Generator set scheduling method and system in electricity-carbon market

Abstract

The invention provides a generator set scheduling method and system in an electricity-carbon market, and the method comprises the steps: obtaining a market response model, a current decision, and a current adjustment range based on the data of the electricity-carbon market and a historical optimal reference point; based on the current decision and the current adjustment range, executing an accelerated solution step to obtain an optimal decision; based on the optimal decision, scheduling the target generator set to obtain a clearing result; the acceleration solving step comprises the following steps: based on a market response model, obtaining trust region sub-problems in a current decision and a current adjustment range, and then carrying out dimensionality reduction solving to obtain a decision adjustment step length; obtaining an update decision amount and an update adjustment range based on the decision adjustment step length; if the updating decision gradient vector is greater than the preset convergence threshold value, executing an accelerated solution step based on the updating decision and the updating adjustment range; otherwise, the accelerated solution is terminated, and the optimal decision is obtained. The scheduling efficiency of the generator set can be improved.

Description

TECHNICAL FIELD

[0001] The application belongs to the field of information technology, and particularly relates to a generator unit scheduling method and system under an electricity-carbon market. BACKGROUND

[0002] With the coordinated development of the electricity-carbon market, the generation cost structure of the traditional unit has changed significantly, which in turn has an impact on the bidding behavior of the electricity-carbon market. In order to maximize the revenue of the power generator in the electricity-carbon market, the traditional method uses a double-layer game model of dynamic carbon emission intensity to describe the strategy optimization problem of the electricity-carbon coupled market. However, the double-layer game model constructed by the traditional method has a large complexity in solving, which greatly reduces the calculation efficiency.

[0003] To solve the above problems, the prior art uses a genetic algorithm to obtain an upper decision model, and uses a particle swarm algorithm to obtain a lower model, so as to obtain an optimal decision including an optimal bid, an optimal generator unit output and an optimal carbon quota transaction amount, and schedule a target generator unit. However, the genetic algorithm and other intelligent optimization algorithms depend on parameter tuning, and it is difficult to guarantee the robustness of the solution. Therefore, the scheduling decision obtained has poor stability, and there may be a large difference between the decision results in different time periods, so that the generator unit cannot form a stable operating condition, and the scheduling efficiency of the generator unit is reduced. Moreover, the prior art cannot guarantee the convergence of the solution, so there may be a case that a long time is required to obtain the scheduling decision, which greatly reduces the scheduling efficiency of the generator unit. SUMMARY

[0004] The application aims to provide a generator unit scheduling method and system under an electricity-carbon market to solve the above technical problems and improve the scheduling efficiency of the generator unit.

[0005] To solve the above technical problems, the application provides a generator unit scheduling method under an electricity-carbon market, which comprises the following steps:

[0006] Obtain electricity-carbon market data and a historical optimal benchmark point, and based on the electricity-carbon market data and the historical optimal benchmark point, obtain a market response model, an initial decision and an initial adjustment range;

[0007] Take the initial decision as a current decision, and take the initial adjustment range as a current adjustment range; based on the current decision and the current adjustment range, perform an accelerated solving step to obtain an optimal decision;

[0008] Based on the optimal decision, obtain a generator unit scheduling strategy, and based on the generator unit scheduling strategy, schedule a target generator unit to obtain a clearing result of the electricity-carbon market;

[0009] The accelerated solving step comprises:

[0010] Based on the market response model, approximate processing is performed under the current decision and the current adjustment range to obtain a trust region sub-problem;

[0011] Based on the trust region sub-problem, a dimension reduction solving process is performed to obtain a decision adjustment step;

[0012] Based on the decision adjustment step and the current decision, an updated decision quantity and an updated adjustment range are obtained under a preset parameter updating strategy; and it is judged whether the gradient vector of the updated decision is greater than a preset convergence threshold;

[0013] If the gradient vector of the updated decision is greater than the preset convergence threshold, the updated decision is taken as the current decision, the updated adjustment range is taken as the current adjustment range, and the accelerating solving step is re-executed based on the current decision and the current adjustment range; otherwise, the accelerating solving step is terminated, and the updated decision quantity is taken as the optimal decision.

[0014] The above scheme uses a market response model to approximate the mapping relationship from the upper-level bidding decision to the lower-level clearing result. Compared with the traditional method of obtaining a scheduling strategy by using a double-layer game model, the market response model has a lower computational complexity, so the time required for the subsequent process of obtaining a scheduling decision based on the market response model is less, which greatly improves the scheduling efficiency of the generator unit. On the basis of the market response model, the current decision and the current adjustment range are used to obtain a dynamic trust region sub-problem, and then the dynamic control of the decision adjustment step is realized, so that the solving process is always performed within an acceptable adjustment range, ensuring the convergence of the decision adjustment step obtained by solving, effectively avoiding the decision deviation caused by model mismatch or search blindness in the traditional method. Therefore, the differences between the scheduling decisions obtained at different times by the scheme are within the acceptable adjustment range, so that the scheduling decision of the generator unit based on the scheme can form a stable operating condition, and the situation of scheduling lag caused by sudden changes in operating conditions is avoided, thereby improving the scheduling efficiency of the generator unit.

[0015] Further, the electric carbon market data and the historical optimal benchmark point are obtained, and based on the electric carbon market data and the historical optimal benchmark point, a market response model, an initial decision and an initial adjustment range are obtained, which includes: obtaining electric carbon market data, and based on the electric carbon market data, performing electric carbon coupling double-layer game modeling processing to obtain a double-layer game model; wherein the double-layer game model includes a lower-layer market clearing model; obtaining a historical optimal benchmark point, and based on the historical optimal benchmark point and the lower-layer market clearing model, obtaining a market response model; based on the historical optimal benchmark point, an initial decision and an initial adjustment range are obtained under a preset electric carbon decision weight.

[0016] Furthermore, the step of obtaining the historical best benchmark point and, based on the historical best benchmark point and the lower-level market clearing model, obtaining the market response model includes: constructing an implicit response function under KKT conditions based on the Lagrangian function of the lower-level market clearing model; obtaining the historical best benchmark point and, based on the implicit response function, performing a second Taylor expansion on the historical best benchmark point to obtain the market response model.

[0017] The above-described scheme replaces the lower-level market clearing model in the traditional two-level game model with a market response model. This avoids the problem of difficulty in solving the problem caused by the complex interaction between the upper-level decision model and the lower-level market clearing model in traditional methods, reduces the time required to obtain generator scheduling strategies, and thus improves the efficiency of generator scheduling. Furthermore, in subsequent processes, this scheme constructs a trust region subproblem based on a simplified market response model of the upper and lower level models to ensure the accuracy of the obtained decision results. This prevents the simplification of the upper and lower level relationships in the two-level game model from causing a decrease in accuracy.

[0018] Furthermore, the two-level game model also includes an upper-level decision model. The approximation process based on the market response model under the current decision and the current adjustment range to obtain the trust region sub-problem includes: obtaining an approximate objective function under the objective function of the upper-level decision model based on the market response model and the current decision; performing a second Taylor expansion based on the approximate objective function to obtain a second approximate model; and obtaining the trust region sub-problem under the current adjustment range based on the second approximate model.

[0019] Further, the step of performing dimensionality reduction based on the trust domain subproblem to obtain the decision adjustment step size includes: performing a dimensionality reduction step based on the trust domain subproblem to obtain the decision adjustment step size; the dimensionality reduction step includes: performing dimensionality reduction projection on the trust domain subproblem using the Lanczos algorithm to obtain a low-dimensional trust domain subproblem; performing a solution process on the low-dimensional trust domain subproblem using the More-Sorensen exact algorithm to obtain a low-dimensional step size; performing high-dimensional recovery processing on the low-dimensional step size to obtain a high-dimensional step size; and determining whether the high-dimensional step size meets the preset solution requirements; if the high-dimensional step size does not meet the preset solution requirements, then increasing the subspace dimension parameter of the Lanczos algorithm according to a preset space expansion threshold, and re-executing the dimensionality reduction step; otherwise, terminating the dimensionality reduction step, and using the high-dimensional step size as the decision adjustment step size.

[0020] The above scheme reduces the dimensionality of the trust domain subproblem, thus decreasing the complexity of solving the obtained trust domain subproblem in the complex environment of the electricity carbon market. This allows the scheme to obtain decision results more quickly, thereby reducing the time required to obtain generator scheduling strategies and improving the efficiency of generator scheduling. Furthermore, after obtaining the high-dimensional step size, the scheme further restricts and adjusts it to ensure that the obtained decision adjustment step size is within a reasonable range. This guarantees that the generator scheduling strategy obtained at each time step will not lead to unstable generator operating conditions and slow scheduling response due to sudden changes in the decision adjustment step size, nor will it cause a delay in the electricity carbon market situation and generator scheduling due to an excessively small decision adjustment step size.

[0021] Furthermore, the step of obtaining the updated decision and the updated adjustment range based on the decision adjustment step size and the current decision, under a preset parameter update strategy, includes: obtaining the model approximation based on the decision adjustment step size and the current decision; and obtaining the updated decision and the updated adjustment range based on the decision adjustment step size and the model approximation, under a preset decision and adjustment radius update strategy.

[0022] Further, the step of obtaining the current decision quantity and the updated adjustment range based on the decision adjustment step size and the model approximation, under a preset decision and trust region radius update strategy, includes: determining whether the model approximation is less than a preset first threshold; if the model approximation is less than the preset first threshold, then performing radius reduction processing based on the current decision and the current adjustment range to obtain an updated decision and an updated adjustment range; otherwise, determining whether the model approximation is less than a preset second threshold; if the model approximation is less than the preset second threshold, then performing decision update processing based on the current decision and the current adjustment range to obtain an updated decision and an updated adjustment range; otherwise, performing radius expansion and decision update processing based on the current decision and the current adjustment range to obtain an updated decision and an updated adjustment range.

[0023] In the above scheme, if the model approximation is less than a preset first threshold, it indicates that the current adjustment range is too large. Therefore, the radius of the current adjustment range is reduced, and the current decision is directly used as the updated decision. Otherwise, if the model approximation is less than a preset second threshold, the decision adjustment step size is added to the current decision as the updated decision, and the current adjustment range is used as the updated adjustment range. If the model approximation is greater than the preset second threshold, it indicates that the current adjustment range is too small. Therefore, the current adjustment range is increased as the updated adjustment range, and the decision adjustment step size is added to the current decision as the updated decision. The above scheme can achieve dynamic adjustment of the decision and the adjustment range, ensuring that when a new generator scheduling decision is obtained in the next time period, the difference between the obtained scheduling decision and the previous time period is within a suitable range. This allows the generators to maintain stable operating conditions based on the scheduling decision and to respond and adjust quickly and in real time based on the electricity market situation, improving the efficiency of generator scheduling in the electricity carbon market.

[0024] This invention also provides a generator scheduling system under an electricity carbon market, comprising: a market response model acquisition module, used to acquire electricity carbon market data and historical optimal benchmarks, and based on the electricity carbon market data and the historical optimal benchmarks, obtain a market response model, an initial decision, and an initial adjustment range; a decision acceleration solution module, used to take the initial decision obtained by the market response model acquisition module as the current decision, and take the initial adjustment range obtained by the market response model acquisition module as the current adjustment range; and based on the current decision and the current adjustment range, perform an acceleration solution step to obtain an optimal decision; and a generator scheduling module, used to obtain a generator scheduling strategy based on the optimal decision obtained by the decision acceleration solution module, and to schedule the target generator based on the generator scheduling strategy. The process involves scheduling to obtain the clearing result of the carbon electricity market. The accelerated solution step includes: based on the market response model, performing approximation processing under the current decision and current adjustment range to obtain a trust region subproblem; based on the trust region subproblem, performing dimensionality reduction processing to obtain the decision adjustment step size; based on the decision adjustment step size and the current decision, obtaining the updated decision quantity and the updated adjustment range under a preset parameter update strategy; and determining whether the gradient vector of the updated decision is greater than a preset convergence threshold; if the gradient vector of the updated decision is greater than the preset convergence threshold, then the updated decision is taken as the current decision, the updated adjustment range is taken as the current adjustment range, and the accelerated solution step is re-executed based on the current decision and the current adjustment range; otherwise, the accelerated solution step is terminated, and the updated decision quantity is taken as the optimal decision.

[0025] Furthermore, the step of acquiring electricity carbon market data and historical optimal benchmarks, and obtaining a market response model, initial decision, and initial adjustment range based on the electricity carbon market data and the historical optimal benchmarks, includes: acquiring electricity carbon market data, and performing electricity carbon coupled two-level game modeling processing based on the electricity carbon market data to obtain a two-level game model; wherein the two-level game model includes a lower-level market clearing model; acquiring historical optimal benchmarks, and obtaining a market response model based on the historical optimal benchmarks and the lower-level market clearing model; and obtaining an initial decision and initial adjustment range based on the historical optimal benchmarks and under preset electricity carbon decision weights.

[0026] Furthermore, the step of obtaining the historical best benchmark point and, based on the historical best benchmark point and the lower-level market clearing model, obtaining the market response model includes: constructing an implicit response function under KKT conditions based on the Lagrangian function of the lower-level market clearing model; obtaining the historical best benchmark point and, based on the implicit response function, performing a second Taylor expansion on the historical best benchmark point to obtain the market response model.

[0027] The above scheme uses a market response model to approximate the mapping relationship between upper-level bidding decisions and lower-level clearing results, simplifying the mapping relationship between the upper and lower levels. This reduces the time required to obtain subsequent scheduling decisions based on the market response model, thus improving the scheduling efficiency of generator units. Furthermore, this scheme uses the current decision and the current adjustment range to obtain a dynamic trust region sub-problem, thereby achieving dynamic control of the decision adjustment step size. This effectively avoids decision biases caused by model mismatch or blind search in traditional methods. Therefore, the differences between scheduling decisions obtained by this scheme at different time periods are all within an acceptable adjustment range, enabling generator units to form stable operating conditions based on the scheduling decisions of this scheme. This prevents scheduling lags caused by sudden changes in operating conditions, further improving the scheduling efficiency of generator units. Attached Figure Description

[0028] Figure 1 A flowchart illustrating the steps of a generator set scheduling method under an electric carbon market, as provided in this embodiment of the invention;

[0029] Figure 2 This is a schematic diagram of the structure for generator set scheduling under an electric carbon market, provided as an embodiment of the present invention. Detailed Implementation

[0030] 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.

[0031] Please see Figure 1 This embodiment provides a generator set dispatching method under the electricity carbon market, including the following steps:

[0032] Step S1: Obtain electricity carbon market data and historical best benchmarks, and based on the electricity carbon market data and historical best benchmarks, obtain the market response model, initial decision and initial adjustment range;

[0033] Step S2: Take the initial decision as the current decision and the initial adjustment range as the current adjustment range; based on the current decision and the current adjustment range, execute the accelerated solution step to obtain the optimal decision;

[0034] Step S3: Based on the optimal decision, obtain the generator set scheduling strategy, and schedule the target generator set according to the generator set scheduling strategy to obtain the clearing result of the electricity carbon market;

[0035] The accelerated solution steps include:

[0036] Based on the market response model, an approximation is performed under the current decision and adjustment range to obtain the trust domain sub-problem;

[0037] Based on the aforementioned trust domain subproblem, a dimensionality reduction solution is performed to obtain the decision adjustment step size;

[0038] Based on the decision adjustment step size and the current decision, under the preset parameter update strategy, the updated decision amount and the update adjustment range are obtained; and it is determined whether the gradient vector of the updated decision is greater than the preset convergence threshold.

[0039] If the gradient vector of the updated decision is greater than the preset convergence threshold, the updated decision is taken as the current decision, the updated adjustment range is taken as the current adjustment range, and the accelerated solution step is re-executed based on the current decision and the current adjustment range; otherwise, the accelerated solution step is terminated, and the updated decision is taken as the optimal decision.

[0040] The above embodiments utilize a market response model to approximate the mapping relationship from upper-level bidding decisions to lower-level clearing results. Compared to traditional methods that use a two-level game model to obtain scheduling strategies, the market response model in this embodiment has lower computational complexity. Therefore, the time required for subsequent scheduling decisions based on the market response model is reduced, significantly improving the scheduling efficiency of generator sets. Based on the market response model, this embodiment uses the current decision and the current adjustment range to obtain a dynamic trust region subproblem, thereby achieving dynamic control of the decision adjustment step size. This ensures that the solution process always operates within an acceptable adjustment range, guaranteeing the convergence of the obtained decision adjustment step size. This effectively avoids decision biases caused by model mismatch or blind search in traditional methods. Therefore, the differences between scheduling decisions obtained in this embodiment at different time periods are all within an acceptable adjustment range, enabling generator sets to form stable operating conditions based on the scheduling decisions of this embodiment. This prevents scheduling lags due to sudden changes in operating conditions, thus improving the scheduling efficiency of generator sets.

[0041] Furthermore, the step of acquiring electricity carbon market data and historical optimal benchmarks, and obtaining a market response model, initial decision, and initial adjustment range based on the electricity carbon market data and the historical optimal benchmarks, includes: acquiring electricity carbon market data, and performing electricity carbon coupled two-level game modeling processing based on the electricity carbon market data to obtain a two-level game model; wherein the two-level game model includes a lower-level market clearing model; acquiring historical optimal benchmarks, and obtaining a market response model based on the historical optimal benchmarks and the lower-level market clearing model; and obtaining an initial decision and initial adjustment range based on the historical optimal benchmarks and under preset electricity carbon decision weights.

[0042] Furthermore, the step of obtaining the historical best benchmark point and, based on the historical best benchmark point and the lower-level market clearing model, obtaining the market response model includes: constructing an implicit response function under KKT conditions based on the Lagrangian function of the lower-level market clearing model; obtaining the historical best benchmark point and, based on the implicit response function, performing a second Taylor expansion on the historical best benchmark point to obtain the market response model.

[0043] The above embodiments, by constructing a market response model to replace the lower-level market clearing model in the traditional two-level game model, avoid the problem of difficulty in solving the problem caused by the complex interaction between the upper-level decision model and the lower-level market clearing model in traditional methods. This reduces the time required to obtain generator scheduling strategies and thus improves the efficiency of generator scheduling. Furthermore, in subsequent processes, this embodiment constructs a trust region subproblem based on a simplified market response model of the upper and lower level models to solve the problem under the current decision. This ensures that the obtained decision results have good accuracy and does not lead to a decrease in accuracy due to the simplification of the upper and lower level relationships in the two-level game model.

[0044] In one embodiment, based on carbon market data, a higher-level decision-making model is constructed with the goal of maximizing overall profit. The objective function of this higher-level decision-making model is:

[0045]

[0046] in, For the profit of power generators in the electricity market, For the revenue of power generator g in the carbon market, Let g be the cost of generator g in the carbon market. The constraints of this upper-level decision-making model include the two-market profit equation constraint, generator behavior constraint, and generator carbon quota calculation and scope constraint. The two-market profit equation constraint is as follows:

[0047]

[0048] in, Let n be the marginal electricity price at time t. Let g be the set of generators on node n. To reduce the cost of segmented power generation, Let λ be the output power of generator unit i in bidding block v during time period t. c To clear prices in the carbon market. and These represent the carbon allowances sold and bought, respectively. h is the block containing carbon allowances sold, and k is the block containing carbon allowances bought. and These represent the carbon quota supply and demand capacity of the generator sets, λ FS and λ FB These refer to the trading prices of carbon allowances under the carbon market stabilization reserve mechanism.

[0049] In the above embodiments, the generator behavior constraints indirectly affect the carbon market by controlling the buying and selling prices in the bidding blocks. and To ensure that the daily net surplus carbon allowance is met: if the selling price in the carbon market in the bidding block... Too low, If the daily net surplus carbon allowance is too large, it may become negative, requiring the purchase of additional carbon allowances, thus decreasing the objective function value; if the purchase price in the carbon market within the bidding block... Too high, then If the value is too large, the daily net surplus carbon allowance may be positive, resulting in wasted capital costs and a decrease in the objective function value. The constraints on the generator behavior are as follows:

[0050]

[0051] in, This refers to the purchase price in the carbon market within the bidding blocks. The selling price in the carbon market for the bidding blocks; and These represent the upper and lower limits corresponding to the buying and selling prices, respectively. The bid price for generator set i in bidding block v; and These represent the upper and lower limits of the unit's price quote.

[0052] In the above embodiments, the carbon quota calculation and scope constraints for power generators are as follows:

[0053]

[0054] in, The daily net surplus carbon allowance reflects the difference between the carbon emission allowances held by a power generator on a given day and its actual carbon emissions. M is the maximum trading capacity of carbon allowances for generator unit i, and u i Let u be the binary variable representing the carbon trading direction of generator set i, and u i =1 indicates that generator unit i mainly sells carbon quotas in the current period, u i =0 indicates that generator set i was mainly purchased in the current period. I is the baseline value for the carbon emission intensity of generator set i. i，t Let be the actual carbon emission intensity of generator unit i during time period t, and In the formula a i b is the coefficient of the carbon emission intensity constant term. i This is the linear coefficient for carbon emission intensity.

[0055] In one embodiment, the lower-level market clearing model constructed based on electricity and carbon market data includes an electricity market clearing model and a carbon market clearing model. The electricity market clearing model aims to minimize the total dispatch cost of the electricity market, and its specific objective function is: Furthermore, the constraints of the electricity market clearing model include power balance constraints, generator output constraints, total output power constraints, generator ramp capacity constraints, line transmission power constraints, node phase angle constraints, and reference node constraints.

[0056] In the above embodiments, the power balance constraint is:

[0057]

[0058] in, This represents the total output power of the generator. Let n be the set of generator sets on node n; Let n be the set of all its neighboring nodes and the node itself. B represents the load demand of node n during time period t.n，m The susceptance between nodes n and m; δ n，t δ is the voltage phase angle of node n during time period t; m，t Let be the voltage phase angle of node m during time interval t; These are the Lagrange multipliers associated with this constraint; The relevant Lagrangian function is constructed in the form of "multiplier × constraint deviation term": when the constraint is satisfied, the multiplier does not affect the objective function; when the constraint is not satisfied, the multiplier will be adjusted through the objective function, forcing the optimization result to return to the constraint feasible region.

[0059] In the above embodiment, the generator output constraint is:

[0060]

[0061] in, Let i be the minimum generating power of generator set i in the price range v; This represents the maximum generating power of generator set i in price segment v. and These are the Lagrange multipliers related to the generator output constraints; and The relevant Lagrangian function is constructed in the form of "multiplier × constraint deviation term": when the constraint is satisfied, the multiplier does not affect the objective function; when the constraint is not satisfied, the multiplier will be adjusted through the objective function, forcing the optimization result to return to the constraint feasible region.

[0062] In the above embodiments, the total output power constraint is:

[0063]

[0064] in, and This represents the upper and lower limits of the total output power of generator set i during time period t; and These are the dual variables related to the total output power constraint; the dual variables reflect the tightness of the constraint and the marginal effect, and their influence logic is consistent with that of the Lagrange multiplier.

[0065] In the above embodiment, the generator ramp capacity constraint is:

[0066]

[0067] in, This is the upper limit of the power descent rate. This is the upper limit of the power rise rate. and These are the relevant dual variables.

[0068] In the above embodiments, the line transmission power constraint is:

[0069]

[0070] Among them, B n，m The susceptance between nodes n and m, Let (n, m) be the maximum transmission capacity of the line. These are the dual variables related to the line transmission power constraint.

[0071] In the above embodiment, the node phase angle constraint is:

[0072]

[0073] Where, δ n，t Let n be the voltage phase angle at node n during time interval t. and These are the dual variables related to the nodal phase angle constraint.

[0074] In the above embodiments, the reference node constraint is:

[0075]

[0076] Where, δ 1,t The phase angle value is the reference node. Constrain the relevant dual variables for the reference node.

[0077] In the above embodiments, the carbon market clearing model constructs an objective function based on maximizing the social welfare of carbon market clearing, and its objective function is: and This represents the amount of carbon emission allowances purchased / sold in the carbon market under bidding blocks k / h. The constraints of the carbon market clearing model include carbon emission balance equation constraints, upper and lower limits on carbon emission trading volume, and upper and lower limits on total carbon emission capacity.

[0078] In the above embodiments, the carbon emission balance equation constraint is:

[0079]

[0080] Where, λ C These are the Lagrange multipliers associated with the carbon emission balance equation constraints.

[0081] In the above embodiments, the upper and lower limits of carbon emission trading volume are constrained as follows:

[0082]

[0083] in, and The upper and lower limits of carbon emission trading volume are related to the amount of carbon emission allowances sold in the carbon market under bidding block h. Related dual variables, and The upper and lower limits of carbon emission trading volume are the amount of carbon emission allowances purchased in the carbon market under bidding block k. Related dual variables.

[0084] In the above embodiments, the upper and lower limits of the total carbon emission capacity are constrained as follows:

[0085]

[0086] in, and The dual variables are related to the upper and lower limits of the total carbon emission capacity constraints.

[0087] It should be noted that the Lagrange multipliers related to each constraint in the above embodiments are essentially the carbon market clearing prices corresponding to each constraint, and the carbon market clearing prices are the marginal prices at which the supply and demand of carbon quotas are balanced.

[0088] In one embodiment, the Lagrangian function of the lower-level market clearing model is:

[0089]

[0090] Where, λ C This indicates the carbon market clearing price; This represents the amount of carbon allowance sold by generator g in block b during period t when there is a carbon allowance surplus. This represents the amount of carbon allowance purchased by generator g in block b during period t when there is a carbon allowance shortage; L n，t P represents the total load demand of node n during time period t; n，m，t λ represents the transmission power of line (n, m) during time interval t; P λ represents the Lagrange multiplier for the power balance constraint in the electricity market, i.e., the nodal marginal price; μ is the logarithmic barrier function parameter, used to control the degree of relaxation of the inequality constraint; λ E The Lagrange multiplier representing the carbon quota balance constraint; α i denoted as the dynamic carbon emission intensity coefficient of the i-th unit.

[0091] In the above embodiment, by taking the partial derivative of the Lagrangian function of the lower-level market clearing model and setting it equal to zero, the implicit response function obtained under the KKT conditions is: and S represents the implicit function vector formed by the KKT conditions. t This represents the system state vector during time period t. Let represent the decision variable vector of the upper-level generator in time period t. Let λ represent the output vector of the response function, and λ represent the Lagrange multiplier vector. This embodiment sets the implicit response function at the historical best baseline. A second Taylor expansion yields the market response model as follows: and and These are the state offset and the decision adjustment, S, respectively. t This represents the current system state. For the current decision of the upper-level decision-making model, and These are the coefficient matrices for the state and the decision, respectively, both of which are calculated from the inverse of the Jacobian matrix.

[0092] Furthermore, the two-level game model also includes an upper-level decision model. The approximation process based on the market response model under the current decision and the current adjustment range to obtain the trust region sub-problem includes: obtaining an approximate objective function under the objective function of the upper-level decision model based on the market response model and the current decision; performing a second Taylor expansion based on the approximate objective function to obtain a second approximate model; and obtaining the trust region sub-problem under the current adjustment range based on the second approximate model.

[0093] Further, the step of performing dimensionality reduction based on the trust domain subproblem to obtain the decision adjustment step size includes: performing a dimensionality reduction step based on the trust domain subproblem to obtain the decision adjustment step size; the dimensionality reduction step includes: performing dimensionality reduction projection on the trust domain subproblem using the Lanczos algorithm to obtain a low-dimensional trust domain subproblem; performing a solution process on the low-dimensional trust domain subproblem using the More-Sorensen exact algorithm to obtain a low-dimensional step size; performing high-dimensional recovery processing on the low-dimensional step size to obtain a high-dimensional step size; and determining whether the high-dimensional step size meets the preset solution requirements; if the high-dimensional step size does not meet the preset solution requirements, then increasing the subspace dimension parameter of the Lanczos algorithm according to a preset space expansion threshold, and re-executing the dimensionality reduction step; otherwise, terminating the dimensionality reduction step, and using the high-dimensional step size as the decision adjustment step size.

[0094] The above embodiment solves the trust domain subproblem by reducing its dimensionality, which reduces the complexity of solving the obtained trust domain subproblem in the complex environment of the electricity carbon market. This allows the embodiment to obtain decision results more quickly, thereby reducing the time required to obtain generator scheduling strategies and improving the efficiency of generator scheduling. Furthermore, after obtaining the high-dimensional step size, this embodiment further restricts and adjusts the high-dimensional step size to ensure that the obtained decision adjustment step size is within a reasonable range. This guarantees that the generator scheduling strategy obtained at each time will not lead to unstable generator operating conditions and slow scheduling response due to sudden changes in the decision adjustment step size, nor will it cause a delay in the electricity carbon market situation and generator scheduling due to an excessively small decision adjustment step size.

[0095] In one embodiment, the market response model is used. Substituting into the objective function of the upper-level decision model, we obtain the quadratic approximation model. Within the current adjustment range, the trust domain subproblem is obtained. Under the Lanczos algorithm, the trust domain subproblem is dimensionality-reduced and projected onto a low-dimensional Krylov subspace. The dimension of the Krylov subspace is determined by the subspace dimension parameter of the Lanczos algorithm. After obtaining the low-dimensional trust domain subproblem through dimensionality reduction projection, this embodiment uses the More-Sorensen exact algorithm for solution processing to obtain a low-dimensional step size. However, the low-dimensional step size cannot be directly applied to the adjustment of the decision. Therefore, this embodiment restores the low-dimensional step size to a high-dimensional step size. If the high-dimensional step size does not meet the solution requirements, the subspace dimension parameter of the Lanczos algorithm is increased under a preset space expansion threshold to re-solve the dimensionality reduction problem, obtaining a low-dimensional trust domain subproblem with a higher dimension than the previous dimensionality reduction solution step.

[0096] Furthermore, the step of obtaining the updated decision and the updated adjustment range based on the decision adjustment step size and the current decision, under a preset parameter update strategy, includes: obtaining the model approximation based on the decision adjustment step size and the current decision; and obtaining the updated decision and the updated adjustment range based on the decision adjustment step size and the model approximation, under a preset decision and adjustment radius update strategy.

[0097] Further, the step of obtaining the current decision quantity and the updated adjustment range based on the decision adjustment step size and the model approximation, under a preset decision and trust region radius update strategy, includes: determining whether the model approximation is less than a preset first threshold; if the model approximation is less than the preset first threshold, then performing radius reduction processing based on the current decision and the current adjustment range to obtain an updated decision and an updated adjustment range; otherwise, determining whether the model approximation is less than a preset second threshold; if the model approximation is less than the preset second threshold, then performing decision update processing based on the current decision and the current adjustment range to obtain an updated decision and an updated adjustment range; otherwise, performing radius expansion and decision update processing based on the current decision and the current adjustment range to obtain an updated decision and an updated adjustment range.

[0098] In the above embodiments, if the model approximation is less than a preset first threshold, it indicates that the current adjustment range is too large. Therefore, the radius of the current adjustment range is reduced, and the current decision is directly used as the update decision. Otherwise, if the model approximation is less than a preset second threshold, the decision adjustment step size is added to the current decision as the update decision, and the current adjustment range is used as the update adjustment range. If the model approximation is greater than the preset second threshold, it indicates that the current adjustment range is too small. Therefore, the current adjustment range is increased as the update adjustment range, and the decision adjustment step size is added to the current decision as the update decision. The above embodiments can realize dynamic adjustment of decisions and adjustment ranges, ensuring that when a new generator scheduling decision is obtained in the next time period, the difference between the obtained scheduling decision and the previous time period is within a suitable range. This allows generators to maintain stable operating conditions based on scheduling decisions and can also respond and adjust quickly and in real time based on the electricity market situation, improving the efficiency of generator scheduling in the electricity carbon market.

[0099] In one embodiment, the model approximation is obtained based on the decision adjustment step size and the current decision. The formula for calculating the model approximation is as follows: Where, ω E ω represents the power decision weights within the current adjustment range. C This represents the carbon decision weights within the current adjustment range.

[0100] In one embodiment, if the model approximation is less than a preset first threshold, it indicates that the current adjustment range is too large. Therefore, the radius of the current adjustment range is reduced to obtain an updated decision and an updated adjustment range. At this time, the current decision is directly used as the updated decision. Otherwise, if the model approximation is less than a preset second threshold, a decision update process is performed based on the current decision and the current adjustment range to obtain an updated decision and an updated adjustment range. At this time, the decision adjustment step size is superimposed on the current decision as the updated decision. Since the current adjustment range is within the appropriate range, the current adjustment range will not be further adjusted, and the current adjustment range is directly used as the updated adjustment range. If the model approximation is greater than the preset second threshold, it indicates that the current adjustment range is too small. In order to enable the decision obtained in the next period to be adjusted in real time and quickly based on the electricity market situation, a radius expansion and decision update process are performed to obtain an updated decision amount and an updated adjustment range. At this time, the current adjustment range is increased to be used as the updated adjustment range, and the decision adjustment step size is superimposed on the current decision as the updated decision.

[0101] The dynamic adjustment process of decision-making and adjustment scope in the above embodiments enables the generator set dispatch decision to remain unchanged when a new generator set dispatch decision is obtained in the next time period. This ensures that the generator set can maintain a stable operating condition based on the dispatch decision, and also prevents the dispatch decision from changing too little, enabling the generator set to adjust quickly and in real time based on the electricity market situation.

[0102] Please see Figure 2This embodiment also provides a generator scheduling system under an electricity carbon market, including: a market response model acquisition module, used to acquire electricity carbon market data and historical optimal benchmarks, and based on the electricity carbon market data and the historical optimal benchmarks, obtain a market response model, an initial decision, and an initial adjustment range; a decision acceleration solution module, used to take the initial decision obtained by the market response model acquisition module as the current decision, and take the initial adjustment range obtained by the market response model acquisition module as the current adjustment range; and based on the current decision and the current adjustment range, perform an acceleration solution step to obtain an optimal decision; and a generator scheduling module, used to obtain a generator scheduling strategy based on the optimal decision obtained by the decision acceleration solution module, and to schedule the target generators based on the generator scheduling strategy. The group performs scheduling to obtain the clearing result of the carbon electricity market; the accelerated solution step includes: based on the market response model, performing approximation processing under the current decision and the current adjustment range to obtain a trust region subproblem; based on the trust region subproblem, performing dimensionality reduction processing to obtain the decision adjustment step size; based on the decision adjustment step size and the current decision, under a preset parameter update strategy, obtaining the updated decision quantity and the updated adjustment range; and determining whether the gradient vector of the updated decision is greater than a preset convergence threshold; if the gradient vector of the updated decision is greater than the preset convergence threshold, then the updated decision is taken as the current decision, the updated adjustment range is taken as the current adjustment range, and the accelerated solution step is re-executed based on the current decision and the current adjustment range; otherwise, the accelerated solution step is terminated, and the updated decision quantity is taken as the optimal decision.

[0103] Furthermore, the step of acquiring electricity carbon market data and historical optimal benchmarks, and obtaining a market response model, initial decision, and initial adjustment range based on the electricity carbon market data and the historical optimal benchmarks, includes: acquiring electricity carbon market data, and performing electricity carbon coupled two-level game modeling processing based on the electricity carbon market data to obtain a two-level game model; wherein the two-level game model includes a lower-level market clearing model; acquiring historical optimal benchmarks, and obtaining a market response model based on the historical optimal benchmarks and the lower-level market clearing model; and obtaining an initial decision and initial adjustment range based on the historical optimal benchmarks and under preset electricity carbon decision weights.

[0104] Furthermore, the step of obtaining the historical best benchmark point and, based on the historical best benchmark point and the lower-level market clearing model, obtaining the market response model includes: constructing an implicit response function under KKT conditions based on the Lagrangian function of the lower-level market clearing model; obtaining the historical best benchmark point and, based on the implicit response function, performing a second Taylor expansion on the historical best benchmark point to obtain the market response model.

[0105] The above embodiments utilize a market response model to approximate the mapping relationship from upper-level bidding decisions to lower-level clearing results, simplifying the mapping relationship between the upper and lower levels. This reduces the time required to obtain subsequent scheduling decisions based on the market response model, thereby improving the scheduling efficiency of generator sets. Furthermore, this embodiment uses the current decision and the current adjustment range to obtain a dynamic trust domain sub-problem, thus achieving dynamic control of the decision adjustment step size. This effectively avoids decision biases caused by model mismatch or blind search in traditional methods. Therefore, the differences between scheduling decisions obtained in different time periods are all within an acceptable adjustment range, enabling generator sets to form stable operating conditions based on the scheduling decisions of this embodiment. This prevents scheduling lags due to sudden changes in operating conditions, further improving the scheduling efficiency of generator sets.

[0106] The above description represents the preferred embodiments of the present invention. It should be noted that those skilled in the art can make various improvements and modifications without departing from the principles of the present invention, and these improvements and modifications are also considered to be within the scope of protection of the present invention.

Claims

1. A generator set dispatching method under an electricity carbon market, characterized by comprising the following steps: Obtain electricity carbon market data and historical best benchmarks, and based on the electricity carbon market data and historical best benchmarks, obtain the market response model, initial decision and initial adjustment range; The initial decision is taken as the current decision, and the initial adjustment range is taken as the current adjustment range; Based on the current decision and the current adjustment range, execute the accelerated solution steps to obtain the optimal decision; Based on the optimal decision, a generator set scheduling strategy is obtained, and the target generator sets are scheduled based on the generator set scheduling strategy to obtain the clearing result of the electricity carbon market. The accelerated solution steps include: Based on the market response model, an approximation is performed under the current decision and adjustment range to obtain the trust domain sub-problem; Based on the aforementioned trust domain subproblem, a dimensionality reduction solution is performed to obtain the decision adjustment step size; Based on the decision adjustment step size and the current decision, under the preset parameter update strategy, the updated decision amount and the update adjustment range are obtained; and it is determined whether the gradient vector of the updated decision is greater than the preset convergence threshold. If the gradient vector of the updated decision is greater than the preset convergence threshold, the updated decision is taken as the current decision, the updated adjustment range is taken as the current adjustment range, and the accelerated solution step is re-executed based on the current decision and the current adjustment range; otherwise, the accelerated solution step is terminated, and the updated decision is taken as the optimal decision.

2. The generator scheduling method under the electric carbon market according to claim 1, characterized in that, the step of acquiring electric carbon market data and historical optimal benchmark points, and obtaining a market response model, initial decision and initial adjustment range based on the electric carbon market data and the historical optimal benchmark points, includes: Acquire electricity carbon market data, and based on the electricity carbon market data, perform electricity carbon coupled two-level game modeling to obtain a two-level game model; wherein, the two-level game model includes a lower-level market clearing model; Obtain the historical best benchmark point, and based on the historical best benchmark point and the lower-level market clearing model, obtain the market response model; Based on the historical best benchmark, the initial decision and initial adjustment range are obtained under the preset carbon decision weights.

3. The generator set scheduling method under the electric carbon market according to claim 2, characterized in that, the step of obtaining the historical optimal benchmark point and obtaining the market response model based on the historical optimal benchmark point and the lower-level market clearing model includes: Based on the Lagrangian function of the lower-level market clearing model, an implicit response function is constructed under KKT conditions; Obtain the historical best benchmark point, and based on the implicit response function, perform a second Taylor expansion on the historical best benchmark point to obtain the market response model.

4. The generator set scheduling method under the electric carbon market according to claim 2, characterized in that the two-layer game model further includes an upper-layer decision model, and the approximate processing based on the market response model under the current decision and the current adjustment range to obtain the trust region sub-problem includes: Based on the market response model and the current decision, an approximate objective function is obtained under the objective function of the upper-level decision model; Based on the approximate objective function, a second Taylor expansion is performed to obtain a second approximate model; Based on the aforementioned quadratic approximation model, within the current adjustment range, the trust region subproblem is obtained.

5. The generator set scheduling method under the electric carbon market according to claim 1, characterized in that, the step of performing dimensionality reduction solution based on the trust domain subproblem to obtain the decision adjustment step size includes: Based on the aforementioned trust domain subproblem, a dimensionality reduction solution step is performed to obtain the decision adjustment step size; The dimensionality reduction solution steps include: Based on the aforementioned trust domain subproblem, a dimension reduction projection is performed using the Lanczos algorithm to obtain a low-dimensional trust domain subproblem. Based on the aforementioned low-dimensional trust domain subproblem, the solution process is performed using the More-Sorensen exact algorithm to obtain the low-dimensional step size. Based on the low-dimensional step size, high-dimensional recovery processing is performed to obtain the high-dimensional step size; and it is determined whether the high-dimensional step size meets the preset solution requirements. If the high-dimensional step size does not meet the preset solution requirements, the subspace dimension parameter of the Lanczos algorithm is increased according to the preset space expansion threshold, and the dimensionality reduction solution step is re-executed; otherwise, the dimensionality reduction solution step is terminated, and the high-dimensional step size is used as the decision adjustment step size.

6. The generator set scheduling method under the electric carbon market according to claim 1, characterized in that, based on the decision adjustment step size and the current decision, obtaining the updated decision and the updated adjustment range under a preset parameter update strategy includes: Based on the decision adjustment step size and the current decision, the model approximation is obtained; Based on the decision adjustment step size and the model approximation, the updated decision and the updated adjustment range are obtained under the preset decision and adjustment radius update strategy.

7. A generator scheduling method under an electric carbon market according to claim 6, characterized in that, based on the decision adjustment step size and the model approximation, obtaining the current decision quantity and the update adjustment range under a preset decision and trust region radius update strategy includes: Determine whether the model approximation is less than a preset first threshold; If the model approximation is less than a preset first threshold, then based on the current decision and the current adjustment range, a radius reduction process is performed to obtain an updated decision and an updated adjustment range; otherwise, it is determined whether the model approximation is less than a preset second threshold. If the model approximation is less than a preset second threshold, then a decision update process is performed based on the current decision and the current adjustment range to obtain an updated decision and an updated adjustment range; otherwise, a radius expansion and decision update process is performed based on the current decision and the current adjustment range to obtain an updated decision and an updated adjustment range.

8. A generator set dispatching system under an electricity carbon market, characterized in that, A generator set dispatching method under an electric carbon market as described in any one of claims 1 to 7, comprising: The market response model acquisition module is used to acquire electricity carbon market data and historical best benchmarks, and based on the electricity carbon market data and the historical best benchmarks, obtain the market response model, initial decision and initial adjustment range; The decision acceleration module is used to take the initial decision obtained by the market response model acquisition module as the current decision and the initial adjustment range obtained by the market response model acquisition module as the current adjustment range; and to perform acceleration solution steps based on the current decision and the current adjustment range to obtain the optimal decision. The generator set scheduling module is used to obtain a generator set scheduling strategy based on the optimal decision obtained by the decision acceleration solution module, and to schedule the target generator set based on the generator set scheduling strategy to obtain the clearing result of the electricity carbon market. The accelerated solution steps include: Based on the market response model, an approximation is performed under the current decision and adjustment range to obtain the trust domain sub-problem; Based on the aforementioned trust domain subproblem, a dimensionality reduction solution is performed to obtain the decision adjustment step size; Based on the decision adjustment step size and the current decision, under the preset parameter update strategy, the updated decision amount and the update adjustment range are obtained; and it is determined whether the gradient vector of the updated decision is greater than the preset convergence threshold. If the gradient vector of the updated decision is greater than the preset convergence threshold, the updated decision is taken as the current decision, the updated adjustment range is taken as the current adjustment range, and the accelerated solution step is re-executed based on the current decision and the current adjustment range; otherwise, the accelerated solution step is terminated, and the updated decision is taken as the optimal decision.

9. A generator set dispatching system under an electricity carbon market according to claim 8, characterized in that, the step of acquiring electricity carbon market data and historical optimal benchmark points, and obtaining a market response model, initial decision, and initial adjustment range based on the electricity carbon market data and the historical optimal benchmark points, includes: Acquire electricity carbon market data, and based on the electricity carbon market data, perform electricity carbon coupled two-level game modeling to obtain a two-level game model; wherein, the two-level game model includes a lower-level market clearing model; Obtain the historical best benchmark point, and based on the historical best benchmark point and the lower-level market clearing model, obtain the market response model; Based on the historical best benchmark, the initial decision and initial adjustment range are obtained under the preset carbon decision weights.

10. A generator set dispatching system under an electric carbon market according to claim 9, characterized in that, obtaining the historical optimal benchmark point and, based on the historical optimal benchmark point and the lower-level market clearing model, obtaining the market response model includes: Based on the Lagrangian function of the lower-level market clearing model, an implicit response function is constructed under KKT conditions; Obtain the historical best benchmark point, and based on the implicit response function, perform a second Taylor expansion on the historical best benchmark point to obtain the market response model.

Images