Electric quantity regulation and control method and system based on quantum algorithm and neural network in electric power market environment

Abstract

The invention relates to the technical field of short-term electric quantity regulation and control, in particular to an electric quantity regulation and control method and system based on a quantum algorithm and a neural network in an electric power market environment. According to the method, a plurality of quantum registers are constructed to form a population and are initialized, the quantum registers are individuals representing market decisions, and the market decisions are sets of declared quantity-price curves of all participants in the electricity market; the quantum code of one participant is selected from each quantum register to serve as a candidate solution, the fitness of the candidate solution is evaluated, and updating variation is conducted on half of the quantum registers with the smaller fitness of the candidate solution so as to iterate the population; when the number of population iterations reaches a set threshold value, selecting an individual which meets the constraint conditions of each participant and has the maximum optimization target as an optimal declaration strategy; obtaining a declaration quantity price decision of each participant through optimal declaration strategy inverse quantum coding; the optimization objective as the individual's market decision is to maximize the sum of producer and consumer residuals. According to the method, the defect of poor power resource configuration efficiency in the prior art is overcome, market decision optimization is performed in combination with the quantum genetic algorithm, declaration reference is provided for each participant, so that clearing power is improved, power configuration is optimized, and configuration efficiency is improved.

Description

Technical Field

[0001] The present invention relates to the technical field of short-term power quantity regulation, and in particular to a power quantity regulation method and system based on quantum algorithms and neural networks in a power market environment. Background Technique

[0002] Power quantity regulation is divided into long-term plans and short-term plans. The long-term plan is the power quantity estimation of power generation units and power consumption units for a relatively long time; the long-term plan is beneficial for power generation units to estimate the consumption in advance, so as to arrange the power generation plan in advance, ensure the power grid power quantity, and avoid power consumption congestion. However, the long-term plan has a low prediction accuracy. Specifically, for a certain day or even a certain moment, there may still be an imbalance between supply and demand. Therefore, it is necessary to combine short-term prediction for supplementation.

[0003] Short-term prediction is to predict the supply and demand on the same day or even at a certain moment on the basis of long-term prediction, so as to adjust the balance point of the current power sales volume and power purchase volume.

[0004] The power market is the execution environment of short-term plans. Power generation units, as power sellers, declare power sales volume and power sales price, and power consumption units, as power buyers, declare power purchase volume and power purchase price. Through market execution regulation, the supply and demand balance is achieved to optimize the declared volume and price of market participants and provide a decision-making basis for participants.

[0005] However, the problem of blind declaration of market participants caused by incomplete market information will lead to the problems of low cleared power quantity and high cleared result cost of market participants, which is not conducive to the optimal allocation and efficient utilization of power resources. Summary of the Invention

[0006] In order to overcome the defect that the power resource allocation efficiency is poor and needs to be further optimized in the above-mentioned existing technology, the present invention proposes a power quantity regulation method based on quantum algorithms and neural networks in a power market environment, which combines the fitness of the declared volume-price curve of power market participants and the quantum genetic algorithm to optimize market decision-making, provides a declaration reference for each participant, thereby increasing the cleared power quantity, optimizing the power quantity allocation, and improving the allocation efficiency.

[0007] A power quantity regulation method based on quantum algorithms and neural networks in a power market environment proposed by the present invention constructs a population by initializing multiple quantum registers. The quantum register is an individual representing market decision-making, and the market decision-making is a set of declared volume-price curves of all participants in the power market;

[0008] Select the quantum encoding of a participant from each quantum register as a candidate solution, evaluate the fitness of the candidate solution, and update and mutate the half of the quantum registers with smaller candidate solution fitness to iterate the population;

[0009] When the population iteration times reach the set threshold, select the individual that meets the constraint conditions of each participant and has the maximum optimization objective as the optimal declaration strategy; through the optimal declaration strategy, the quantity-price decision of each participant is obtained by inverse quantum coding; the optimization objective of the individual's market decision is to maximize the sum of producer surplus and consumer surplus.

[0010] Preferably, the participants include power plants, virtual power plants, and grid enterprises; the fitness value of a power plant is the electricity sales profit under its declared quantity-price curve, the fitness value of a virtual power plant is the power generation cost under its declared quantity-price curve, and the fitness value of a grid enterprise is the sum of the electricity purchase cost and transportation cost under its declared quantity-price curve.

[0011] Preferably, a large model is used to evaluate the fitness of candidate solutions. The large model is trained based on the historical data of the electricity market for evaluating the fitness of the declared quantity and price of participants; the training samples of the large model are the quantum state encodings of the declared quantity-price curves of participants with marked fitness values; the large model uses a quantum neural network.

[0012] Preferably, the electricity sales profit of a power plant is the difference between the total electricity sales price and the power generation cost. The total electricity sales price is the product of the total electricity sales volume and the clearing price; the clearing price is the electricity price that maximizes the optimization objective when the declared electricity quantities of participants are known.

[0013] Preferably, the optimization objective H(t) of the market decision at time t is:

[0014]

[0015] q su,m (t) represents the total cleared electricity quantity of producer m at time t, that is, the sum of the declared electricity quantities of producer m in each period at time t; q(t, b) represents the cleared electricity quantity in the b-th period at time t; λ su,m (q(t, b)) represents the declared price when the cleared electricity quantity of producer m in the b-th period at time t is q(t, b); producers include power plants and virtual power plants;

[0016] q co,j (t) represents the total cleared electricity quantity of grid enterprise j at time t, that is, the sum of the declared electricity quantities of grid enterprise j in each period at time t; λ co,j (q(t, b)) represents the declared price when the purchased electricity quantity of grid enterprise j in the b-th period at time t is q(t, b).

[0017] Preferably, the way to iterate the population is as follows: copy the qubits of the half of the quantum registers with larger candidate solution fitness to the half of the quantum registers with smaller candidate solution fitness, pair up the updated quantum registers two by two, and exchange half of the qubits of each pair of quantum registers; rotate the qubits in the exchanged quantum registers; all quantum registers form a mutant population.

[0018] Preferably, when iterating the population, after copying and updating the quantum registers with relatively small fitness values of the candidate solutions, according to the sorting of the fitness values of the candidate solutions before the update, the qubits in the latter half of the adjacent two quantum registers are exchanged.

[0019] Preferably, the way to rotate the qubits in the exchanged quantum register is as follows: extract qubits from the quantum register according to the set rotation ratio, and use any one or more of the Rx, Ry, and Rz gates to rotate the extracted qubits, and the rotation angle of each gate is a set value.

[0020] The present invention also proposes a power quantity regulation system and a storage medium based on a quantum algorithm and a neural network in a power market environment, which provide a carrier for the above-mentioned power quantity regulation method based on a quantum algorithm and a neural network in a power market environment, facilitating the popularization and application of this method.

[0021] A power quantity regulation system based on a quantum algorithm and a neural network in a power market environment proposed by the present invention includes a memory and a processor. A computer program is stored in the memory, and the processor is connected to the memory. The processor is configured to execute the computer program to implement the above-mentioned power quantity regulation method based on a quantum algorithm and a neural network in a power market environment.

[0022] A storage medium proposed by the present invention stores a computer program, and when the computer program is executed, it is used to implement the above-mentioned power quantity regulation method based on a quantum algorithm and a neural network in a power market environment.

[0023] The advantages of the present invention are as follows:

[0024] (1) A power quantity regulation method based on a quantum algorithm and a neural network in a power market environment proposed by the present invention uses a quantum register as an individual representing market decisions, realizes the mutation of market decision policies through population iteration, and then combines the market surplus as an optimization goal to select the optimal market decision. The present invention obtains information of power sellers and power buyers based on power market participants, and then generates the declared quantity-price curves of power sellers and power buyers based on the optimization goal of power quantity allocation, fully considering the known market information, and realizing the optimal allocation of power quantity under the condition of complete information. Moreover, the present invention combines population iteration with quantum computing, greatly improving the decision-making efficiency.

[0025] (2) The present invention sets a fitness for evaluating the adaptability of the declared volume-price curve of participants in the market. During the population iteration process, mutations are preferentially performed on individuals with smaller fitness, that is, quantum registers. As the population iterates, more market decisions with high fitness are gradually generated, providing more and better selection objects for the subsequent screening of the optimal declaration decision, thus ensuring the high adaptability of the optimal declaration decision to the market, ensuring the accuracy of short-term prediction and the high-level balance between the power supply and demand sides, ensuring the power sales revenue of the power seller and the production capacity of the power purchaser, and improving the optimization and allocation efficiency.

[0026] (3) In the present invention, participants are divided into power plants, virtual power plants, and grid enterprises, and different fitness representation methods are formulated according to the characteristics of each type of participant, thus ensuring the accuracy of fitness calculation.

[0027] (4) The present invention takes the sum of producer surplus and consumer surplus as the optimization goal, closely adheres to the purpose of optimizing the allocation of power resources, and ensures the superiority of the final decision.

[0028] (5) The present invention constructs a population based on quantum registers and calculates the fitness value in combination with a quantum neural network, shortening the calculation time of the fitness value, greatly improving the iteration speed and efficiency of the population, and improving the power allocation calculation efficiency. BRIEF DESCRIPTION OF THE DRAWINGS

[0029] Figure 1 is a flowchart of a power quantity regulation method based on a quantum algorithm and a neural network in a power market environment;

[0030] Figure 2 is a detailed flowchart of a power quantity regulation method based on a quantum algorithm and a neural network in a power market environment. DETAILED DESCRIPTION OF THE INVENTION

[0031] Next, the technical solutions in the embodiments of the present invention will be clearly and completely described in conjunction with the accompanying drawings in the embodiments of the present invention. Obviously, the described embodiments are only a part of the embodiments of the present invention, rather than all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those of ordinary skill in the art without creative efforts shall fall within the protection scope of the present invention.

[0032] Participants in the power futures market include: power sellers and power purchasers; at the same moment, the power sales volume of the power seller is equal to the power purchase volume of the power purchaser.

[0033] Specifically, power sellers include power plants and virtual power plants; power purchasers are mainly grid enterprises.

[0034] The set of power plants is denoted as I, the set of virtual power plants is denoted as O, and the set of grid enterprises is denoted as J.

[0035] Reference Figure 1 、 Figure 2 In the present invention, the combination of the declared volume-price curves of all participants at a single moment is used as the market decision. First, a quantum variational convolutional neural network for evaluating the fitness of the market decision is trained based on historical data. Its input is the quantum encoding of the market decision at a single moment, and the output is the fitness of the market decision. Then, a quantum register representing the market decision is constructed and initialized. Using the quantum register as a population individual, a genetic algorithm is used for population iteration. During the population iteration process, the mutation is performed on the half of the individuals with smaller fitness values;

[0036] Calculate the optimization objectives of each population, and select the population with the largest optimization objective as the optimal declaration strategy.

[0037] The training samples {|Ψ α ,|y α >} of the variational quantum convolutional neural network are constructed by combining the historical data of the electricity futures market, where |Ψ α > is the quantum state encoding of the declared volume-price curve at a certain moment, α is the participant in the electricity futures market, α ∈ I ∪ J ∪ O; |y α > represents the fitness corresponding to the declared volume-price curve;

[0038] The declared volume-price curve is denoted as [S α1 ,S α2 ,…,S αN ;j α1 ,j α2 ,…,j αN , S α1 , S α2 and S αN respectively represent the declared electricity quantities of the participant α in the electricity futures market at the 1st, 2nd, …, Nth time periods at the corresponding moment, j α1 , j α2 and j αN respectively represent the declared electricity prices of the participant α at the 1st, 2nd, …, Nth time periods at the corresponding moment; N is the number of time periods divided by a moment;

[0039]

[0040] C(q cg,o (t),u cg,o (t))=U(q cg,o (t))×u cg,o (t)

[0041] C(q wg,o (t),u wg,o (t))=U(q wg,o (t))×u wg,o (t)

[0042] C(q curt,o (t),u curt,o (t)) = U(q curt,o (t)) × u curt,o (t)

[0043] Among them, p i represents the winning bid price of power plant i, that is, the electricity clearing price at this moment; P i represents the winning bid electricity quantity of power plant i, and C(P i ) represents the power generation cost of power plant i;

[0044] C(q cg,o (t),u cg,o (t)) is the actual power generation cost of virtual power plant o, U(q cg,o (t)) is the power generation cost of the virtual power plant, u cg,o (t) is the power generation start flag variable, 1 represents start, and 0 represents not started; C(q wg,o (t),u wg,o (t)) is the actual cost of wind and light curtailment inside the virtual power plant, U(q wg,o (t)) is the cost of wind and light curtailment inside the virtual power plant, u wg,o (t) is the wind and light curtailment cost coefficient; C(q curt,o (t),u curt,o (t)) is the actual cost of load interruption inside the virtual power plant, U(q curt,o (t)) is the cost of load interruption inside the virtual power plant, u curt,o (t) is the load interruption flag variable inside the virtual power plant, 1 represents interrupted load, and 0 represents non-interrupted load;

[0045] P j,b represents the electricity price declared by grid enterprise j in the b-th period at this moment, Q j,b represents the electricity purchase quantity declared by grid enterprise j in the b-th period at this moment; γ j,b is a 0-1 variable representing the winning bid situation of grid enterprise j in the b-th period, 0 represents not winning the bid, and 1 represents winning the bid; S w represents the unit price of transmission cost of the w-th transmission line; B represents the set of periods at a single moment, B = {1, 2, ……, N}; W represents the set of transmission line segments;

[0046] It can be seen that for a power plant, the fitness |y α > of the quantum state encoding |Ψ α > of its declared quantity-price curve at a single moment is the electricity sales profit at this moment;

[0047] For a virtual power plant, the quantum state encoding |Ψ αFitness |y α is the power generation cost at this moment;

[0048] For grid enterprises, the quantum state encoding |Ψ of the declared quantity-price curve at a single moment α Fitness |y α is the sum of the electricity purchase cost and the transportation cost at this moment.

[0049] The optimization objective of the electricity clearing price R(t) at time t is to maximize the total sum H(t) of consumer surplus and producer surplus;

[0050]

[0051] q su,m (t) represents the total cleared electricity of producer m at time t, that is, the sum of the declared electricity quantities of producer m in each period at time t; q(t, b) represents the cleared electricity quantity in the b-th period at time t; λ su,m (q(t, b)) represents the declared price when the cleared electricity quantity of producer m in the b-th period at time t is q(t, b); Producers include power plants and virtual power plants;

[0052] q co,j (t) represents the total cleared electricity of grid enterprise j at time t, that is, the sum of the declared electricity quantities of grid enterprise j in each period at time t; λ co,j (q(t, b)) represents the declared price when the purchased electricity quantity of grid enterprise j in the b-th period at time t is q(t, b).

[0053] In this embodiment, a quantum variational neural network is first trained on the training samples {|Ψ α >, |y α >}. The input of the quantum variational neural network is the quantum encoding of the declared quantity-price curves of each electricity futures market participant at a single moment, and the output of the quantum variational neural network is the fitness of the input.

[0054] In this embodiment, a quantum genetic algorithm is used to optimize the declared plans of each participant at a single moment. The optimization process of the declared quantity-price curve includes the following steps:

[0055] S1. Initialize the basic parameters of the quantum genetic algorithm, including: the number d of quantum registers, the maximum number τ of iterations, the length r of the quantum chromosome, and the quantum rotation angle ε;

[0056] d quantum registers constitute the population of the quantum genetic algorithm; A single quantum register is an individual representing the market decision, that is, the quantum state representing the set of the declared quantity-price curves of all participants;

[0057] In a specific embodiment, it can be set that: d is an integer multiple of 4, and r is an even number;

[0058] S2, initialize d quantum registers to random quantum states to form an initial population; calculate the optimization target H(t) of each quantum register in the population;

[0059] The quantum bit of a single quantum register is denoted by {δ α ; α∈I∪O∪J}: α represents the participants in the electricity futures market, I represents the set of power plants, O represents the set of virtual power plants, and J represents the set of power grid companies;

[0060] δ α =[|c α1 >|c α2 >…|c αr >]

[0061] δ α represents the gene encoded on participant α, that is, the quantum encoding of the declared quantity-price curve of participant α; c α1 、c α2 and c αr Represents gene δ α The states of the 1st, 2nd, and rth quantum bits in ;

[0062] S3. Select a participant's declared quantity-price curve from each quantum register as a candidate solution, input the candidate solution into the trained variational quantum neural network, and predict the fitness of the candidate solution;

[0063] S4. Sort the quantum registers in descending order of the fitness of the candidate solutions. The fitness of the candidate solutions of the first half of the quantum registers is greater than that of the candidate solutions of the second half of the quantum registers. Reset the second half of the quantum registers to the reference state, and use the quantum cloning machine to copy the quantum bits of the first half of the quantum registers to the second half of the quantum registers.

[0064] That is, the quantum bits of the half of the quantum registers with larger fitness values ​​are copied to the half of the quantum registers with smaller fitness values.

[0065] S5. Perform a crossover operation in the updated second half of the quantum registers, that is, pair the second half of the quantum registers in pairs, and extract half of the quantum bits from each pair of quantum registers for exchange.

[0066] In this implementation, a set of transition parameters a is constructed during the specific implementation, a∈{1,2,…,d / 4}; all a are traversed, and the last r / 2 qubits of the d / 2+2a-1th quantum register and the last r / 2 qubits of the d / 2+2ath quantum register are exchanged using a quantum exchange gate:

[0067] S6. Rotate each qubit in the latter half of the quantum register after crossover with the same probability to achieve quantum gene mutation and generate a mutant population containing d quantum registers, and calculate the optimization objective H(t) of each quantum register in the mutant population;

[0068] The specific quantum gene mutation is as follows: Use any one or more of the Rx, Ry, and Rz gates to operate on any qubit in the quantum register, and the rotation angle of each gate is the quantum rotation angle ε;

[0069] For each mutation of the population, half of the individuals remain unchanged to ensure the heredity of the population. In the mutant population, only the mutated individuals need to be calculated for the optimization objective. For the individuals that remain unchanged, their optimization objectives can also directly adopt the calculated values before population iteration.

[0070] S7. Determine whether the number of population mutation times reaches the maximum iteration number τ;

[0071] If not, return to step S3;

[0072] If yes, select the quantum register that satisfies the participant's constraint conditions and corresponds to the maximum optimization objective H(t) as the optimal declaration strategy. Under the optimal declaration strategy, the fitness of the participant is the highest, and the optimal declaration strategy is retained as the participant's declaration sample.

[0073] That is, perform inverse quantum encoding on the optimal declaration strategy to obtain the declaration quantity-price curves of each participant in the final declaration strategy.

[0074] It should be noted that in order to avoid situations where the declared price is too high or too low, or the declared electricity quantity is too much or too little in the participant's declaration quantity-price decision-making, the participant can pre-set the declared price range and declared electricity quantity range for each time period as constraint conditions; so as to finally select the market decision corresponding to the maximum optimization objective, that is, the individual, under the condition of satisfying the constraint conditions of each participant.

[0075] In specific implementation, during population initialization and population mutation, the coding genes of individuals can be restricted so that the declared quantity and price of the participant should always meet the corresponding constraint conditions; or population mutation can be performed without constraints, and then after the population iteration is completed, first screen the feasible individuals according to the participant's constraint conditions, and then select the individual corresponding to the maximum optimization objective among the feasible individuals as the optimal declaration strategy.

[0076] Of course, for those skilled in the art, the present invention is not limited to the details of the above exemplary embodiments, but also includes the same or similar structures that can be implemented in other specific forms without departing from the spirit or basic characteristics of the present invention. Therefore, from any point of view, the embodiments should be regarded as exemplary and non-limiting. The scope of the present invention is defined by the appended claims rather than the above description. Therefore, all changes falling within the meaning and scope of the equivalent elements of the claims are intended to be embraced within the present invention. Any reference signs in the claims should not be construed as limiting the claims involved.

[0077] For specific implementation, the moment t can be defined at 96 moments, that is, the 24 hours of a day are divided into 96 moments, each moment being 15 minutes, and each moment is further divided into multiple time periods. In this way, the declared volume-price mapping relationship for each time period at each moment forms the declared volume-price curve at that moment.

[0078] In addition, it should be understood that although this specification is described in terms of embodiments, not every embodiment only contains an independent technical solution. This narrative manner of the specification is only for clarity. Those skilled in the art should regard the specification as a whole, and the technical solutions in each embodiment can also be appropriately combined to form other embodiments that can be understood by those skilled in the art.

[0079] The technologies, shapes, and structures not detailedly described in the present invention are all well-known technologies.

Claims

1. A method for power control based on quantum algorithm and neural network in power market environment, characterized in that: Construct multiple quantum registers to form a population and initialize them. The quantum registers are individuals that represent market decisions. Market decisions are the collection of declared quantity and price curves of all participants in the electricity market. Select a participant's quantum code from each quantum register as a candidate solution, evaluate the fitness of the candidate solution, and update and mutate half of the quantum registers with smaller fitness of the candidate solution to iterate the population; When the number of population iterations reaches the set threshold, the individual that meets the constraints of each participant and has the largest optimization goal is selected as the optimal declaration strategy; the declaration quantity and price decisions of each participant are obtained through the inverse quantum coding of the optimal declaration strategy; the optimization goal of the individual market decision is to maximize the sum of producer surplus and consumer surplus.

2. The method for controlling power based on quantum algorithm and neural network in the power market environment according to claim 1 is characterized in that: Participants include power plants, virtual power plants and power grid companies; the fitness value of a power plant is the profit from electricity sales under its declared quantity-price curve, the fitness value of a virtual power plant is the cost of electricity generation under its declared quantity-price curve, and the fitness value of a power grid company is the sum of its electricity purchase cost and transportation cost under its declared quantity-price curve.

3. The method for controlling power based on quantum algorithm and neural network in the power market environment according to claim 2 is characterized in that: A large model is used to evaluate the fitness of candidate solutions. The large model is trained based on historical data of the electricity market to evaluate the fitness of participants' declared quantities and prices. The training samples of the large model are quantum state encodings of the participants' declared quantity and price curves with marked fitness values. The large model uses a quantum neural network.

4. The method for power control based on quantum algorithm and neural network in the power market environment as claimed in claim 2 is characterized in that: The profit of a power plant from selling electricity is the difference between the total electricity selling price and the cost of power generation. The total electricity selling price is the product of the total electricity selling volume and the clearing price. The clearing price is the electricity price that maximizes the optimization objective when the electricity volume declared by the known participants is known.

5. The method for power control based on quantum algorithm and neural network in the power market environment as claimed in claim 4 is characterized in that: The optimization goal H(t) of market decision at time t is: q su,m (t) represents the total electricity cleared by producer m at time t, that is, the total electricity reported by producer m in each time period at time t; q(t,b) represents the electricity cleared in the bth time period at time t; λ su,m (q(t,b)) represents the declared price when the cleared electricity quantity of producer m is q(t,b) in the bth period at time t; producers include power plants and virtual power plants; q co,j (t) represents the total cleared electricity of grid enterprise j at time t, that is, the total electricity declared by grid enterprise j in each period at time t; co,j (q(t,b)) represents the declared price when the power grid enterprise j purchases the power q(t,b) in the bth period at time t.

6. The method for power control based on quantum algorithm and neural network in the power market environment according to claim 1 is characterized in that: The method of iterating the population is as follows: the quantum bits of half of the quantum registers with larger candidate solution fitness are copied to half of the quantum registers with smaller candidate solution fitness, and the updated quantum registers are paired, and half of the quantum bits of each pair of quantum registers are extracted for exchange; the quantum bits in the exchanged quantum registers are rotated; all quantum registers form a mutant population.

7. The method for power control based on quantum algorithm and neural network in the power market environment according to claim 6 is characterized in that: When iterating the population, after the half of the quantum registers with smaller candidate solution fitness are copied and updated, the candidate solutions are sorted according to their fitness before the update, and the quantum bits of the second half of the two adjacent quantum registers are exchanged.

8. The method for power control based on quantum algorithm and neural network in the power market environment as claimed in claim 6 is characterized in that: The method of rotating the quantum bits in the exchanged quantum register is as follows: extracting quantum bits from the quantum register according to the set rotation ratio, and rotating the extracted quantum bits using any one or more gates of Rx, Ry, and Rz, with the rotation angle of each gate being a set value.

9. A power control system based on quantum algorithm and neural network in the power market environment, characterized in that: It includes a memory and a processor, the memory stores a computer program, the processor is connected to the memory, and the processor is used to execute the computer program to implement the power control method based on quantum algorithm and neural network in the power market environment as described in any one of claims 1 to 8.

10. A storage medium, characterized in that: A computer program is stored, and when the computer program is executed, it is used to implement the power control method based on quantum algorithm and neural network in the power market environment as described in any one of claims 1 to 8.

Images