A quantum computing method and system for unit commitment problem based on QUBO model

Abstract

The application relates to the technical field of power system optimization scheduling, in particular to a quantum computing method and system for a unit commitment problem based on a QUBO model. First, a unit commitment model considering power generation operation cost, start-stop cost, power balance, rotating reserve, upper and lower limits of unit output and climbing constraint is established; then the unit commitment model is converted into a QUBO Hamilton function by adopting an output expression mode of'minimum technical output + additional output', a controlled bit weight binary coding mode and a climbing constraint bit sharing mechanism; further, a standard QUBO matrix is extracted and mapped into an Ising matrix to input a quantum real machine or a quantum-like optimizer; finally, a plurality of groups of candidate solutions output are subjected to solution restoration, feasibility verification and cost comparison, so that an optimization scheduling scheme meeting the unit commitment operation constraint is obtained.

Description

Technical Field

[0001] This invention relates to the field of power system optimization scheduling technology, specifically a quantum computing method and system for the unit combination problem based on the QUBO model. Background Technology

[0002] With the large-scale integration of new energy sources and the increasing complexity of power system operation, the computational scale and difficulty of unit combination problems are constantly increasing. Traditional solution methods include dynamic programming, Lagrange relaxation, mixed integer programming, branch and bound, heuristic algorithms, simulated annealing, genetic algorithms, and commercial solvers. These methods can achieve good results on small- or medium-sized problems, but when the number of units, scheduling periods, and constraint types increase, they often face problems such as rapid growth in computation time, sensitivity to initial values ​​or parameters, and difficulty in quickly adapting to new computing hardware. The QUBO model is a type of optimization model that uses 0-1 variables as decision variables, a quadratic polynomial as the objective function, and does not explicitly include constraints. Its standard form can be expressed as: Since the QUBO model can be equivalently mapped to the Ising model, it can be processed by solution platforms such as simulated annealing, quantum annealing, coherent Ising machines, and optical quantum computers. If the constraints of start-up, shutdown, output, reserve, and ramp-up in the unit combination problem can be uniformly transformed into QUBO penalty terms, and the number of auxiliary variables and the numerical range of the Q matrix can be further reduced, then a new solution path adapted to quantum computing and quantum-like computing hardware can be provided for the unit combination problem.

[0003] However, the following problems still exist in the transformation of existing unit combination problems to QUBO models: First, the upper and lower limits of unit output are usually introduced with auxiliary variables, leading to a rapid increase in the number of qubits; second, if inequality constraints are expanded using ordinary binary relaxation variables, the constraint range may be unexpectedly expanded because the upper bound of the binary bit weights is greater than the actual constraint range; third, if ramp constraints are modeled separately for uphill and downhill ramps, auxiliary variables will be used repeatedly. Therefore, it is necessary to propose a compact QUBO modeling, matrix generation, and quantum solution method for unit combination problems. Summary of the Invention

[0004] The purpose of this invention is to provide a quantum computing method and system for the unit combination problem based on the QUBO model. This method first establishes a unit combination model considering power generation operating costs, start-up and shutdown costs, power balance, spinning reserve, upper and lower limits of unit output, and ramp constraints. Then, it transforms the unit combination model into a QUBO Hamiltonian function using an output expression method of "minimum technical output + additional output," a controlled bit weight binary encoding method, and a ramp constraint bit sharing mechanism. Furthermore, it extracts the standard QUBO matrix and maps it to the Ising matrix, inputting it into a quantum simulator or a quantum optimizer-like system. Finally, it performs solution restoration, feasibility verification, and cost comparison on multiple candidate solutions to obtain an optimized scheduling scheme that satisfies the unit combination operation constraints, thereby solving the problems mentioned in the background art.

[0005] To achieve the above objectives, the present invention provides the following technical solution: This invention relates to a quantum computing method for the unit combination problem based on the QUBO model, the method comprising the following steps: ① Construct a basic optimization model for unit combination; ② Controlled binary encoding is applied to the unit output variables, standby slack variables, and ramp slack variables; ③ Transform the objective function and constraints of the unit combination basic optimization model into a QUBO Hamiltonian function; ④ Extract the standard QUBO matrix from the QUBO Hamiltonian function and perform a mapping from QUBO to the Ising model; ⑤ Input the Ising matrix into a quantum computer, optical quantum computer, Ising machine, or quantum-like optimizer to obtain several candidate solutions; ⑥ Perform QUBO solution recovery, unit start-up and shutdown status and output plan restoration, physical constraint verification and optimal solution selection on candidate solutions.

[0006] In step ①, assume the system includes Taiwanese unit The scheduling period. Taiwanese unit in The start / stop status for each time period is as follows: ,in This indicates that the generator unit has been started. Indicates unit shutdown; Taiwanese unit in The active power output for each time period is ;No. The system load for each time period is ;No. The minimum technical output and maximum output of the unit are respectively , ;No. The limits for the uphill and downhill ramps of the unit are as follows: , ,in ;No. The operating cost coefficient of the Taiwanese unit is , , Start-up costs and downtime costs are respectively , .

[0007] The objective of the constructed unit combination model is to minimize the total system operating cost within the scheduling cycle. The objective function is expressed as follows: In the formula, the first term represents the unit operating cost, and the second term represents the unit start-up and shutdown cost. For scenarios that only consider hot starts or incorporate shutdown costs into start-up costs, we can set... Or merge it into In the case of linearization and cost simplification, it can be made .

[0008] The constraints of the constructed model include: (1) System power balance constraints In the formula, the total output of all online units in each time period should be equal to the system load demand in that time period.

[0009] (2) Rotational spare constraint In the formula, For the first The system's rotating standby requirements for a given time period can be set according to the system's safety margin, for example, taking... ,in This is a reserve ratio.

[0010] (3) Maximum and minimum output constraints of the unit This constraint ensures that the unit output is 0 when it is shut down, and that the unit output is between the minimum technical output and the maximum output when it is started up.

[0011] (4) Unit ramping constraints In the formula, This indicates the change in unit output between adjacent time periods.

[0012] (5) Minimum continuous operation and downtime constraints of the unit When the minimum continuous operation and downtime of the unit need to be considered, set and The first Minimum continuous operating time and minimum continuous shutdown time of the unit. This can be determined based on unit startup variables. and halting variables Establish the following constraints: The above constraints can be included in the QUBO penalty term depending on the specific case size.

[0013] In step ②, to adapt to the QUBO model, the unit output is written as a structure of "minimum technical output + additional output": in, For the additional output of the unit above the minimum technical output, the following conditions must be met: Will Represented as a weighted sum of binary variables: in, To add a power output position. Compared to ordinary... Unlike binary power expansion, this invention uses controlled bit weight encoding, where the first few bits are set according to binary powers and the last bit is set to the residual value, ensuring that the sum of all bit weights is strictly equal to the upper bound of the variable. For example, when the upper bound of an auxiliary variable is 13, the bit weights can be set to 1, 2, 4, 6 instead of 1, 2, 4, 8, thus avoiding the actual representable range of the variable being expanded from 13 to 15. In this way, the solver can be prevented from using redundant binary ranges to relax the original constraints.

[0014] because exist It may still be due to Non-zero output results in unreasonable output; therefore, this invention includes a shutdown-based additional output suppression penalty: When the unit is shut down, this penalty item forces all binary bits of the additional output to be 0; when the unit is started up, this penalty item automatically becomes invalid and does not affect the unit's output regulation.

[0015] In step ③, power balance, spinning reserve, and ramping constraints are converted into QUBO penalty terms.

[0016] (1) QUBO penalty term for power balance constraint in, This is the penalty coefficient for power balance constraints.

[0017] (2) QUBO penalty term for rotational spare constraint Transform the rotational spare inequality constraint into an equality constraint: In the formula, As a rotational spare relaxation variable, its upper bound can be set as follows: Will Controlled bit weight expansion is used: This leads to the rotational spare penalty term: (3) QUBO penalty term for ramp constraint This invention no longer sets two sets of auxiliary variables separately for uphill and downhill sections, but instead uses a shared uphill relaxation variable. The change in output between adjacent time periods is defined as follows: Original climbing constraints Equivalent to: Introducing shared hill relaxation variables : in, Will Controlled bit weight expansion is used: In this case, the last weight also takes the remaining value, such that: This leads to the hill-climbing constraint penalty term: This form uses a set of shared auxiliary variables to simultaneously express the uphill and downhill boundaries, reducing the number of auxiliary variables and qubits.

[0018] For constraints such as minimum continuous operation and minimum continuous shutdown, controlled position weight slack variables can be introduced in the same way, and corresponding penalty terms can be constructed. , .

[0019] Finally, the QUBO Hamiltonian function of the unit combination problem can be expressed as: In the formula, These are the weighting coefficients of the objective function, used to adjust the relative magnitudes between the original cost terms and the constraint penalty terms; each This is the penalty coefficient. For constraints that are not considered, the corresponding Hamiltonian function can be set to 0. Since... , , Both are linearly represented by 0-1 variables, and the objective function and constraint penalty terms are at most quadratic polynomials. Therefore... It can be organized into a standard QUBO format.

[0020] In step ④, all start / stop variables, additional output variables, reserve relaxation variables, and ramp relaxation variables are made into a unified QUBO binary variable vector: Will After unfolding, we get: Because the QUBO variables satisfy The linear terms can be incorporated into the diagonal of the Q matrix, resulting in: The Q matrix is ​​then symmetricized. The final standard QUBO model is obtained: When input is required for Ising type hardware, the QUBO variable will be used. Mapped to Ising spin variable ,in: This leads to the Ising matrix and bias terms.

[0021] For low-precision quantum hardware, matrix scaling, precision-fit splitting, and truncation of minimal elements can be further performed to reduce the impact of numerical range on the stability of hardware solutions.

[0022] In steps ⑤ and ⑥, the constructed QUBO matrix or Ising matrix is ​​input into a quantum machine, optical quantum computer, Ising machine, simulated annealer, or other quantum optimizer for solving, obtaining several candidate solutions. Since the output of a quantum machine typically contains multiple sets of candidate spin vectors, this invention performs the following verification process for each set of candidate solutions: (1) Restore the Ising spin solution output by the quantum machine to the QUBO binary solution; (2) Reconstruct the unit start-up and shutdown state matrix based on the variable mapping relationship. and unit output matrix ; (3) Check the power balance deviation, output upper and lower limit exceedance, rotational reserve deficit, ramp exceedance and other constraints that need to be considered in each time period; (4) Calculate operating costs, start-up and shutdown costs, and total costs; (5) Compare the costs of candidate solutions that meet the constraints and select the final unit combination optimization scheme. Through the above process, even if the quantum simulator outputs multiple sets of candidate solutions at once, it is possible to select feasible scheduling schemes that meet the physical constraints. In actual implementation, some of the quantum simulator's candidate solutions are completely consistent with the standard answers obtained by classical exact solvers such as Gurobi.

[0023] As a further preferred embodiment of the present invention, the present invention also provides a quantum computing method and system for the unit combination problem based on the QUBO model, comprising: Model building module: Constructs a basic optimization model for unit combination; Encoding module: Performs controlled binary encoding on unit output variables, standby slack variables, and ramp slack variables; Transformation module: Transforms the objective function and constraints of the unit combination basic optimization model into a QUBO Hamiltonian function; Extraction module: Extracts the standard QUBO matrix from the QUBO Hamiltonian function and performs a mapping from QUBO to the Ising model; Quantum computing module: Input the Ising matrix into a quantum machine, optical quantum computer, Ising machine or quantum optimizer to obtain several candidate solutions; The filtering module performs QUBO solution recovery, unit start-up and shutdown status and output plan restoration, physical constraint verification, and selection of the optimal solution for candidate solutions.

[0024] Compared with the prior art, the beneficial effects of the present invention are: (1) This invention transforms the unit combination problem into a QUBO matrix form, so that the mixed integer optimization problem, which originally contained multiple types of constraints, can be processed by quantum computers, optical quantum computers, Ising machines, simulated annealers and other types of quantum optimizers.

[0025] (2) The present invention adopts the unit output expression method of "minimum technical output + additional output" and avoids the problem of the unit still having output after shutdown by suppressing the additional output during shutdown penalty term. Compared with the method of constructing the upper and lower limits of output relaxation variables separately, the number of auxiliary variables is reduced.

[0026] (3) The present invention adopts a controlled bit weight binary encoding method, which sets the last binary bit weight as the remaining value, so that the maximum representable value of the auxiliary variable is strictly equal to the upper bound of the constraint, avoiding the unintended expansion of the constraint feasible region caused by ordinary binary expansion, and improving the accuracy of the QUBO model expression.

[0027] (4) The present invention rewrites the unit's uphill and downhill constraints into a shared uphill relaxation variable expression, and reduces the size of the QUBO variable through the bit sharing mechanism, thereby improving the efficiency of quantum bit utilization.

[0028] (5) This invention not only constructs the QUBO model, but also establishes a mechanism for restoring and verifying the quantum real machine candidate solution to the unit start-up and shutdown state and output plan. It can perform power balance, rotation reserve, output upper and lower limits and ramping constraint verification on multiple candidate solutions one by one, and compare them with classical standard solutions such as Gurobi, thereby improving the interpretability and credibility of quantum solution results.

[0029] (6) In the embodiments, the present invention can generate a standard QUBO matrix file and further convert it into an Ising matrix input to the quantum real machine; among the several candidate solutions output by the quantum real machine, some candidate solutions can be completely consistent with the standard scheduling results obtained by Gurobi after recovery and verification, indicating that the method can be effectively used for the quantum solution of the unit combination problem. Attached Figure Description

[0030] Figure 1 This is a flowchart of the unit combination optimization method based on the QUBO model of the present invention. Figure 2 This is the Hamiltonian evolution diagram calculated by the optical quantum computer in this invention. Detailed Implementation

[0031] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. 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 skilled in the art without creative effort are within the scope of protection of the present invention.

[0032] Example 1: A quantum computing method for the unit combinator problem based on the QUBO model, comprising the following steps: ① Construct a basic optimization model for unit combination; ② Controlled binary encoding is applied to the unit output variables, standby slack variables, and ramp slack variables; ③ Transform the objective function and constraints of the unit combination basic optimization model into a QUBO Hamiltonian function; ④ Extract the standard QUBO matrix from the QUBO Hamiltonian function and perform a mapping from QUBO to the Ising model; ⑤ Input the Ising matrix into a quantum computer, optical quantum computer, Ising machine, or quantum-like optimizer to obtain several candidate solutions; ⑥ Perform QUBO solution recovery, unit start-up and shutdown status and output plan restoration, physical constraint verification and optimal solution selection on candidate solutions.

[0033] In step ①, assume the system includes Taiwanese unit The scheduling period. Taiwanese unit in The start / stop status for each time period is as follows: ,in This indicates that the generator unit has been started. Indicates unit shutdown; Taiwanese unit in The active power output for each time period is ;No. The system load for each time period is ;No. The minimum technical output and maximum output of the unit are respectively , ;No. The limits for the uphill and downhill ramps of the unit are as follows: , ,in ;No. The operating cost coefficient of the Taiwanese unit is , , Start-up costs and downtime costs are respectively , .

[0034] The objective of the constructed unit combination model is to minimize the total system operating cost within the scheduling cycle. The objective function is expressed as follows: In the formula, the first term represents the unit operating cost, and the second term represents the unit start-up and shutdown cost. For scenarios that only consider hot starts or incorporate shutdown costs into start-up costs, we can set... Or merge it into In the case of linearization and cost simplification, it can be made .

[0035] The constraints of the constructed model include: (1) System power balance constraints In the formula, the total output of all online units in each time period should be equal to the system load demand in that time period.

[0036] (2) Rotational spare constraint In the formula, For the first The system's rotating standby requirements for a given time period can be set according to the system's safety margin, for example, taking... ,in This is a reserve ratio.

[0037] (3) Maximum and minimum output constraints of the unit This constraint ensures that the unit output is 0 when it is shut down, and that the unit output is between the minimum technical output and the maximum output when it is started up.

[0038] (4) Unit ramping constraints In the formula, This indicates the change in unit output between adjacent time periods.

[0039] (5) Minimum continuous operation and downtime constraints of the unit When the minimum continuous operation and downtime of the unit need to be considered, set and The first Minimum continuous operating time and minimum continuous shutdown time of the unit. This can be determined based on unit startup variables. and halting variables Establish the following constraints: The above constraints can be included in the QUBO penalty term depending on the specific case size.

[0040] In step ②, to adapt to the QUBO model, the unit output is written as a structure of "minimum technical output + additional output": in, For the additional output of the unit above the minimum technical output, the following conditions must be met: Will Represented as a weighted sum of binary variables: in, To add a power output position. Compared to ordinary... Unlike binary power expansion, this invention uses controlled bit weight encoding, where the first few bits are set according to binary powers and the last bit is set to the residual value, ensuring that the sum of all bit weights is strictly equal to the upper bound of the variable. For example, when the upper bound of an auxiliary variable is 13, the bit weights can be set to 1, 2, 4, 6 instead of 1, 2, 4, 8, thus avoiding the actual representable range of the variable being expanded from 13 to 15. In this way, the solver can be prevented from using redundant binary ranges to relax the original constraints.

[0041] because exist It may still be due to Non-zero output results in unreasonable output; therefore, this invention includes a shutdown-based additional output suppression penalty: When the unit is shut down, this penalty item forces all binary bits of the additional output to be 0; when the unit is started up, this penalty item automatically becomes invalid and does not affect the unit's output regulation.

[0042] In step ③, power balance, spinning reserve, and ramping constraints are converted into QUBO penalty terms.

[0043] (1) QUBO penalty term for power balance constraint in, This is the penalty coefficient for power balance constraints.

[0044] (2) QUBO penalty term for rotational spare constraint Transform the rotational spare inequality constraint into an equality constraint: In the formula, As a rotational spare relaxation variable, its upper bound can be set as follows: Will Controlled bit weight expansion is used: This leads to the rotational spare penalty term: (3) QUBO penalty term for ramp constraint This invention no longer sets two sets of auxiliary variables separately for uphill and downhill sections, but instead uses a shared uphill relaxation variable. The change in output between adjacent time periods is defined as follows: Original climbing constraints Equivalent to: Introducing shared hill relaxation variables : in, Will Controlled bit weight expansion is used: In this case, the last weight also takes the remaining value, such that: This leads to the hill-climbing constraint penalty term: This form uses a set of shared auxiliary variables to simultaneously express the uphill and downhill boundaries, reducing the number of auxiliary variables and qubits.

[0045] For constraints such as minimum continuous operation and minimum continuous shutdown, controlled position weight slack variables can be introduced in the same way, and corresponding penalty terms can be constructed. , .

[0046] Finally, the QUBO Hamiltonian function of the unit combination problem can be expressed as: In the formula, These are the weighting coefficients of the objective function, used to adjust the relative magnitudes between the original cost terms and the constraint penalty terms; each This is the penalty coefficient. For constraints that are not considered, the corresponding Hamiltonian function can be set to 0. Since... , , Both are linearly represented by 0-1 variables, and the objective function and constraint penalty terms are at most quadratic polynomials. Therefore... It can be organized into a standard QUBO format.

[0047] In step ④, all start / stop variables, additional output variables, reserve relaxation variables, and ramp relaxation variables are made into a unified QUBO binary variable vector: Will After unfolding, we get: Because the QUBO variables satisfy The linear terms can be incorporated into the diagonal of the Q matrix, resulting in: The Q matrix is ​​then symmetricized. The final standard QUBO model is obtained: When input is required for Ising type hardware, the QUBO variable will be used. Mapped to Ising spin variable ,in: This leads to the Ising matrix and bias terms.

[0048] For low-precision quantum hardware, matrix scaling, precision-fit splitting, and truncation of minimal elements can be further performed to reduce the impact of numerical range on the stability of hardware solutions.

[0049] In steps ⑤ and ⑥, the constructed QUBO matrix or Ising matrix is ​​input into a quantum machine, optical quantum computer, Ising machine, simulated annealer, or other quantum optimizer for solving, obtaining several candidate solutions. Since the output of a quantum machine typically contains multiple sets of candidate spin vectors, this invention performs the following verification process for each set of candidate solutions: (1) Restore the Ising spin solution output by the quantum machine to the QUBO binary solution; (2) Reconstruct the unit start-up and shutdown state matrix based on the variable mapping relationship. and unit output matrix ; (3) Check the power balance deviation, output upper and lower limit exceedance, rotational reserve deficit, ramp exceedance and other constraints that need to be considered in each time period; (4) Calculate operating costs, start-up and shutdown costs, and total costs; (5) Compare the costs of candidate solutions that meet the constraints and select the final unit combination optimization scheme. Through the above process, even if the quantum simulator outputs multiple sets of candidate solutions at once, it is possible to select feasible scheduling schemes that meet the physical constraints. In actual implementation, some of the quantum simulator's candidate solutions are completely consistent with the standard answers obtained by classical exact solvers such as Gurobi.

[0050] As a further embodiment of the present invention, the present invention also provides a quantum computing method and system for the unit combination problem based on the QUBO model, comprising: Model building module: Constructs a basic optimization model for unit combination; Encoding module: Performs controlled binary encoding on unit output variables, standby slack variables, and ramp slack variables; Transformation module: Transforms the objective function and constraints of the unit combination basic optimization model into a QUBO Hamiltonian function; Extraction module: Extracts the standard QUBO matrix from the QUBO Hamiltonian function and performs a mapping from QUBO to the Ising model; Quantum computing module: Input the Ising matrix into a quantum machine, optical quantum computer, Ising machine or quantum optimizer to obtain several candidate solutions; The filtering module performs QUBO solution recovery, unit start-up and shutdown status and output plan restoration, physical constraint verification, and selection of the optimal solution for candidate solutions.

[0051] Example 2: This embodiment uses a generator combinator problem with 4 generator units and 3 scheduling periods as an example to illustrate the QUBO modeling, matrix generation, quantum solution, and result verification process of this invention. It should be noted that this embodiment is only for explaining this invention and does not limit the scope of protection of this invention; this method can be extended to generator combinator problems with more generator units, longer scheduling cycles, and complex generator combinator problems involving minimum start / stop times and network flow constraints.

[0052] This embodiment sets the number of generator units. Number of scheduling periods The load for each time period is as follows: The coefficient of the quadratic term of operating cost is taken as follows The linear cost coefficient is: The fixed operating cost coefficient is: Startup costs and downtime costs are calculated as follows: The maximum and minimum technical output of the unit are respectively taken as: The unit's ramp-up limit is: The unit's downhill ramp limit is taken as follows: Rotational standby requirements The calculation shows that the right-hand side of the spare constraints for each time period is: In this embodiment, the unit output variable is constructed according to the "minimum technical output + additional output" structure described in this invention, and penalties for suppressing additional output during shutdown, power balance, spinning reserve, and shared ramp-up are set respectively. The penalty coefficients can be: The objective function weights can be chosen. This is used to improve the visibility of cost items in the QUBO objective. Therefore, we get: After performing quadratic term decomposition and linear term merging on the above QUBO Hamiltonian function, the standard QUBO matrix is ​​obtained. In the current embodiment, the derived QUBO matrix has the following dimensions. The matrix has a minimum element value of -8172, a maximum element value of 3510, and a range of 11682. The QUBO matrix can be saved as a CSV file for subsequent solving using quantum simulators or quantum-like optimizers.

[0053] This embodiment further maps the QUBO matrix to the Ising matrix and performs precision-adaptive decomposition to obtain Ising problem instances that can be input into a quantum simulator. After the quantum simulator or quantum-like optimizer outputs several sets of candidate spin solutions, they are recovered and verified in the following way: First, the Ising spin solutions are recovered as QUBO binary vectors; second, the linear mapping relationship between QUBO variables and unit start-up / shutdown states and unit output is used to recover the unit start-up / shutdown state table and output plan table; then, the power balance deviation, output upper and lower limit breaches, spinning reserve deficit, and ramp constraint breaches are calculated item by item for each time period; finally, the generation cost, start-up / shutdown cost, and total cost are calculated, and the costs of candidate solutions that meet the constraints are compared to obtain the final unit combination optimization scheme.

[0054] After constructing a unit combination model based on the above parameters and extracting the QUBO matrix, Gurobi was used as the classical standard solver to solve the same model. The QUBO matrix was then input into a quantum simulation machine for further solving. Experimental results show that among the multiple candidate solutions returned by the quantum simulation machine, a few candidate solutions, after reduction and physical constraint verification, completely match the scheduling results obtained by Gurobi in terms of unit start-up and shutdown states, unit output plans, and total costs. This demonstrates that the QUBO modeling method, controlled bit weight encoding method, bit sharing ramp constraint, and candidate solution verification process described in this invention can effectively support the quantum solution of the unit combination problem.

[0055] Table 1 shows the unit parameters in this embodiment.

[0056] Table 1 Unit Parameter Table Table 2 shows the load and reserve requirements in this embodiment.

[0057] Table 2 Load and Reserve Demand Table It should be noted that the time taken to transform and extract the unit combination problem into a QUBO matrix is ​​34.0 ms. The optical quantum computer performs calculations based on the extracted QUBO matrix.

[0058] Table 3 Table 3 Comparison of Time Consumption for Each Calculation Scheme Tables 4 and 5 show the optimal start-stop plan and power output plan obtained after optimization in this embodiment. Table 4 Table 4 Unit Start-up and Shutdown Planning Table Table 5 Table 5 Unit Output Planning Table The Hamiltonian evolution diagram calculated by the optical quantum computer is as follows: Figure 2 : The duration of a single evolution is 2.4422 ms. The classical computer using Gurobi takes 0.1466 seconds to solve the MILP problem and obtains the optimal solution of 316 elements. The classical computer using Gurobi takes 1.7438 seconds to solve the QUBO model. The time to derive the model into a QUBO matrix using quaddecomp is 0.0340 seconds. Based on this matrix, the computation time on the optical quantum computer is 2.4422 ms, and the optimal solution of 316 elements is obtained.

[0059] Compared to classical optimization algorithms, the computation speed of optical quantum computers is more than 60 times faster, verifying that this method can significantly improve the solution speed of unit combination optimization control problems. With the increase in the maximum number of superimposed qubits of optical quantum computers and the improvement and breakthrough of parallel computing capabilities, it will play an even greater role in solving massive distributed resource optimization control problems.

[0060] The innovative technical features of this invention in the above algorithm include: reducing the demand for auxiliary variables for the upper and lower limits of unit output constraints through the "minimum technical output + additional output" structure; avoiding the range of binary relaxation variables from exceeding the actual limit of the constraints through controlled bit weight encoding; realizing bit sharing of the upper and lower ramp constraints through shared ramp relaxation variables; and realizing the complete transformation of the unit combination problem from the classical mixed integer model to the quantum solvable model through QUBO matrix extraction, Ising mapping, quantum real machine candidate solution recovery and physical constraint verification process.

[0061] It will be apparent to those skilled in the art that the present invention is not limited to the details of the exemplary embodiments described above, and that the invention can be implemented in other specific forms without departing from its spirit or essential characteristics. Therefore, the embodiments should be considered in all respects as exemplary and non-limiting, and the scope of the invention is defined by the appended claims rather than the foregoing description. Thus, all variations falling within the meaning and scope of equivalents of the claims are intended to be included within the present invention. No reference numerals in the claims should be construed as limiting the scope of the claims.

[0062] Furthermore, it should be understood that although this specification describes embodiments, not every embodiment contains only one independent technical solution. This narrative style is merely for clarity. Those skilled in the art should consider the specification as a whole, and the technical solutions in each embodiment can also be appropriately combined to form other embodiments that can be understood by those skilled in the art.

Claims

1. A quantum computing method for the unit combination problem based on the QUBO model, characterized in that, Includes the following steps: ① Construct a basic optimization model for unit combination; ② Controlled binary encoding is applied to the unit output variables, standby slack variables, and ramp slack variables; ③ Transform the objective function and constraints of the unit combination basic optimization model into a QUBO Hamiltonian function; ④ Extract the standard QUBO matrix from the QUBO Hamiltonian function and perform a mapping from QUBO to the Ising model; ⑤ Input the Ising matrix into a quantum computer, optical quantum computer, Ising machine, or quantum-like optimizer to obtain several candidate solutions; ⑥ Perform QUBO solution recovery, unit start-up and shutdown status and output plan restoration, physical constraint verification and optimal solution selection on candidate solutions.

2. The quantum computing method for the unit combination problem based on the QUBO model according to claim 1, characterized in that, Step ① specifically includes: assuming the system contains... Taiwanese unit The scheduling period, the first Taiwanese unit in The start / stop status for each time period is as follows: ,in This indicates that the generator unit has been started. Indicates unit shutdown; Taiwanese unit in The active power output for each time period is ;No. The system load for each time period is ;No. The minimum technical output and maximum output of the unit are respectively , ;No. The limits for the uphill and downhill ramps of the unit are as follows: , ,in ;No. The operating cost coefficient of the Taiwanese unit is , , Start-up costs and downtime costs are respectively , ; The objective of the constructed unit combination model is to minimize the total system operating cost within the scheduling cycle. The objective function is expressed as follows: In the formula, the first term is the unit operating cost, and the second term is the unit start-up and shutdown cost. For scenarios that only consider hot start or include shutdown costs in the start-up cost, let Or merge it into In the case of linearization simplification cost scenarios, let ; The constraints of the constructed unit combination model include: (1) System power balance constraints In the formula, the total output of all online units in each time period is equal to the system load demand in that time period; (2) Rotational spare constraint In the formula, For the first The system's rotating standby requirements for each time period are set according to the system's safety margin. ,in This is a reserve ratio; (3) Maximum and minimum output constraints of the unit This constraint ensures that the unit output is 0 when the unit is shut down, and that the unit output is between the minimum technical output and the maximum output when the unit is started up. (4) Unit ramping constraints In the formula, This indicates the change in unit output between adjacent time periods; (5) Minimum continuous operation and downtime constraints of the unit When the minimum continuous operation and downtime of the unit need to be considered, set and The first The minimum continuous operating time and minimum continuous shutdown time of the unit are determined based on the unit startup variables. and halting variables Establish the following constraints: 。 3. The quantum computing method for the unit combination problem based on the QUBO model according to claim 2, characterized in that, In step ②, to adapt to the QUBO model, the unit output is written as a structure of "minimum technical output + additional output": in, For the additional output of the unit above the minimum technical output, the following conditions must be met: Will Represented as a weighted sum of binary variables: in, To add output weights, controlled weight encoding is used, where the first few bits are set according to binary powers, and the last bit is set to the residual value, so that the sum of all weights is strictly equal to the upper bound of the variable. Set an additional output suppression penalty for shutdown: When the unit is shut down, this penalty item forces all binary bits of the additional output to be 0; when the unit is started up, this penalty item automatically becomes invalid and does not affect the unit's output regulation.

4. The quantum computing method for the unit combination problem based on the QUBO model according to claim 1, characterized in that, In step ③, power balance, spinning reserve, and ramping constraints are respectively converted into QUBO penalty terms: (1) QUBO penalty term for power balance constraint in, This is the penalty coefficient for power balance constraints; (2) QUBO penalty term for rotational spare constraint Transform the rotational spare inequality constraint into an equality constraint: In the formula, As a rotational spare relaxation variable, its upper bound is set as follows: Will Controlled bit weight expansion: This leads to the rotational spare penalty term: in, The penalty coefficient for rotational spare constraint; (3) QUBO penalty term for ramp constraint Using shared ramp relaxation variables, the output change between adjacent time periods is defined as follows: Original climbing constraints Equivalent to: Introducing shared hill relaxation variables : in, Will Controlled bit weight expansion is used: In this case, the last weight also takes the remaining value, such that: This leads to the hill-climbing constraint penalty term: in, This represents the penalty coefficient for the hill-climbing constraint. For the minimum duration of operation and minimum duration of shutdown constraints, controlled position weight slack variables are introduced in the same way, and corresponding penalty terms are constructed. , ; Finally, the QUBO Hamiltonian function of the unit combination problem is expressed as: In the formula, These are the weighting coefficients of the objective function, used to adjust the relative magnitudes between the original cost term and the constraint penalty term; for constraint terms that are not considered, the corresponding Hamiltonian function is set to 0, and... , , Both are linearly represented by 0-1 variables, and the objective function and constraint penalty terms are at most quadratic polynomials. Therefore... Organize it into a standard QUBO format.

5. A quantum computing method for the unit combination problem based on the QUBO model according to claim 1, characterized in that, In step ④, all start / stop variables, additional output variables, reserve relaxation variables, and ramp relaxation variables are made into a unified QUBO binary variable vector: Will After unfolding, we get: Since the QUBO variable satisfies By incorporating the linear terms into the diagonal of matrix Q, we obtain: And the Q matrix is ​​symmetricized: The final standard QUBO model is obtained: ; When input is required for Ising type hardware, the QUBO variable will be used. Mapped to Ising spin variable ,in: This leads to the Ising matrix and bias terms; For low-precision quantum hardware, matrix scaling, precision-adaptive splitting, and truncation of minimal elements are performed to reduce the impact of numerical range on the stability of hardware solutions.

6. The quantum computing method for the unit combination problem based on the QUBO model according to claim 1, characterized in that, In step ⑥, the following verification process is performed on each group of candidate solutions: (1) Restore the Ising spin solution output by the quantum machine to the QUBO binary solution; (2) Reconstruct the unit start-up and shutdown state matrix based on the variable mapping relationship. and unit output matrix ; (3) Check the power balance deviation, output upper and lower limit exceedance, rotational reserve deficit, ramp exceedance and other constraints that need to be considered in each time period; (4) Calculate operating costs, start-up and shutdown costs, and total costs; (5) Compare the costs of candidate solutions that meet the constraints and select the final unit combination optimization scheme.

7. A system employing the quantum computing method for the unit combination problem based on the QUBO model as described in any one of claims 1-6, characterized in that, include: Model building module: Builds the basic optimization model for unit combination; Encoding module: Performs controlled binary encoding on unit output variables, standby slack variables, and ramp slack variables; Transformation module: Transforms the objective function and constraints of the unit combination basic optimization model into a QUBO Hamiltonian function; Extraction module: Extracts the standard QUBO matrix from the QUBO Hamiltonian function and performs a mapping from QUBO to the Ising model; Quantum computing module: Input the Ising matrix into a quantum machine, optical quantum computer, Ising machine or quantum optimizer to obtain several candidate solutions; The filtering module performs QUBO solution recovery, unit start-up and shutdown status and output plan restoration, physical constraint verification, and selection of the optimal solution for candidate solutions.

Images