Power transaction method, device and system based on Markov optimization

Abstract

The invention discloses a power transaction method, device and system based on Markov optimization, and belongs to the technical field of power transaction, the transaction method comprises the following steps: in a day-ahead stage: constructing a Markov unit combination model corresponding to a Markov chain state transition matrix to determine day-ahead output results of two power generation parties; fixing a start-stop decision in the Markov unit combination model by using a day-ahead start-stop result of a conventional energy power generation party to determine a day-ahead electricity price; calculating the day-ahead payment amount under each possible state by using the day-ahead electricity price for the day-ahead output result; the complexity of the problem is reduced; and the excessive conservative property of a model solution is avoided. In the real-time stage, real-time output results corresponding to the two power generation parties are obtained, and the real-time electricity price is determined; calculating a real-time payment difference corresponding to a difference value between the real-time output result and the day-ahead output result by using the real-time electricity price, and carrying out day-ahead-real-time overall settlement; and the day-ahead settlement and the real-time settlement are linked and unified, so that the overall settlement result is more accurate.

Description

Technical Field

[0001] The present invention belongs to the technical field of power trading, and more specifically, relates to a power trading method, device and system based on Markov optimization. Background Art

[0002] Renewable energy generation, such as wind power and photovoltaic power generation, depends on weather conditions and is uncertain. In the process of electricity market clearing, renewable energy generation itself cannot be dispatched like conventional generation. When renewable energy generation is insufficient, conventional energy generation is generally used to meet the corresponding demand. Therefore, in the unit combination problem of day-ahead electricity market clearing, it is necessary to coordinate the power generation of renewable energy and conventional energy.

[0003] Existing methods include deterministic methods, stochastic programming methods, and robust optimization methods. Deterministic methods consider the expected value of wind or photovoltaic power generation; stochastic programming methods aim to minimize the expected cost on the probability distribution of uncertainty; and robust optimization methods seek the optimal solution in an uncertain set that does not require probability information. In the prior art, after the day-ahead power market is cleared, the start-stop decisions of conventional energy power generators are often fixed based on the solution of the deterministic method to obtain the corresponding economic dispatch problem, and the marginal electricity price is obtained through the solution of its dual problem for settlement in the power market.

[0004] However, the uncertainty of renewable energy makes the day-ahead electricity market clearing process more complicated and causes inaccurate settlement results. Summary of the invention

[0005] In view of the above defects or improvement needs of the prior art, the present invention provides a power trading method, device and system based on Markov optimization, which aims to solve the technical problem that the uncertainty of renewable energy makes the day-ahead power market clearing process more complicated and causes inaccurate settlement results.

[0006] To achieve the above object, according to one aspect of the present invention, a power trading method based on Markov optimization is provided, comprising:

[0007] S1: Use the historical power generation of renewable energy generators to construct the state transfer matrix of the Markov chain to predict the output results of the renewable energy generators on the first day before the day; establish and solve the Markov unit commitment model corresponding to the output results on the first day before the day, and obtain the start-stop results on the day before the day and the output results on the second day before the day corresponding to the conventional energy generators;

[0008] S2: fixing the start / stop decision of the conventional energy power generation party in the Markov unit combination model based on the day-ahead start / stop result to obtain the corresponding first economic dispatch problem; solving the first economic dispatch problem and performing sensitivity analysis to obtain the day-ahead electricity price;

[0009] S3: Calculate the day-ahead payment amount of the renewable energy power generation party and the conventional energy power generation party under each possible state by using the day-ahead electricity price, the first day-ahead output result and the second day-ahead output result;

[0010] S4: Obtain the first real-time output result corresponding to the renewable energy power generation party, and establish a corresponding second economic dispatch problem given the real-time start and stop status of the conventional energy power generation unit; solve the second economic dispatch problem to obtain the second real-time output result corresponding to the conventional energy power generation party;

[0011] S5: Perform sensitivity analysis on the second economic dispatch problem to obtain a real-time electricity price;

[0012] S6: Calculate the real-time payment difference corresponding to the difference between the first day-ahead output result and the first real-time output result using the real-time electricity price, and conduct the day-ahead-real-time overall settlement of the renewable energy power generator; Calculate the real-time payment difference corresponding to the difference between the second day-ahead output result and the second real-time output result using the real-time electricity price, and conduct the day-ahead-real-time overall settlement of the conventional energy power generator.

[0013] Furthermore, the S1 includes:

[0014] S11: Aggregate the historical power generation of the renewable energy power generator, and assume that the distribution of the historical power generation obeys a Markov chain, and use the historical power generation of the renewable energy power generator to construct a state transfer matrix of the Markov chain, so as to predict the output result of the renewable energy power generator corresponding to the first day before;

[0015] S12: Establish a Markov unit combination model corresponding to the first day-ahead output result, and select a set of start-stop decisions and multiple sets of economic dispatch decisions for the conventional energy power generation units according to the net demand state within the dispatch time range to minimize the total day-ahead expected cost.

[0016] S13: Solving the Markov unit commitment model corresponding to the first day-ahead output result to obtain the day-ahead start / stop result corresponding to the conventional energy power generation party and the second day-ahead output result.

[0017] Furthermore, the objective function of the Markov unit commitment model is:

[0018]

[0019] The system demand constraint in the constraint condition of the Markov unit commitment model is expressed as:

[0020]

[0021] Among them, J DA represents the day-ahead cost, T represents the time number, t represents the time index, n represents the state of the system's net demand; I represents the number of conventional energy generators, i represents the index of the conventional energy generator, u represents the i (t) represents the binary variable corresponding to the start and stop operation of generator i, S i is the startup cost, x i (t) represents the power generation state of generator i, represents the no-load cost when generator i is online, represents the probability of being in state n at time t. The power quotation of generator i contains multiple segments, and the price of segment b (1≤b≤B) is C i,b The maximum power is p i,bmax , the day-ahead power generation of generator i in hour t period b state n is The sum of the power of generator i in each section is D DA (t) represents the forecast value of the total system demand at hour t on the previous day, Represents the net demand of the system.

[0022] Furthermore, the S12 includes: solving the Markov unit combination model by using a branch-and-cut method to obtain the day-ahead start-stop result and the second-day-ahead output result corresponding to the conventional energy power generation party.

[0023] Further, the S2 includes:

[0024] S21: fixing the start / stop decision of the conventional power generation unit in the Markov unit combination model according to the day-ahead start / stop result corresponding to the conventional energy power generation party, and obtaining the corresponding first economic dispatch problem;

[0025] S22: The first economic dispatch problem is processed by linear programming and sensitivity analysis to obtain boundary values, and the day-ahead electricity price of the power generator under each state is calculated according to the dual variables in the boundary values.

[0026] Furthermore, S5 includes: processing the second economic dispatch problem using sensitivity analysis to obtain boundary values, and calculating the real-time electricity price of the power generator in each state according to the dual variables in the boundary values.

[0027] In one embodiment, the S6 includes:

[0028] The real-time electricity price is multiplied by the difference between the output result of the second day before and the second real-time output result corresponding to the renewable energy power generation party as the real-time payment difference of the renewable energy power generation party, and the day-ahead-real-time overall settlement of the renewable energy power generation party is performed;

[0029] The product of the real-time electricity price and the difference between the first day-ahead output result and the first real-time output result corresponding to the conventional energy power generation party is used as the real-time payment difference of the conventional energy power generation party, and the day-ahead-real-time overall settlement of the conventional energy power generation party is performed.

[0030] Furthermore, built into the blockchain platform, the S6 also includes:

[0031] When the renewable energy power generation party objects to the liquidated damages collected by the blockchain platform and files an arbitration application, the smart contract in the blockchain platform is used to vote on the arbitration application; wherein the liquidated damages are generated due to the large deviation between the day-ahead output result and the real-time output;

[0032] The actual amount of liquidated damages that the renewable energy generator ultimately needs to pay will be determined and settled based on the voting results.

[0033] According to another aspect of the present invention, there is provided a power trading device based on Markov optimization, comprising:

[0034] The day-ahead output calculation module is used to construct the state transfer matrix of the Markov chain using the historical power generation of the renewable energy power generation party to predict the day-ahead output result corresponding to the renewable energy power generation party; establish and solve the Markov unit combination model corresponding to the first day-ahead output result to obtain the day-ahead start-stop result corresponding to the conventional energy power generation party and the second day-ahead output result;

[0035] a day-ahead electricity price calculation module, for fixing the start-stop decision of the conventional energy power generation party in the Markov unit combination model based on the second day-ahead start-stop result, and obtaining the corresponding first economic dispatch problem; solving the first economic dispatch problem and performing sensitivity analysis to obtain the day-ahead electricity price;

[0036] a day-ahead settlement calculation module, configured to calculate the day-ahead payment amounts of the renewable energy power generator and the conventional energy power generator under each possible state by using the day-ahead electricity price, the first day-ahead output result, and the second day-ahead output result;

[0037] A real-time output calculation module is used to obtain a first real-time output result corresponding to the renewable energy power generation party, and establish a corresponding second economic dispatch problem given the start / stop status of the conventional energy power generation unit (which may be derived from the day-ahead start / stop result); solve the second economic dispatch problem to obtain a second real-time output result corresponding to the conventional energy power generation party;

[0038] A real-time electricity price calculation module, used for performing sensitivity analysis on the second economic dispatch problem to obtain a real-time electricity price;

[0039] A real-time settlement calculation module is used to use the real-time electricity price to calculate the real-time payment difference corresponding to the difference between the first day-ahead output result and the first real-time output result, and to perform the day-ahead-real-time overall settlement of the renewable energy power generation party; and to use the real-time electricity price to calculate the real-time payment difference corresponding to the difference between the second day-ahead output result and the second real-time output result, and to perform the day-ahead-real-time overall settlement of the conventional energy power generation party.

[0040] According to another aspect of the present invention, there is provided an electricity trading system, comprising a memory and a processor, wherein the memory stores a computer program, and the processor implements the steps of the above method when executing the computer program.

[0041] According to another aspect of the present invention, a computer-readable storage medium is provided, on which a computer program is stored, and when the computer program is executed by a processor, the steps of the above method are implemented.

[0042] In general, the above technical solutions conceived by the present invention can achieve the following beneficial effects compared with the prior art:

[0043] (1) The present invention is directed to a hybrid power system containing renewable energy. In the day-ahead stage: a Markov unit combination model corresponding to the Markov chain is constructed to determine the day-ahead output results corresponding to the two power generators and the day-ahead start / stop results of the conventional energy power generator; the day-ahead start / stop results of the conventional energy power generator are used to fix the start / stop decision of the conventional energy power generator in the Markov unit combination model and a sensitivity analysis is performed to determine the day-ahead electricity price; the day-ahead electricity price is used to calculate the day-ahead payment amount under each possible state for the two day-ahead output results. In the real-time stage: the real-time output results corresponding to the renewable energy power generator are obtained, a real-time economic dispatch problem model is established given the start / stop state of the conventional energy power generator (which can be derived from the day-ahead start / stop results), and the real-time output results corresponding to the conventional energy power generator are calculated, and a sensitivity analysis is performed to determine the real-time electricity price; the real-time electricity price is used to calculate the real-time payment difference corresponding to the difference between the real-time output result and the day-ahead output result and a day-ahead-real-time unified settlement is performed for the two power generators. The present invention takes into account the uncertainty of renewable energy in the day-ahead clearing problem, and summarizes all the information of previous moments in a probabilistic sense through the state at a certain moment, thereby reducing the complexity of the problem and avoiding the over-conservatism of the model solution. During the settlement process, the impact of renewable energy uncertainty on the day-ahead price is included, and the day-ahead and real-time settlements are connected and unified based on the real-time determined renewable energy status, making the overall settlement results more accurate.

[0044] (2) The characteristics of the Markov unit combination model described in this scheme are as follows: the state transfer matrix is ​​used to predict the output result of the renewable energy power generation party on the first day before the day; a set of start-stop decisions and multiple sets of economic dispatch decisions of conventional energy power generation units are selected according to the net demand state within the dispatch time range to minimize the total expected cost on the first day before the day; the state at a certain moment summarizes the information of all previous moments in a probabilistic sense, which reduces the complexity of the overall problem compared to the random programming method; compared to the robust optimization method, the excessive conservatism of the model and solution is avoided because the probabilistic information is taken into account.

[0045] (3) This scheme uses the branch-and-cut method to solve the Markov unit combination model. Commercial optimization solvers such as Gurobi and COPT do not provide a solution method for random processes. This scheme includes the state probability as a weight in the objective function, and the system demand constraint only needs to be established for those states with non-zero probability, and the ramp rate constraint and state transition only apply to those states with non-zero probability. Given the state transition probability, the calculated state probability, the objective equation and the constraints, the problem is expressed from a Markov decision process to a mixed integer linear programming problem (i.e., Markov optimization), which can be effectively solved using the branch-and-cut method, accelerating the solution while ensuring convergence.

[0046] (4) This scheme fixes the start-stop decision of the conventional power generation unit according to the start-stop result of the conventional energy power generation party on the second day before, obtains the corresponding economic dispatch problem and uses linear programming and sensitivity analysis to process the economic dispatch problem to obtain the boundary value, and calculates the day-ahead electricity price of the power generation party under each state according to the dual variables in the boundary value; the real-time electricity price takes into account the determined renewable energy state and the output uncertainty of renewable energy, so that the electricity price calculation result is more accurate.

[0047] (5) This solution uses the product of the real-time electricity price and the difference between the day-ahead output result and the real-time output result as the real-time payment difference, which matches the actual application scenario and is simple to operate.

[0048] (6) This scheme provides arbitration for possible deviation disputes of renewable energy. While reducing costs and improving system operation efficiency, arbitration can protect the interests of renewable energy generators when unavoidable unexpected weather conditions occur, or when major maintenance plan adjustments, major equipment accidents in power grids and transmission power stations affect renewable energy generation, thereby ensuring the enthusiasm of renewable energy generators to participate in power trading. BRIEF DESCRIPTION OF THE DRAWINGS

[0049] Figure 1 A flow chart of a power trading method based on Markov optimization provided in Example 1 of the present invention;

[0050] Figure 2 A schematic diagram of state transition corresponding to a renewable energy power generation party provided in Example 1 of the present invention;

[0051] Figure 3 A schematic diagram of a power trading system provided in Embodiment 1 of the present invention;

[0052] Figure 4 A schematic diagram of the layered architecture of the blockchain platform provided in Example 1 of the present invention;

[0053] Figure 5 A specific schematic diagram of arbitration in the power trading method based on Markov optimization provided in Embodiment 1 of the present invention;

[0054] Figure 6 A specific flow chart of arbitration in the power trading method based on Markov optimization provided in Example 1 of the present invention. DETAILED DESCRIPTION

[0055] In order to make the purpose, technical solutions and advantages of the present invention more clearly understood, the present invention is further described in detail below in conjunction with the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are only used to explain the present invention and are not intended to limit the present invention. In addition, the technical features involved in the various embodiments of the present invention described below can be combined with each other as long as they do not conflict with each other.

[0056] Example 1

[0057] like Figure 1 As shown, this embodiment provides a power trading method based on Markov optimization, including: S1-S6.

[0058] S1: Use the historical power generation of renewable energy power generation to construct the state transfer matrix of the Markov chain to predict the output results of the renewable energy power generation on the first day before the day; solve the Markov unit combination model corresponding to the output results on the first day before the day, and obtain the start-stop results on the first day before the day and the output results on the second day before the day corresponding to the conventional energy power generation.

[0059] S2: Based on the day-ahead start / stop results, the start / stop decisions of conventional energy power generation parties in the Markov unit combination model are fixed to obtain the corresponding first economic dispatch problem; the first economic dispatch problem is solved and sensitivity analysis is performed to obtain the day-ahead electricity price.

[0060] S3: Calculate the day-ahead payment amount of renewable energy power generators and conventional energy power generators under each possible state using the day-ahead electricity price, the first day-ahead output result, and the second day-ahead output result.

[0061] S4: Obtain the first real-time output result corresponding to the renewable energy power generation party, and establish the corresponding second economic dispatch problem given the real-time start and stop status of the conventional energy power generation unit (which can be derived from the day-ahead start and stop result); solve the second economic dispatch problem to obtain the second real-time output result corresponding to the conventional energy power generation party.

[0062] S5: Conduct sensitivity analysis on the second economic dispatch problem to obtain the real-time electricity price.

[0063] S6: Calculate the real-time payment difference corresponding to the difference between the output result of the first day before and the first real-time output result using the real-time electricity price, and conduct the day-ahead-real-time overall settlement for renewable energy power generators; Calculate the real-time payment difference corresponding to the difference between the output result of the second day before and the second real-time output result using the real-time electricity price, and conduct the day-ahead-real-time overall settlement for conventional energy power generators.

[0064] As an optional implementation, S1 includes: S11: Aggregate the historical power generation of renewable energy power generators, and assume that the distribution of historical power generation obeys the Markov chain, and use the historical power generation of renewable energy power generators to construct the state transfer matrix of the Markov chain, which is used to predict the output results of the renewable energy power generators on the first day before the day. S12: Establish a Markov unit combination model corresponding to the output results on the first day before the day, and select a set of start and stop decisions and multiple sets of economic dispatch decisions for conventional energy power generators according to the net demand state within the scheduling time range to minimize the total expected cost on the day before the day. S13: Solve the Markov unit combination model corresponding to the output results on the first day before the day to obtain the start and stop results on the day before the day and the output results on the second day before the day corresponding to the conventional energy power generator.

[0065] Specifically, in this Markov chain, the capacity of renewable energy generators is evenly divided into multiple intervals, the mean of each interval is represented by a state, and the states are arranged in ascending order of mean. The transfer matrix is ​​represented by π as follows:

[0066]

[0067] Among them, the (m, n)th element of the matrix π mn It represents the ratio of the number of observed transitions from state m to state n to the number of occurrences of state m.

[0068] Its state probability is calculated as follows:

[0069]

[0070] in, represents the probability that the total power generation of renewable energy generators is in state n at time t+1, represents the probability of being in state m at time t. The state transition of the total power generation of renewable energy generators is as follows: Figure 2 shown.

[0071] Since the total power generation of renewable energy generators is equal to the sum of the power generation of each renewable energy generator, the power generation state of each renewable energy generator corresponds to the total state, that is:

[0072]

[0073] in, represents the total (expected) power generation of renewable energy generators at t hours and n states; K represents the number of renewable energy generators, k is the index of the renewable energy generator, is the day-ahead (expected) power generation of renewable energy generator k at hour t under state n.

[0074] As an optional implementation, S12 includes: the Markov unit commitment model selects a set of unit start-stop decisions and multiple sets of economic dispatch decisions according to the net demand state of the system within the dispatch time range to minimize the total expected cost before the day, such as 24 hours before the day. The objective function of the Markov unit commitment model is:

[0075]

[0076] Among them, J DA represents the day-ahead cost, T represents the time number, t represents the time index, n represents the state of the system's net demand; I represents the number of conventional energy generators, i represents the index of the conventional energy generator, u represents the i (t) represents the binary variable corresponding to the start-up operation of generator i, S i is the startup cost, x i (t) represents the start / stop decision of generator i, “1” represents online, “0” represents offline, represents the no-load cost when generator i is online, represents the probability of being in state n at time t. The power quotation of generator i contains multiple segments, and the price of segment b (1≤b≤B) is C i,b The maximum power is p i,bmax , the day-ahead power generation of generator i in hour t period b state n is The sum of the power of generator i in each section (economic dispatch decision) is

[0077] The constraints include system demand constraints and conventional generator constraints. Specifically, the system demand constraints can be expressed as:

[0078]

[0079] Among them, D DA(t) represents the forecast value of the total system demand at hour t on the day before, assuming that the system demand is inelastic and ignoring the uncertainty. It should be noted that as a Markov decision problem, the economic dispatch decision of conventional generators at hour t is It depends only on the state n of the renewable energy generator or the net demand of the system at time t.

[0080] Generator constraints include starting constraints, generator capacity constraints, ramp rate constraints, and minimum rise / fall time constraints.

[0081] Specifically, the startup constraints can be expressed as:

[0082]

[0083] Since the generator quotation contains multiple sections, the power generation of each section cannot exceed the maximum value of the section, which can be expressed as:

[0084]

[0085] The sum of the power generation of all quoted sections of an engine is equal to the power generation of the generator, that is:

[0086]

[0087] The generator capacity constraint can be expressed as:

[0088]

[0089] where p imin and p imax are the minimum and maximum power generation of generator i respectively.

[0090] The power generation change of generator i between t and t+1 hours cannot exceed its ramp rate Δ i Since the net demand of the system may be in different states at these two times, the ramp rate constraint should be satisfied for all possible state transitions. Therefore, the ramp rate constraint can be expressed as:

[0091]

[0092] The shortest startup / shutdown time constraint can be expressed as:

[0093]

[0094] As an optional implementation, S13 includes: solving the Markov unit commitment model using a branch-and-cut method to obtain the day-ahead start / stop results and the day-ahead output results corresponding to the conventional energy power generation party.

[0095] Specifically, the branch-and-cut method combines the branch-and-bound method and the cutting plane method. The branch-and-cut method is effective in solving deterministic mixed integer linear problems and has been widely used by the International Organization for Standardization, utilities, semiconductor manufacturers, etc. In addition, the existence of commercial optimization solvers such as Gurobi and COPT reduces the programming time and debugging time for solving such problems. However, these software do not provide a solution method for random processes. For the method used in the present invention, the state probability is included as a weight in the objective function (4), and the system demand constraint (5) only needs to be established for those states with non-zero probability, and the ramp rate constraints (10) and (11) only apply to those state transitions with non-zero probability. Given the state transition probability, the state probability calculated based on (2), the objective equation and the constraints, the problem is expressed from a Markov decision process to a mixed integer linear programming problem (i.e., a Markov optimization problem), which can be effectively solved using the branch-and-cut method, accelerating the solution while ensuring convergence. Thus, a set of unit start-stop decisions x is obtained. i (t) and multiple sets of economic dispatch decisions that depend on state n

[0096] As an optional implementation, S2 includes: S21: fixing the start and stop decisions of conventional power generating units in the Markov unit combination model according to the corresponding day-ahead start and stop results of conventional energy power generators, and obtaining the corresponding first economic dispatch problem; S22: using linear programming and sensitivity analysis to process the first economic dispatch problem to obtain boundary values, and calculating the day-ahead electricity price of the power generator in each state according to the dual variables in the boundary values.

[0097] Specifically, considering a small change in the net demand of the system at time t and state n, based on the sensitivity formula of the linear programming problem, the target value J is relative to The change in can be measured by the dual variable (shadow price) of constraint (5), expressed as Right now:

[0098]

[0099] Among them, the dual variable is the boundary value (sensitivity) of (5), where λ n (t) represents the day-ahead electricity price for n conditions at hour t.

[0100] S3: Calculate the day-ahead payment amount for renewable energy generators and conventional energy generators under each possible state using the day-ahead electricity price, the output result of the first day and the output result of the second day. Based on (14), we can get λ n (t)p i,n (t) is the day-ahead electricity cost of generator i in state n, expressed as So:

[0101]

[0102] Similarly, the day-ahead electricity cost of renewable energy generator k at state n is expressed as:

[0103]

[0104] Note: Since there are multiple possible states in the day-ahead phase, it is difficult to settle directly. Therefore, settlement will be made in the real-time phase (S6) according to the real-time state (a state corresponding to the real-time renewable energy power generation power).

[0105] S4: Obtain the first real-time output result corresponding to the renewable energy power generation party, and establish the corresponding second economic dispatch problem given the real-time start and stop status of the conventional energy power generation unit (which can be derived from the day-ahead start and stop result); solve the second economic dispatch problem to obtain the second real-time output result corresponding to the conventional energy power generation party.

[0106] Specifically, assuming that the power generation capacity (or net system demand) of renewable energy generators is determined in real time (hour t'), there is no need to consider state changes. Given the real-time start and stop status of conventional energy generators (which can be derived from the day-ahead start and stop results), the objective function of the real-time economic dispatch problem can be expressed as:

[0107]

[0108] Among them, J RT represents the real-time cost, I' represents the number of online conventional energy generators, and i is the index of the conventional energy generator represents the real-time power generation of generator i, Indicates the real-time power generation of segment b.

[0109] The system requirement constraints can be expressed as:

[0110]

[0111] Among them, D RT (t') represents the total real-time system demand, Represents the real-time power generation of renewable energy generator k.

[0112] The generator capacity constraint can be expressed as:

[0113]

[0114] The power generation constraints of each section are expressed as:

[0115]

[0116] The total power generated by the generator is expressed as:

[0117]

[0118] The ramp rate constraint can be expressed as:

[0119]

[0120] The second economic dispatch problem is solved by linear programming method to obtain the second real-time output result corresponding to the conventional energy power generation party.

[0121] As an optional implementation, S5 includes: processing the second economic dispatch problem using sensitivity analysis to obtain boundary values, and calculating the real-time electricity price of the power generator in each state according to the dual variables in the boundary values.

[0122] Specifically, based on the sensitivity formula of the linear programming problem, solving the dual variable (shadow price) corresponding to (18), we can get λ RT (t'), which is the real-time electricity price.

[0123] In one embodiment, S6 includes: using the real-time electricity price and the product of the difference between the second day-ahead output result corresponding to the renewable energy power generation party and the second real-time output result as the payment difference for the real-time settlement of the renewable energy power generation party, and performing the day-ahead-real-time overall settlement of the renewable energy power generation party; using the real-time electricity price and the product of the difference between the first day-ahead output result corresponding to the conventional energy power generation party and the first real-time output result as the payment difference for the real-time settlement of the conventional energy power generation party, and performing the day-ahead-real-time overall settlement of the conventional energy power generation party.

[0124] Specifically, assuming that the power generation of the renewable energy generator in real time (hour t') corresponds to a day-ahead state n' (i.e., real-time state). For generator i, the total electricity fee can be expressed as the day-ahead electricity fee corresponding to the real-time state plus the electricity fee for over-generation in real time or minus the electricity fee for under-generation in real time, as shown below:

[0125]

[0126] Among them, W i Tot (t') represents the total electricity cost of generator i, represents the day-ahead electricity cost of generator i, W i RT (t') represents the real-time electricity cost (deviation adjustment cost) of generator i, represents the real-time power generation of generator i, represents the day-ahead power generation of generator i.

[0127] Similarly, for renewable energy generator k, it is also required to pay deviation adjustment fees. In addition, when the power generation is less than the previous day's power, a deviation penalty is also required. Then,

[0128]

[0129] in, represents the total electricity cost of renewable energy generator k, represents the day-ahead electricity cost of renewable energy generator k, represents the real-time electricity cost (deviation adjustment fee) of renewable energy generator k, represents the deviation penalty of renewable energy generator k, λ penal Indicates the unit penalty price, represents the real-time power generation of renewable energy generator k, represents the day-ahead power generation of renewable energy generator k in state n'.

[0130] As an optional implementation method, it is built into the blockchain platform and includes after S6: when the renewable energy power generation party objects to the liquidated damages collected by the blockchain platform and files an arbitration application, the smart contract in the blockchain platform is used to vote on the arbitration application; the liquidated damages are generated due to the large deviation between the day-ahead output result and the real-time output; the actual amount of liquidated damages that the renewable energy power generation party ultimately needs to pay is determined and settled according to the voting results.

[0131] The blockchain platform uses a Hyperledger Fabric consortium chain. The layered architecture of the blockchain platform includes five layers: application layer, smart contract layer, consensus layer, network layer, and data layer. Figure 3 and Figure 4 The power generator, power purchaser and arbitrator call the smart contract on the smart contract layer through the application provided to each party by the application layer, such as Figure 5 As shown. Smart contracts are executable computer programs that are responsible for the automatic execution of various transaction transactions, including output contracts, settlement contracts, arbitration contracts, and query contracts. The network layer is mainly composed of a channel. The sorting node and peer nodes will join the channel, and peer nodes of the same type will also form an organization. After the smart contract submits the transaction, the sorting node will sort the transaction, the consensus protocol of the consensus layer will cooperate with the peer nodes to verify and execute the transaction, and finally the transaction information will be passed to the data layer through the peer node, encrypted by the encryption algorithm, stored in the ledger, and the world state will be updated.

[0132] Specifically, Figure 6 As shown in the figure, after the settlement is completed, the renewable energy generator can call the relevant smart contract through the application to initiate an arbitration application. The arbitration application needs to include the transaction information and the reason for applying for arbitration.

[0133] After receiving the arbitration request, the arbitrator will review the transaction information and vote. Specifically, the arbitrator will check the transaction contract and some more transaction information, especially the relevant weather conditions. The relevant weather conditions of the power generation party during the transaction will be recorded. For wind power generation, the weather condition is wind speed; for photovoltaic power generation, the weather condition is irradiance. After verifying the information required for arbitration, the arbitrator votes through the smart contract. Furthermore, the vote is Boolean: if the arbitrator believes that the applicant is responsible, the vote is "yes" and the smart contract records it as 0; if the arbitrator believes that the applicant is not responsible, the vote is "no" and the smart contract records it as 1.

[0134] The smart contract determines whether the number of votes exceeds 2 / 3 of the total number of arbitrators. If so, the voting is completed, the smart contract counts the votes, and determines the amount of liquidated damages that the power generator needs to pay according to the voting results; otherwise, the voting continues and goes to the previous step. Specifically, the arbitrator will check the transaction contract and some more transaction information, especially the relevant weather conditions. The relevant weather conditions of the power generator during the transaction will be recorded. For wind power generation, the weather condition is wind speed; for photovoltaic power generation, the weather condition is irradiance. After verifying the information required for arbitration, the arbitrator votes through the smart contract.

[0135] Furthermore, the vote is of Boolean type: if the arbitrator believes that the applicant is responsible, the vote is "yes" and the smart contract records it as 0; if the arbitrator believes that the applicant is not responsible, the vote is "no" and the smart contract records it as 1.

[0136] After the vote counting is completed, if there are more "yes" votes, the applicant can be exempted from the deviation penalty, otherwise the applicant's deviation penalty will remain the calculated deviation penalty. At the same time, the applicant needs to pay the arbitration fee after the arbitration is completed. Furthermore, the final deviation penalty that the applicant needs to pay is as follows:

[0137]

[0138] in V represents the final penalty for deviation that applicant k needs to pay after arbitration. s The number of votes indicating “yes”, V total Indicates the total number of votes.

[0139] Furthermore, after the arbitration is concluded, the applicant is also required to pay arbitration fees, which are calculated based on the voting results. Specifically, in order to ensure the fairness and efficiency of the arbitration process, it is stipulated that only arbitrators who vote and whose votes are consistent with the final arbitration results are eligible to receive arbitration fees, while arbitrators who vote but whose votes are inconsistent with the final arbitration results or arbitration members who did not vote cannot receive arbitration fees. This can encourage arbitrators to participate in voting and encourage them to vote according to the actual situation of the transaction.

[0140] Furthermore, the calculation formula for arbitration costs is as follows:

[0141]

[0142] Among them, W k V It represents the arbitration fee that applicant k needs to pay, and p0 represents the unit arbitration fee.

[0143] Example 2

[0144] This embodiment provides a power trading device based on Markov optimization, including: a day-ahead output calculation module, a day-ahead electricity price calculation module, a day-ahead settlement calculation module, a real-time output calculation module, a real-time electricity price calculation module and a real-time settlement calculation module.

[0145] The day-ahead output calculation module is used to construct the state transfer matrix of the Markov chain using the historical power generation of renewable energy power generation parties to predict the day-ahead output results corresponding to the renewable energy power generation parties; establish and solve the Markov unit combination model corresponding to the day-ahead output results to obtain the day-ahead start and stop results and the day-ahead output results corresponding to the conventional energy power generation parties.

[0146] The day-ahead electricity price calculation module is used to fix the start-stop decision of the conventional energy power generation party in the Markov unit combination model based on the start-stop result the day before, and obtain the corresponding first economic dispatch problem; solve the first economic dispatch problem and perform sensitivity analysis to obtain the day-ahead electricity price;

[0147] The day-ahead settlement calculation module is used to calculate the day-ahead payment amount of renewable energy power generators and conventional energy power generators under various possible states using the day-ahead electricity price, the first day-ahead output result and the second day-ahead output result.

[0148] The real-time output calculation module is used to obtain the first real-time output result corresponding to the renewable energy power generation party, and establish the corresponding second economic dispatch problem under the given start-stop status of the conventional energy power generation unit (which can be derived from the start-stop result of the day before); solve the second economic dispatch problem to obtain the second real-time output result corresponding to the conventional energy power generation party.

[0149] The real-time electricity price calculation module is used to perform sensitivity analysis on the second economic dispatch problem to obtain the real-time electricity price.

[0150] The real-time settlement calculation module is used to use the real-time electricity price to calculate the real-time payment difference corresponding to the difference between the output result of the first day before and the first real-time output result, and to conduct the day-ahead-real-time overall settlement of renewable energy power generation parties; and to use the real-time electricity price to calculate the real-time payment difference corresponding to the difference between the output result of the second day before and the second real-time output result, and to conduct the day-ahead-real-time overall settlement of conventional energy power generation parties.

[0151] Example 3

[0152] This embodiment provides an electricity trading system, including a memory and a processor, wherein the memory stores a computer program, and the processor implements the steps of the above method when executing the computer program.

[0153] Example 4

[0154] This embodiment provides a computer-readable storage medium on which a computer program is stored. When the computer program is executed by a processor, the steps of the above method are implemented.

[0155] It will be easily understood by those skilled in the art that the above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions and improvements made within the spirit and principles of the present invention should be included in the protection scope of the present invention.

Claims

1. A method for power trading based on Markov optimization, characterized in that: include: S1: Use the historical power generation of renewable energy generators to construct the state transfer matrix of the Markov chain to predict the output results of the renewable energy generators on the first day before the day; establish and solve the Markov unit commitment model corresponding to the output results on the first day before the day, and obtain the start-stop results on the day before the day and the output results on the second day before the day corresponding to the conventional energy generators; S2: fixing the start / stop decision of the conventional energy power generation party in the Markov unit combination model based on the day-ahead start / stop result to obtain the corresponding first economic dispatch problem; solving the first economic dispatch problem and performing sensitivity analysis to obtain the day-ahead electricity price; S3: Calculate the day-ahead payment amount of the renewable energy power generation party and the conventional energy power generation party under each possible state by using the day-ahead electricity price, the first day-ahead output result and the second day-ahead output result; S4: Obtain the first real-time output result corresponding to the renewable energy power generation party, and establish a corresponding second economic dispatch problem given the real-time start and stop status of the conventional energy power generation unit; solve the second economic dispatch problem to obtain the second real-time output result corresponding to the conventional energy power generation party; S5: Perform sensitivity analysis on the second economic dispatch problem to obtain a real-time electricity price; S6: Calculate the real-time payment difference corresponding to the difference between the first day-ahead output result and the first real-time output result using the real-time electricity price, and conduct the day-ahead-real-time overall settlement of the renewable energy power generator; Calculate the real-time payment difference corresponding to the difference between the second day-ahead output result and the second real-time output result using the real-time electricity price, and conduct the day-ahead-real-time overall settlement of the conventional energy power generator.

2. The power trading method based on Markov optimization according to claim 1, characterized in that: The S1 includes: S11: Aggregate the historical power generation of the renewable energy power generator, and assume that the distribution of the historical power generation obeys a Markov chain, and use the historical power generation of the renewable energy power generator to construct a state transfer matrix of the Markov chain, so as to predict the output result of the renewable energy power generator corresponding to the first day before; S12: establishing a Markov unit commitment model corresponding to the first day-ahead output result, wherein the Markov unit commitment model is used to select a set of start-stop decisions and multiple sets of economic dispatch decisions for the conventional energy generating units according to the net demand state within the dispatch time range to minimize the total day-ahead expected cost; S13: Solving the Markov unit commitment model corresponding to the first day-ahead output result to obtain the day-ahead start / stop result corresponding to the conventional energy power generation party and the second day-ahead output result.

3. The power trading method based on Markov optimization according to claim 2, characterized in that: The S12 includes: the objective function of the Markov unit commitment model is: The system demand constraint in the constraint condition of the Markov unit commitment model is expressed as: Among them, J DA represents the day-ahead cost, T represents the time number, t represents the time index, n represents the state of the system's net demand; I represents the number of conventional energy generators, i represents the index of the conventional energy generator, u represents the i (t) represents the binary variable corresponding to the start and stop operation of generator i, S i is the startup cost, x i (t) represents the power generation state of generator i, represents the no-load cost when generator i is online, represents the probability of being in state n at time t. The power quotation of generator i contains multiple segments, and the price of segment b (1≤b≤B) is C i,b The maximum power is p i,bmax , the day-ahead power generation of generator i in hour t period b state n is The sum of the power of generator i in each section is D DA (t) represents the forecast value of the total system demand at hour t on the previous day, Represents the net demand of the system.

4. The power trading method based on Markov optimization according to claim 2, characterized in that: The S13 includes: solving the Markov unit combination model by using the branch-cut method to obtain the day-ahead start-stop result and the day-ahead output result corresponding to the conventional energy power generation party.

5. The power trading method based on Markov optimization according to claim 1, characterized in that: The S2 includes: S21: fixing the start / stop decision of the conventional power generation unit in the Markov unit combination model according to the day-ahead start / stop result corresponding to the conventional energy power generation party, and obtaining the corresponding first economic dispatch problem; S22: The first economic dispatch problem is processed by linear programming and sensitivity analysis to obtain boundary values, and the day-ahead electricity price of the power generator under each state is calculated according to the dual variables in the boundary values.

6. The power trading method based on Markov optimization according to claim 5, characterized in that: The S5 includes: processing the second economic dispatch problem by using sensitivity analysis to obtain a boundary value, and calculating the real-time electricity price of the power generator according to the dual variables in the boundary value.

7. The power trading method based on Markov optimization according to claim 1, characterized in that: The S6 includes: The real-time electricity price is multiplied by the difference between the output result of the second day before and the second real-time output result corresponding to the renewable energy power generation party as the real-time payment difference of the renewable energy power generation party, and the day-ahead-real-time overall settlement of the renewable energy power generation party is performed; The product of the real-time electricity price and the difference between the first day-ahead output result and the first real-time output result corresponding to the conventional energy power generation party is used as the real-time payment difference of the conventional energy power generation party, and the day-ahead-real-time overall settlement of the conventional energy power generation party is performed.

8. The power trading method based on Markov optimization according to any one of claims 1 to 7, characterized in that: Built into the blockchain platform, the S6 also includes: When the renewable energy power generation party objects to the liquidated damages collected by the blockchain platform and files an arbitration application, the smart contract in the blockchain platform is used to vote on the arbitration application; the liquidated damages are generated due to a large deviation between the day-ahead output result and the real-time output; the actual amount of liquidated damages that the renewable energy power generation party ultimately needs to pay is determined and settled according to the voting results.

9. A power trading device based on Markov optimization, characterized in that: include: The day-ahead output calculation module is used to construct the state transfer matrix of the Markov chain using the historical power generation of the renewable energy power generation party to predict the day-ahead output result corresponding to the renewable energy power generation party; establish and solve the Markov unit combination model corresponding to the first day-ahead output result to obtain the day-ahead start-stop result corresponding to the conventional energy power generation party and the second day-ahead output result; a day-ahead electricity price calculation module, for fixing the start-stop decision of the conventional energy power generation party in the Markov unit combination model based on the second day-ahead start-stop result, and obtaining the corresponding first economic dispatch problem; solving the first economic dispatch problem and performing sensitivity analysis to obtain the day-ahead electricity price; a day-ahead settlement calculation module, configured to calculate the day-ahead payment amounts of the renewable energy power generator and the conventional energy power generator under each possible state by using the day-ahead electricity price, the first day-ahead output result, and the second day-ahead output result; A real-time output calculation module is used to obtain a first real-time output result corresponding to the renewable energy power generation party, and establish a corresponding second economic dispatch problem given the start and stop status of the conventional energy power generation unit; solve the second economic dispatch problem to obtain a second real-time output result corresponding to the conventional energy power generation party; A real-time electricity price calculation module, used for performing sensitivity analysis on the second economic dispatch problem to obtain a real-time electricity price; A real-time settlement calculation module is used to use the real-time electricity price to calculate the real-time payment difference corresponding to the difference between the first day-ahead output result and the first real-time output result, and to perform the day-ahead-real-time overall settlement of the renewable energy power generation party; and to use the real-time electricity price to calculate the real-time payment difference corresponding to the difference between the second day-ahead output result and the second real-time output result, and to perform the day-ahead-real-time overall settlement of the conventional energy power generation party.

10. An electric power trading system, comprising a memory and a processor, wherein the memory stores a computer program, characterized in that: When the processor executes the computer program, the steps of the method according to any one of claims 1 to 8 are implemented.

Images