Virtual power plant low-carbon optimization operation method considering carbon transaction

Abstract

The invention relates to the technical field of optimization operation, in particular to a virtual power plant low-carbon optimization operation method considering carbon transaction, which comprises the following steps: S1, virtual power plant framework design and modeling: designing an operation transaction framework of a low-carbon virtual power plant, and modeling each unit in the virtual power plant; s2, depicting the uncertainty of wind and light output: depicting the uncertainty of wind and light output through scene generation and reduction; s3, obtaining an income and carbon emission optimization objective: constructing an objective function taking income maximization and carbon emission minimization as objectives; s4, optimizing an output scheme by using a DQN algorithm: optimizing a next-day output scheme by using the DQN algorithm; s5, a deviation punishment mechanism is implemented, wherein when the actual wind and light output deviates from the prediction, the deviation punishment mechanism is implemented; according to the method, the next-day output prediction and scheduling optimization capability of the virtual power plant is improved, and achievement of revenue maximization and carbon emission minimization is ensured.

Description

Technical Field

[0001] The present invention relates to the field of optimized operation technology, and in particular to a low-carbon optimized operation method of a virtual power plant taking carbon trading into consideration. Background Art

[0002] In recent years, distributed energy generation based on wind and solar renewable energy has developed rapidly around the world. Although these distributed power sources based on renewable energy can reduce the use of thermal power units to a certain extent, distributed energy often has defects such as large random volatility and small single-machine capacity, making it difficult to participate independently in power market operations. The emergence of virtual power plants has effectively solved the problems of volatility and capacity of distributed power sources.

[0003] Virtual power plants can use the flexible adjustment characteristics of their energy storage equipment to compensate for the deviation in wind and solar power output, minimize the uncertainty of overall output and the penalty for unbalanced electricity, and ensure overall benefits. On the one hand, virtual power plants optimize the coordinated operation of their internal resources according to their possible real-time output, control their carbon emissions, and participate in carbon market transactions with their remaining carbon emission quotas to obtain more benefits. It can be seen that the electricity trading and carbon trading of virtual power plants have coupling characteristics, and it is necessary to coordinate and optimize their operations in the electricity market and the carbon market to improve overall benefits. As a new type of market entity, virtual power plants can further promote the development of new energy and exert their environmental benefits by participating in carbon trading, thereby improving the overall benefits of virtual power plants. Therefore, studying the optimal operation strategy of virtual power plants that coordinate electricity trading and carbon trading has important engineering application value.

[0004] In response to the problems existing in the low-carbon operation optimization method of virtual power plants mentioned above, a low-carbon optimization operation method of virtual power plants considering carbon trading is proposed, and the optimization operation method of virtual power plants is determined. Summary of the invention

[0005] The present invention provides a low-carbon optimized operation method of a virtual power plant taking carbon trading into consideration.

[0006] A low-carbon optimization operation method of a virtual power plant considering carbon trading comprises the following steps:

[0007] S1, Virtual Power Plant Framework Design and Modeling: Design the operation and trading framework of the low-carbon virtual power plant and model each unit in the virtual power plant;

[0008] S2, characterization of wind and solar power output uncertainty: characterizes the uncertainty of wind and solar power output through the generation and reduction of scenarios;

[0009] S3, benefit and carbon emission optimization objectives: construct an objective function with the goal of maximizing benefits and minimizing carbon emissions;

[0010] S4, DQN optimizes the output plan: Use the DQN algorithm to optimize the next day's output plan;

[0011] S5, implementation of the deviation penalty mechanism: when there is a deviation between the actual wind and solar power output and the forecast, the deviation penalty mechanism is implemented.

[0012] Optionally, the virtual power plant framework design and modeling in S1 includes:

[0013] S11, Dispatchable Generator Set Modeling: The dispatchable generator set consists of gas turbines. Based on the minimum and maximum power, minimum and maximum output, and set time period, the dispatchable generator set modeling is combined with the startup and shutdown costs of the gas turbines, which is expressed as:

[0014]

[0015] in, They represent the minimum and maximum electrical energy generated by the gas turbine, They represent the minimum and maximum output of the gas turbine respectively, and τ is the set time period;

[0016]

[0017] Among them, n j is the number of linearization segments of the cost function, take n j =3, C su , C sd are the cost of starting and shutting down the gas turbine, K ij is the slope of the linearization segment;

[0018] S12, Energy storage device modeling: Energy storage device modeling is based on the state transition equation. The electric energy generated is determined by the operating state of the energy storage device itself, which is expressed as:

[0019]

[0020] in, is the electrical energy in the energy storage device at time t, is the electric energy in the energy storage device at time t-1, P t SC (t) is the charging power of the energy storage device, is the discharge power of the energy storage device, η SC , η SD are the charging and discharging efficiency of the energy storage device respectively;

[0021] S13, Wind power modeling: Wind power output is determined by wind speed, and the wind speed distribution function satisfies the Weibull distribution, which is expressed as:

[0022]

[0023] Where v is the actual wind speed, is the shape parameter of wind power, θ is the scale parameter of wind turbine;

[0024] where v in <v t <v out ;

[0025] Among them, P WP is the maximum output of the wind turbine, is the rated power of the wind turbine, v t is the actual wind speed of the wind turbine, v rat It is the rated wind speed during normal operation of the wind turbine;

[0026] S14, Photovoltaic modeling: Photovoltaic output is determined by light intensity and climate. The light intensity conforms to the Beta distribution function, which is expressed as:

[0027]

[0028] Among them, θ is the light intensity of the photovoltaic unit, α is the shape parameter of the distribution function, and β is the distribution parameter of the distribution function;

[0029]

[0030] Among them, P PV is the maximum output of the photovoltaic unit, is the rated power of the photovoltaic system, θ t is the actual light intensity of the photovoltaic unit.

[0031] Optionally, the uncertainty characterization of wind and solar power output in S2 includes:

[0032] S21, scenario generation: Latin hypercube sampling is used to generate a power scenario set for fluctuating power sources, the error probability distribution is divided into multiple equal probability intervals, and a sample is randomly selected in each interval to finally generate a scenario set;

[0033] S22, scene reduction: A fast backward reduction method based on probability distance calculates the distance between scenes, gradually merges similar scenes and updates the probability until the number of scenes meets the requirement.

[0034] Optionally, the scene generation in S21 includes:

[0035] S211, error probability distribution division: distribute the error probabilities of wind power and photovoltaic power generation in m equal probability spaces, and the probability of each interval is 1 / m;

[0036] S212, randomly select probability points: in the i-th probability interval Randomly select a point p from i , expressed as:

[0037]

[0038] Where s is a randomly distributed number between [0,1];

[0039] S213, generate a scene set: randomly select probability points in each probability interval in turn, obtain m groups of sampling samples, and finally obtain a scene set M.

[0040] Optionally, the scene reduction in S22 includes:

[0041] S221, calculate geometric distance: calculate the geometric distance between two scenes in the scene set M, expressed as:

[0042]

[0043] Where, d is the Euclidean distance;

[0044] S222, find the nearest scene pair: find the scene j that is closest to scene i from the scene set M, that is, min{d(m i ,m j ),i≠j};

[0045] S223, minimum probability distance scene selection: repeatedly calculate the geometric distance, find the closest scene pair, and obtain the scene i with the smallest sum of probability distances with the remaining scenes j;

[0046] S224, update scene probability: update the scene probability, the new scene probability is equal to the probability of i plus the probability of j, and obtain a new scene set M;

[0047] S225, repeat the scene reduction process: repeatedly calculate the geometric distance, find the nearest scene pair, select the scene with the minimum probability distance, and update the scene probability until the number of remaining scenes meets the requirement of the number of scene reductions.

[0048] Optionally, the revenue and carbon emission optimization objectives in S3 include:

[0049] S31, benefit maximization objective function: Construct an objective function with benefit maximization as the goal, expressed as:

[0050]

[0051] Among them, F is the total profit obtained by the virtual power plant from participating in electricity market transactions, is the day-ahead market electricity sales revenue of the virtual power plant during period t, is the income or cost of the virtual power plant participating in carbon market transactions during period t, is the profit or cost caused by the deviation between the daily power of the virtual power plant and the planned power during period t, They are the maintenance cost of renewable energy units, energy storage system cost, demand response cost and gas turbine cost during period t respectively;

[0052]

[0053] in, is the day-ahead electricity market price during period t, P t DA is the power sold by the virtual power plant to the day-ahead power market during period t, P t ESS is the charging and discharging power of the energy storage system in the period t the day before, is the power sold by the gas turbine in the virtual power plant to the day-ahead power market during period t, P t DR is the power that the adjustable load in the virtual power plant responds to in the day-ahead power market during period t;

[0054]

[0055] Among them, γ A is the price of free carbon emission allowance in the carbon trading market, γ BE The price of the carbon emission reduction share generated by renewable energy units in the CCER trading market;

[0056]

[0057] in, is the deviation penalty electricity price during period t, ε + , ε - is the deviation penalty coefficient;

[0058]

[0059] in, are the operating cost coefficients of wind turbines and photovoltaic units, are the actual power generation of the i-th wind turbine and photovoltaic unit in period t, k is the number of wind turbines, and q is the number of photovoltaic units;

[0060]

[0061] Among them, δ ESS is the unit charging and discharging cost of the energy storage system;

[0062]

[0063] in, is the operation management coefficient of the gas turbine, is the unit power generation fuel cost of the gas turbine;

[0064]

[0065] in, is the demand response compensation coefficient, P t DR is the demand response power in the t period the day before;

[0066] S32, Carbon emission minimization objective function: Construct an objective function with carbon emission minimization as the goal, expressed as:

[0067]

[0068] Among them, CAR is the total carbon emissions generated during the operation of the virtual power plant. t is the carbon emissions of the virtual power plant during period t.

[0069] Optionally, the constraints of the revenue maximization objective function and the carbon emission minimization objective function include:

[0070] Gas turbine constraints:

[0071]

[0072] Among them, v k is the operating state of gas turbine k, 1 means start and 0 means shut down. is the output of gas turbine k participating in the day-ahead market during time period t, Δ MT is the ramp constraint of the gas turbine, is the output of gas turbine k participating in the balancing market in the scenario ω in time period t, P MT,max , P MT,min They represent the minimum output of the gas turbine, is the limit of the regulating power of gas turbine k;

[0073] Energy storage device constraints:

[0074]

[0075] in, is the stored energy of battery q in time period t, represents the charging power of battery q in time period t, represents the discharge power of battery q in time period t, η SC Indicates the charging loss of the battery, P SCmax Indicates the maximum charge limit of the battery, P SDmax Indicates the maximum discharge limit of the battery, Q min , Q maxIndicates the minimum and maximum storage capacity of the battery;

[0076] Power balance constraints:

[0077]

[0078] in, They represent the output of wind power m and photovoltaic power n in the day-ahead market period t, They represent the power of battery q participating in the day-ahead market charging and discharging during period t, is the output of micro gas turbine k in the day-ahead market during period t;

[0079] Power balancing constraints of VPP in the balancing market:

[0080]

[0081] in, They represent the output of wind power m and photovoltaic power n in the balancing market in the scenario ω of time period t, They represent the positive and negative deviations between the actual output of the virtual power plant and the declared power, respectively. They represent the power of battery q participating in the charging and discharging of the balancing market in the scenario ω during time period t, The output of micro gas turbine k participating in the balancing market in scenario ω during time period t.

[0082] Optionally, the DQN optimization output solution in S4 includes:

[0083] S41, DQN algorithm parameter initialization: Initialize the learning rate, number of training times, and neural network weights in the DQN algorithm;

[0084] S42, obtaining the initial state of the virtual power plant: starting the first virtual power plant scheduling, obtaining the environmental state of the virtual power plant in operation, that is, the initial state of each device in the virtual power plant;

[0085] S43, the agent takes scheduling actions: the agent takes actions under the constraints, including discharging energy storage and increasing the operating power of the gas unit, calculates the reward under the current action, that is, the target value of the virtual power plant scheduling model, and updates the current environmental state, and puts the current work, reward, and environmental state experience values ​​into the experience pool;

[0086] S44, experience pool sample accumulation: the agent repeatedly takes scheduling actions until the sample capacity in the experience pool reaches the set threshold;

[0087] S45, experience pool sampling and neural network weight update: randomly extracting a predetermined number of samples from the experience pool to update the weights of the double Q neural network;

[0088] S46, performing the next round of training based on the new weights: performing the next round of training based on the updated neural network weights until the maximum number of training times is reached;

[0089] S47, find the best scheduling plan: find the scheduling plan with the largest objective function value in all iterations as the optimal output plan of the virtual power plant on the next day.

[0090] Optionally, the implementation of the deviation penalty mechanism in S5 includes:

[0091] S51, calculate deviation power: calculate the deviation power caused by the uncertainty of distributed power generation, and determine the difference between the actual power generation of the virtual power plant and the planned power generation on the day before, expressed as:

[0092]

[0093] Where ΔP t is the deviation power between the planned power generation in the day and the actual power generation in the day during period t, is the power generation capacity of the i-th renewable generator set in the t-th period of the day before;

[0094] S52, deviation adjustment: using energy storage equipment and gas turbines to compensate for the deviation of electricity, and reducing the deviation by discharging energy storage or adjusting the power of gas turbines;

[0095] S53, calculating the adjusted deviation power: calculating the remaining deviation power after the deviation is adjusted.

[0096] Optionally, the calculated adjusted deviation power in S53 is expressed as:

[0097]

[0098] Among them, P t RM It is the deviation power generation after compensation of the dispatchable units in the virtual power plant during period t. Its positive value indicates that the actual transaction power is greater than the previous day's bid power. Its negative value indicates that the actual transaction power is less than the previous day's bid power. It is necessary to purchase power from the real-time market to make up the difference. The power of the energy storage device to compensate for the bidding deviation of the renewable energy unit during the t period of the day, P is the power of the gas turbine to compensate for the bidding deviation of the renewable energy unit during the t period of the day, t DR,cs It is the power given to compensate the bidding deviation of renewable energy units in response to the adjustable load demand during period t within the day. μ is a ±1 variable. When μ=1, the ESS equipment is in a discharging state during period t. When μ=-1, the ESS equipment is in a charging state during period t.

[0099] Beneficial effects of the present invention:

[0100] The present invention, by designing the operation and trading framework of the virtual power plant and accurately modeling each unit, can effectively cope with the uncertainty of wind and solar power output, reduce the output deviation caused by the fluctuation of distributed power sources, and optimize the output plan of the virtual power plant by using scenario generation and reduction technology, combined with dispatchable units such as energy storage equipment and gas turbines, thereby improving the flexibility and accuracy of electricity market participation. At the same time, the introduction of the DQN algorithm further improves the next-day output prediction and scheduling optimization capabilities of the virtual power plant, ensuring the realization of maximum benefits and minimum carbon emissions.

[0101] The present invention effectively couples the electricity trading of virtual power plants with carbon trading by taking into account the mechanism of carbon trading. Through precise carbon emission control and trading, the overall economic benefits of virtual power plants are further improved. The implementation of the deviation penalty mechanism ensures that when the wind and solar power output does not match the forecast, the virtual power plant can be flexibly adjusted through the balancing market, reducing the losses caused by deviations and improving the economic benefits and environmental friendliness of operations. BRIEF DESCRIPTION OF THE DRAWINGS

[0102] In order to more clearly illustrate the technical solutions in the present invention or the prior art, the drawings required for use in the embodiments or the description of the prior art will be briefly introduced below. Obviously, the drawings in the following description are only for the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying creative work.

[0103] Figure 1 This is a schematic diagram of the operating method of an embodiment of the present invention;

[0104] Figure 2 A schematic diagram of the virtual power plant framework design and modeling according to an embodiment of the present invention. DETAILED DESCRIPTION

[0105] The present invention is described in detail below in conjunction with the accompanying drawings and specific embodiments. At the same time, it is explained here that in order to make the embodiments more detailed, the following embodiments are the best and preferred embodiments, and those skilled in the art may also adopt other alternatives to implement some known technologies; and the accompanying drawings are only for more specific description of the embodiments, and are not intended to specifically limit the present invention.

[0106] It should be noted that the references to "one embodiment", "an embodiment", "an exemplary embodiment", "some embodiments" and the like in the specification indicate that the embodiments described may include specific features, structures or characteristics, but not every embodiment may include the specific features, structures or characteristics. In addition, when a specific feature, structure or characteristic is described in conjunction with an embodiment, it should be within the knowledge of a person skilled in the art to implement such feature, structure or characteristic in conjunction with other embodiments (whether or not explicitly described).

[0107] In general, a term can be understood, at least in part, from its use in context. For example, depending, at least in part, on the context, the term "one or more" as used herein can be used to describe any feature, structure, or characteristic in the singular sense, or can be used to describe a combination of features, structures, or characteristics in the plural sense. Additionally, the term "based on" can be understood as not necessarily intended to convey an exclusive set of factors, but can instead, depending, at least in part, on the context, allow for the presence of other factors that are not necessarily explicitly described.

[0108] like Figure 1-Figure 2 As shown, a low-carbon optimization operation method of a virtual power plant considering carbon trading includes the following steps:

[0109] 1. Design the operation and trading framework of the low-carbon virtual power plant and model each unit in the virtual power plant.

[0110] From the perspective of modeling, the main components of VPP can be divided into four parts: dispatchable generators, controllable loads, distributed power sources and energy storage equipment. The dispatchable generators in VPP are usually small generators that generate electricity from fossil energy such as natural gas or oil, such as micro gas turbines, diesel generators, etc. The controllable loads include residential and industrial electricity users equipped with communication control equipment, which can induce the demand elasticity of VPP through time-of-use electricity prices or direct control. Distributed power sources mainly include distributed wind power and distributed photovoltaics, which can use renewable energy such as wind power and photovoltaics to obtain electricity. In addition, distributed wind power and distributed photovoltaics are not dispatchable, and there is a deviation between their actual output and predicted output. Energy storage equipment can reduce the imbalance costs generated by distributed wind power and distributed photovoltaics by adjusting its own operating status. The VPP studied in this paper is mainly composed of gas turbines, distributed photovoltaics, distributed wind power and energy storage equipment. The mathematical description of the four components is as follows:

[0111] (1) Gas turbine:

[0112]

[0113] In the formula, They represent the minimum and maximum electrical energy generated by the gas turbine respectively; They represent the minimum and maximum output of the gas turbine respectively; τ is the set time period.

[0114] The cost function of the gas turbine is expressed by (3):

[0115]

[0116] Where n j is the number of linearization segments of the cost function, take n j =3; C su , C sd are the cost of starting and shutting down the gas turbine respectively; K ij is the slope of the linearization segment.

[0117] (2) Energy storage equipment:

[0118] The modeling of energy storage devices is based on the state transition equation. The electric energy generated is determined by the operating state of the energy storage device itself. The energy storage device can both release and generate electric energy. The electric energy (4) of the energy storage device is expressed as:

[0119]

[0120] In the formula, is the electric energy in the energy storage device at time t; is the electric energy in the energy storage device at time t-1; P t SC (t) is the charging power of the energy storage device; is the discharge power of the energy storage device; η SC , η SD are the charging and discharging efficiencies of the energy storage device, respectively.

[0121] (3) Wind power:

[0122] Wind power output is random and uncertain, and is determined by wind speed. The wind speed distribution function satisfies the Weibull distribution, as shown in formula (5):

[0123]

[0124] Where v is the actual wind speed; is the shape parameter of wind power; θ is the scale parameter of wind turbine, which can be obtained from the historical data of wind turbine.

[0125] The actual output of wind power is shown in (6):

[0126]

[0127] Where P WP is the maximum output of the wind turbine; is the rated power of the wind turbine; v t is the actual wind speed of the wind turbine; v rat is the rated wind speed of the wind turbine when it is in normal operation. out , or below the cut-in wind speed v in When the wind turbines cannot work normally.

[0128] (4) Photovoltaic:

[0129] Photovoltaic output is random and uncertain, and is determined by light intensity and climate. Light intensity conforms to the Beta distribution function and satisfies formula (7):

[0130]

[0131]

[0132] Wherein, θ is the illumination intensity of the photovoltaic unit; α is the shape parameter of the distribution function; β is the distribution parameter of the distribution function, all of which can be obtained from the historical data of the photovoltaic unit.

[0133] The actual photovoltaic output is shown in (10):

[0134]

[0135] Where P PV The maximum output of the photovoltaic unit; is the rated power of the photovoltaic unit; θ t is the actual light intensity of the photovoltaic unit.

[0136] 2. Describe the uncertainty of wind power output through scene generation and reduction.

[0137] Currently, the scenario set generation method and probability interval method are commonly used to deal with the uncertainty of the system. This paper adopts the scenario set generation method to describe the uncertainty of wind power and photovoltaic power generation units. Since the scenario obtained by the scenario set generation method is too large and the calculation is large, it is necessary to reduce the basic scenario set and obtain a few scenario sets with typical representative significance.

[0138] (1) Scene generation:

[0139] This paper uses Latin hypercube sampling to generate a power scenario set for fluctuating power sources. Assuming that m samples are extracted, the specific steps are as follows:

[0140] Step 1: Distribute the error probabilities of wind power and photovoltaic power generation into m equally probable spaces, with the probability of each interval being 1 / m.

[0141] Step 2: In the i-th probability interval Randomly select a point p from i . p i The calculation formula is shown in formula (11):

[0142]

[0143] Where: s is a randomly distributed number between [0,1].

[0144] Step 3: Extract the probability points in the second step in each probability interval in turn, obtain m groups of sampling samples, and finally obtain the scene set M.

[0145] (2) Scene reduction:

[0146] This paper adopts fast backward reduction based on probability distance to perform scene reduction. The specific scene reduction operation steps are as follows:

[0147] Step 1: Calculate the geometric distance between each scene in the scene set M, as shown in formula (12):

[0148]

[0149] Where: d is the Euclidean distance;

[0150] Step 2: Find the scene j that is closest to scene i from the scene set M, that is, min{d(m i ,m j ),i≠j};

[0151] Step 3: Repeat the above steps to obtain the scene i with the smallest sum of probability distances to the remaining scenes j;

[0152] Step 4: Update the scene probability. The new scene probability is equal to the probability of i plus the probability of j, and the new scene set M is obtained.

[0153] Step 5: Repeat the above steps until the number of remaining scenes meets the requirement of scene reduction.

[0154] 3. Construct an objective function with the goal of maximizing benefits and minimizing carbon emissions.

[0155] (1) Objective function:

[0156] This paper builds a model with the goal of minimizing carbon emissions and maximizing benefits. The benefit maximization objective function is as follows:

[0157]

[0158] Where F is the total profit obtained by the virtual power plant from participating in electricity market transactions; is the day-ahead market electricity sales revenue of the virtual power plant during period t; The revenue or cost of the virtual power plant participating in carbon market transactions during period t; The profit or cost caused by the deviation between the daily power of the virtual power plant and the planned power during period t; They are the maintenance cost of renewable energy units, energy storage system cost, demand response cost and gas turbine cost during period t.

[0159] The actual carbon emissions of the virtual power plant come from the operation of the gas turbine. The carbon emission minimization objective function is shown in formula (14):

[0160]

[0161] Among them, CAR is the total carbon emissions generated during the operation of the virtual power plant; CAR t is the carbon emissions of the virtual power plant during period t.

[0162] The day-ahead market electricity sales revenue is shown in formula (15):

[0163]

[0164] in, is the day-ahead electricity market price during period t; P t DA P is the power sold by the virtual power plant to the day-ahead power market during period t; t ESS is the charging and discharging power of the energy storage system in the period t the day before; is the power sold by the gas turbine in the virtual power plant to the day-ahead power market during period t; t DR It is the power that the adjustable load in the virtual power plant responds to in the day-ahead electricity market during period t.

[0165] The benefits or costs of carbon trading. According to the Measures, the national certified voluntary emission reduction can be used to offset the settlement of carbon emission quotas each year, and the offset ratio shall not exceed 5% of the carbon emission quotas to be settled. Therefore, during the operation of the virtual power plant, the carbon emissions generated by the gas turbine are first offset by the free carbon emission quota. If the actual carbon emissions are greater than the free carbon emission quota, the CCER of the renewable energy unit can offset part of the carbon emissions, and the rest needs to be offset by purchasing carbon emission rights in the carbon trading market. At the same time, the excess free carbon emission quotas and CCERs can be traded in the carbon market to obtain income. The specific calculation formula is shown in formula (17):

[0166]

[0167] Among them, γ A is the price of free carbon emission allowance in the carbon trading market; BEThe price of the carbon emission reduction share generated by renewable energy units in the CCER trading market.

[0168] The profit or cost of the intraday market is shown in formula (18):

[0169]

[0170] in, is the deviation penalty electricity price during period t; + , ε - are the deviation penalty coefficients, which are 0.5 and 2.5 respectively.

[0171] The maintenance cost of renewable energy units is shown in formula (20):

[0172]

[0173] in, are the operating cost coefficients of wind turbines and photovoltaic units, respectively; are the actual power generation of the i-th wind turbine and photovoltaic unit in period t; k is the number of wind turbines; q is the number of photovoltaic units.

[0174] The energy storage operation cost is shown in formula (21):

[0175]

[0176] Among them, δ ESS It is the unit charging and discharging cost of the energy storage system.

[0177] The cost of the gas unit is shown in formula (22):

[0178]

[0179] in, is the operation management coefficient of the gas turbine; is the unit power generation fuel cost of the gas turbine.

[0180] The demand response cost is shown in formula (23):

[0181]

[0182] in, is the demand response compensation coefficient; P t DR is the demand response power in period t the day before.

[0183] (2) Constraints:

[0184] The gas turbine constraints are as follows:

[0185]

[0186] Among them, v k is the operating state of gas turbine k, 1 means start and 0 means shut down; is the output of gas turbine k participating in the day-ahead market during time period t; Δ MT is the ramp constraint of the gas turbine; P is the output of gas turbine k participating in the balancing market in the scenario ω in time period t; MT,max , P MT,min They represent the minimum output of the gas turbine respectively; is the limit of the regulating power of gas turbine k.

[0187] The energy storage device constraints are as follows:

[0188]

[0189] in, is the stored energy of battery q in time period t; represents the charging power of battery q in time period t; represents the discharge power of battery q in time period t; η SC Indicates the charging loss of the battery; P SCmax Indicates the maximum charging limit of the battery; P SDmax Indicates the maximum discharge limit of the battery; Q min , Q max Indicates the minimum and maximum battery capacity.

[0190] The power balance constraints are as follows:

[0191]

[0192] in, They represent the output of wind power m and photovoltaic power n in the day-ahead market period t respectively; They represent the charging and discharging power of battery q in the day-ahead market during period t respectively; is the output of micro gas turbine k in the day-ahead market during period t.

[0193] The power balancing constraints of VPPs in the balancing market are as follows:

[0194]

[0195] in, They represent the output of wind power m and photovoltaic power n in the scenario ω of time period t when they participate in the balancing market; Respectively represent the positive and negative deviations between the actual output of the virtual power plant and the declared power; They represent the power of battery q participating in charging and discharging in the balancing market in the scenario ω during time period t; The output of micro gas turbine k participating in the balancing market in scenario ω during time period t.

[0196] 4. Use the DQN algorithm to optimize the next-day effort plan.

[0197] In the day-ahead optimization scheduling stage, the DQN algorithm solves the model based on the 24h power forecast values ​​of wind power and photovoltaics. However, in the intraday optimization scheduling stage, the actual 24h wind power and photovoltaic power curves cannot be directly known. Therefore, rolling optimization is performed in hours during intraday optimization scheduling. In each stage, the DQN algorithm adjusts the output action of the adjustable units in the virtual power plant according to the actual power generation of wind power and photovoltaics at the current moment and the forecast values ​​at the next two moments, reducing deviation penalties and achieving optimal scheduling. The solution process of the virtual power plant optimization scheduling solution based on the DQN algorithm is as follows:

[0198] Step 1: Initialize the learning rate, number of trainings, neural network weights and other parameters in the DQN algorithm:

[0199] Step 2: Start the first virtual power plant dispatch and obtain the environmental status of the virtual power plant during operation, that is, the initial status of each device in the virtual power plant;

[0200] Step 3: The agent takes actions under constraints, such as discharging energy storage and increasing the operating power of gas units; calculates the reward for the current action, i.e., the target value of the virtual power plant scheduling model, and updates the current environmental state; puts the current set of experience values ​​of work, reward, and environmental state into the experience pool;

[0201] Step 4: Repeat step 3 to obtain sufficient sample capacity in the experience pool;

[0202] Step 5: When the number of samples in the experience pool is sufficient, randomly extract small batches of samples from the experience pool to update the weights of the double Q neural network;

[0203] Step 6: Perform the next round of training based on the new neural network weights until the maximum number of training times is reached;

[0204] Step 7: Find the scheduling solution with the largest objective function value in all iterations.

[0205] 5. When there is a deviation between the actual wind and solar power output and the forecast, a deviation penalty mechanism will be implemented.

[0206] The intraday deviation penalty mechanism refers to when the virtual power plant is unable to participate in electricity market transactions according to the day-ahead dispatch plan during actual operation, it will cause power abandonment losses or power shortage penalties.

[0207] The deviation power between the day-ahead market and the intraday market caused by the uncertainty of distributed generation power is shown in formula (39):

[0208]

[0209] Where ΔP t is the deviation power between the planned power generation in the day and the actual power generation in the day during period t; is the power generation capacity of the i-th renewable generator set in the t-th period of the day before.

[0210] For the processing of deviation power, first use energy storage, gas turbines and other units to adjust the deviation. Then, for the deviation power that still exists after the adjustment, it is traded in the balancing market based on the penalty electricity price. The calculation formula for the deviation power after adjustment is as follows:

[0211]

[0212] Among them, P t RM It is the deviation power generation after compensation of dispatchable units in the virtual power plant during period t. A positive value indicates that the actual transaction power is greater than the previous day's bid power, and a negative value indicates that the actual transaction power is less than the previous day's bid power, and it is necessary to purchase power from the real-time market to make up the difference. The power of the energy storage device to compensate for the bid deviation of the renewable energy unit during period t within the day; P is the power of the gas turbine to compensate for the bidding deviation of the renewable energy unit during the t period of the day; t DR,cs It is the power given to compensate the bidding deviation of renewable energy units in response to the adjustable load demand during period t within the day; μ is a ±1 variable. When μ=1, the ESS equipment is in a discharging state during period t; when μ=-1, the ESS equipment is in a charging state during period t.

[0213] The present invention covers any substitution, modification, equivalent method and scheme made on the essence and scope of the present invention. In order to make the public have a thorough understanding of the present invention, specific details are described in detail in the following preferred embodiments of the present invention, but those skilled in the art can fully understand the present invention without the description of these details. In addition, in order to avoid unnecessary confusion about the essence of the present invention, well-known methods, processes, procedures, components and circuits are not described in detail.

[0214] The above is only a preferred embodiment of the present invention. It should be pointed out that for ordinary technicians in this technical field, several improvements and modifications can be made without departing from the principle of the present invention. These improvements and modifications should also be regarded as the scope of protection of the present invention.

Claims

1. A low-carbon optimization operation method for a virtual power plant considering carbon trading, characterized in that: The following steps are involved: S1, Virtual Power Plant Framework Design and Modeling: Design the operation and trading framework of the low-carbon virtual power plant and model each unit in the virtual power plant; S2, characterization of wind and solar power output uncertainty: characterizes the uncertainty of wind and solar power output through the generation and reduction of scenarios; S3, benefit and carbon emission optimization objectives: construct an objective function with the goal of maximizing benefits and minimizing carbon emissions; S4, DQN optimizes the output plan: Use the DQN algorithm to optimize the next day's output plan; S5, implementation of the deviation penalty mechanism: when there is a deviation between the actual wind and solar power output and the forecast, the deviation penalty mechanism is implemented.

2. A low-carbon optimization operation method for a virtual power plant considering carbon trading according to claim 1, characterized in that: The virtual power plant framework design and modeling in S1 include: S11, Dispatchable Generator Set Modeling: The dispatchable generator set consists of gas turbines. Based on the minimum and maximum power, minimum and maximum output, and set time period, the dispatchable generator set modeling is combined with the startup and shutdown costs of the gas turbines, which is expressed as: in, They represent the minimum and maximum electrical energy generated by the gas turbine, They represent the minimum and maximum output of the gas turbine respectively, and τ is the set time period; Among them, n j is the number of linearization segments of the cost function, take n j =3, C su , C sd are the cost of starting and shutting down the gas turbine, K ij is the slope of the linearization segment; S12, Energy storage device modeling: Energy storage device modeling is based on the state transition equation. The electric energy generated is determined by the operating state of the energy storage device itself, which is expressed as: in, is the electrical energy in the energy storage device at time t, is the electric energy in the energy storage device at time t-1, P t SC (t) is the charging power of the energy storage device, is the discharge power of the energy storage device, η SC , η SD are the charging and discharging efficiency of the energy storage device respectively; S13, Wind power modeling: Wind power output is determined by wind speed, and the wind speed distribution function satisfies the Weibull distribution, which is expressed as: Where v is the actual wind speed, is the shape parameter of wind power, θ is the scale parameter of wind turbine; where v in <v t <v out ; Among them, P WP is the maximum output of the wind turbine, is the rated power of the wind turbine, v t is the actual wind speed of the wind turbine, v rat It is the rated wind speed during normal operation of the wind turbine; S14, Photovoltaic modeling: Photovoltaic output is determined by light intensity and climate. The light intensity conforms to the Beta distribution function, which is expressed as: Among them, θ is the light intensity of the photovoltaic unit, α is the shape parameter of the distribution function, and β is the distribution parameter of the distribution function; Among them, P PV is the maximum output of the photovoltaic unit, is the rated power of the photovoltaic system, θ t is the actual light intensity of the photovoltaic unit.

3. A low-carbon optimization operation method of a virtual power plant considering carbon trading according to claim 2, characterized in that: The uncertainty characterization of wind and solar power output in S2 includes: S21, scenario generation: Latin hypercube sampling is used to generate a power scenario set for fluctuating power sources, the error probability distribution is divided into multiple equal probability intervals, and a sample is randomly selected in each interval to finally generate a scenario set; S22, scene reduction: A fast backward reduction method based on probability distance calculates the distance between scenes, gradually merges similar scenes and updates the probability until the number of scenes meets the requirement.

4. A low-carbon optimization operation method for a virtual power plant considering carbon trading according to claim 3, characterized in that: The scene generation in S21 includes: S211, error probability distribution division: distribute the error probabilities of wind power and photovoltaic power generation in m equal probability spaces, and the probability of each interval is 1 / m; S212, randomly select probability points: in the i-th probability interval Randomly select a point p from i , expressed as: Where s is a randomly distributed number between [0,1]; S213, generate a scene set: randomly select probability points in each probability interval in turn, obtain m groups of sampling samples, and finally obtain a scene set M.

5. A low-carbon optimization operation method for a virtual power plant considering carbon trading according to claim 4, characterized in that: The scene reduction in S22 includes: S221, calculate geometric distance: calculate the geometric distance between two scenes in the scene set M, expressed as: Where, d is the Euclidean distance; S222, find the nearest scene pair: find the scene j that is closest to scene i from the scene set M, that is, min{d(m i ,m j ),i≠j}; S223, minimum probability distance scene selection: repeatedly calculate the geometric distance, find the closest scene pair, and obtain the scene i with the smallest sum of probability distances with the remaining scenes j; S224, update scene probability: update the scene probability, the new scene probability is equal to the probability of i plus the probability of j, and obtain a new scene set M; S225, repeat the scene reduction process: repeatedly calculate the geometric distance, find the nearest scene pair, select the scene with the minimum probability distance, and update the scene probability until the number of remaining scenes meets the requirement of the number of scene reductions.

6. A low-carbon optimization operation method for a virtual power plant considering carbon trading according to claim 5, characterized in that: The revenue and carbon emission optimization goals in S3 include: S31, benefit maximization objective function: Construct an objective function with benefit maximization as the goal, expressed as: Among them, F is the total profit obtained by the virtual power plant from participating in electricity market transactions, is the day-ahead market electricity sales revenue of the virtual power plant during period t, is the income or cost of the virtual power plant participating in carbon market transactions during period t, is the profit or cost caused by the deviation between the daily power of the virtual power plant and the planned power during period t, They are the maintenance cost of renewable energy units, energy storage system cost, demand response cost and gas turbine cost during period t respectively; in, is the day-ahead electricity market price during period t, P t DA is the power sold by the virtual power plant to the day-ahead power market during period t, P t ESS is the charging and discharging power of the energy storage system in the period t the day before, is the power sold by the gas turbine in the virtual power plant to the day-ahead power market during period t, P t DR is the power that the adjustable load in the virtual power plant responds to in the day-ahead power market during period t; Among them, γ A is the price of free carbon emission allowance in the carbon trading market, γ BE The price of the carbon emission reduction share generated by renewable energy units in the CCER trading market; in, is the deviation penalty electricity price during period t, ε + , ε - is the deviation penalty coefficient; in, are the operating cost coefficients of wind turbines and photovoltaic units, are the actual power generation of the i-th wind turbine and photovoltaic unit in period t, k is the number of wind turbines, and q is the number of photovoltaic units; Among them, δ ESS The unit charging and discharging cost of the energy storage system; in, is the operation management coefficient of the gas turbine, is the unit power generation fuel cost of the gas turbine; in, is the demand response compensation coefficient, P t DR is the demand response power in the t period the day before; S32, Carbon emission minimization objective function: Construct an objective function with carbon emission minimization as the goal, expressed as: Among them, CAR is the total carbon emissions generated during the operation of the virtual power plant. t is the carbon emissions of the virtual power plant during period t.

7. A low-carbon optimization operation method of a virtual power plant considering carbon trading according to claim 6, characterized in that: The constraints of the profit maximization objective function and the carbon emission minimization objective function include: Gas turbine constraints: Among them, v k is the operating state of gas turbine k, 1 means start and 0 means shut down. is the output of gas turbine k participating in the day-ahead market during time period t, Δ MT is the ramp constraint of the gas turbine, is the output of gas turbine k participating in the balancing market in the scenario ω in time period t, P MT,max , P MT,min They represent the minimum output of the gas turbine, is the limit of the regulating power of gas turbine k; Energy storage device constraints: in, is the stored energy of battery q in time period t, represents the charging power of battery q in time period t, represents the discharge power of battery q in time period t, η SC Indicates the charging loss of the battery, P SCmax Indicates the maximum charge limit of the battery, P SDmax Indicates the maximum discharge limit of the battery, Q min , Q max Indicates the minimum and maximum storage capacity of the battery; Power balance constraints: in, They represent the output of wind power m and photovoltaic power n in the day-ahead market period t, They represent the power of battery q participating in the day-ahead market charging and discharging during period t, is the output of micro gas turbine k in the day-ahead market during period t; Power balancing constraints of VPP in the balancing market: in, They represent the output of wind power m and photovoltaic power n in the balancing market in the scenario ω of time period t, They represent the positive and negative deviations between the actual output of the virtual power plant and the declared power, respectively. They represent the power of battery q participating in the charging and discharging of the balancing market in the scenario ω during time period t, The output of micro gas turbine k participating in the balancing market in scenario ω during time period t.

8. A low-carbon optimization operation method for a virtual power plant considering carbon trading according to claim 7, characterized in that: The DQN optimization output scheme in S4 includes: S41, DQN algorithm parameter initialization: Initialize the learning rate, number of training times, and neural network weights in the DQN algorithm; S42, obtaining the initial state of the virtual power plant: starting the first virtual power plant scheduling, obtaining the environmental state of the virtual power plant in operation, that is, the initial state of each device in the virtual power plant; S43, the agent takes scheduling actions: the agent takes actions under the constraints, including discharging energy storage and increasing the operating power of the gas unit, calculates the reward under the current action, that is, the target value of the virtual power plant scheduling model, and updates the current environmental state, and puts the current work, reward, and environmental state experience values ​​into the experience pool; S44, experience pool sample accumulation: the agent repeatedly takes scheduling actions until the sample capacity in the experience pool reaches the set threshold; S45, experience pool sampling and neural network weight update: randomly extracting a predetermined number of samples from the experience pool to update the weights of the double Q neural network; S46, performing the next round of training based on the new weights: performing the next round of training based on the updated neural network weights until the maximum number of training times is reached; S47, find the best scheduling plan: find the scheduling plan with the largest objective function value in all iterations as the optimal output plan of the virtual power plant on the next day.

9. A low-carbon optimization operation method for a virtual power plant considering carbon trading according to claim 8, characterized in that: The implementation of the deviation penalty mechanism in S5 includes: S51, calculate deviation power: calculate the deviation power caused by the uncertainty of distributed power generation, and determine the difference between the actual power generation of the virtual power plant and the planned power generation on the day before, expressed as: Where ΔP t is the deviation power between the planned power generation in the day and the actual power generation in the day during period t, is the power generation capacity of the i-th renewable generator set in the t-th period of the day before; S52, deviation adjustment: using energy storage equipment and gas turbines to compensate for the deviation of electricity, and reducing the deviation by discharging energy storage or adjusting the power of gas turbines; S53, calculating the adjusted deviation power: calculating the remaining deviation power after the deviation is adjusted.

10. A low-carbon optimization operation method for a virtual power plant considering carbon trading according to claim 9, characterized in that: The calculated adjusted deviation power in S53 is expressed as: Among them, P t RM It is the deviation power generation after compensation of the dispatchable units in the virtual power plant during period t. Its positive value indicates that the actual transaction power is greater than the previous day's bid power. Its negative value indicates that the actual transaction power is less than the previous day's bid power. It is necessary to purchase power from the real-time market to make up the difference. The power of the energy storage device to compensate for the bidding deviation of the renewable energy unit during the t period of the day, P is the power of the gas turbine to compensate for the bidding deviation of the renewable energy unit during the t period of the day, t DR,cs It is the power given to compensate the bidding deviation of renewable energy units in response to the adjustable load demand during period t within the day. μ is a ±1 variable. When μ=1, the ESS equipment is in a discharging state during period t. When μ=-1, the ESS equipment is in a charging state during period t.

Images