper and lower limits of the line power flow constraint, and the closer the upper and lower limits of the semi-continuous constraint are, the better; The semi-continuous constraint is transformed into the same form as the line power flow constraint, as shown in formula (3): Construct a subproblem that contains only 0-1 variables: Among them, the constraint term is the unit commitment problem containing only I i,t , W i,t and Y i,t The constraint H(I)≥0; The above objective function is transformed into a function containing only I i,t The linear expression L(I i,t ), solve such subproblems and obtain the solution S of 0-1 variables.

3. The unit commitment accelerated optimization method based on machine learning to eliminate redundant constraints according to claim 2 is characterized in that: The solution based on the sub-problem and the optimal solution of the original unit combination problem is used to obtain the unit state variable offset of each unit in each time period; the characteristic demand D of the original unit combination problem is changed to obtain multiple optimal solutions, the relationship between the demand D and the offset is obtained, and a prediction regression model is learned; specifically: the demand D is changed, the unit combination problem is solved, and several optimal solutions I are obtained. * , calculate each I * The distance from the sub-problem solution S containing only 0-1 variables; Count the degree of deviation of each unit in each time period to form the relationship between the demand D and the deviation vector v; The SCUC model determined by the demand d∈D has a functional relationship So that: Among them, v j is each component of the deviation degree v of each unit in each time period.

4. The unit commitment accelerated optimization method based on machine learning to eliminate redundant constraints according to claim 3 is characterized in that: The minimum continuous start-stop time constraint is specifically:

5. A unit commitment accelerated optimization system based on machine learning to eliminate redundant constraints based on the unit commitment accelerated optimization method based on machine learning to eliminate redundant constraints according to claim 1, characterized in that: include: A first acquisition module, wherein the first acquisition module acquires semi-continuous constraints of upper and lower limits of output of the SCUC model unit and line power flow constraints; An introduction module, wherein if the semi-continuous constraints of the upper and lower limits of the unit output and the line power flow constraints in the SCUC model are both established, the upper and lower limits of the two constraints are very close, and then a sub-problem containing only 0-1 variables is introduced and solved; A second acquisition module, wherein the second acquisition module adds the solution of the sub-problem containing only 0-1 variables to the objective function of the original unit combination problem in a penalty manner, so that the solution of the sub-problem is close to the optimal solution of the original unit combination problem; An offset data acquisition module, which obtains the unit state variable offset of each unit in each time period based on the solution of the sub-problem and the optimal solution of the original unit combination problem; changes the characteristic demand D of the original unit combination problem, obtains multiple optimal solutions, obtains the relationship between the demand D and the offset, and learns it into a prediction regression model; A judgment module, wherein the judgment module is given a new unit combination problem, predicts the unit state variable offset of each unit in each time period of the new unit combination problem to be solved based on a learned prediction model, identifies and removes redundant constraints in the minimum continuous start-stop time constraints, so as to realize accelerated optimization of unit combination.

6. A terminal device, comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, characterized in that: When the processor executes the computer program, the steps of the method according to any one of claims 1 to 4 are implemented.

7. A computer-readable storage medium storing a computer program, characterized in that: When the computer program is executed by a processor, the steps of the method according to any one of claims 1 to 4 are implemented.

Images