Medium-and-long-term transaction declaration decision-making method and system for photovoltaic power station

Abstract

The invention discloses a photovoltaic power station medium and long term transaction declaration decision-making method and system, and belongs to the technical field of power transaction. According to the photovoltaic power station medium and long term transaction declaration decision-making method, weather condition data in a transaction period is predicted by constructing a weather prediction simulation model, an output feature extraction model, an index generation model, a declaration scene simulation model and an optimal output decision-making model, and typical features in photovoltaic output and market electricity price data are extracted; obtaining an output electricity price index combined sample; clustering processing is carried out, sampling is carried out, and photovoltaic output and electricity price scenes are obtained; and finally, obtaining the declaration decision information of the photovoltaic power station by taking the maximum income as the target, thereby fully considering the new energy power generation characteristics, forming a reasonably designed power transaction mode, and ensuring that the new energy fairly participates in the market transaction, so that the photovoltaic power station can avoid the risk and has sustainable income.

Description

Technical Field

[0001] The present invention relates to a medium - and long - term trading declaration decision - making method and system for a photovoltaic power station, belonging to the technical field of power trading. Background Art

[0002] At the present stage, medium - and long - term power trading still mainly focuses on electricity quantity. The dispatching agency needs to formulate corresponding power generation plans according to the contract electricity quantities of each power plant to ensure that each power plant generates the electricity quantities stipulated in the contract.

[0003] The power grid mainly decomposes based on the principles of "power balance and typical - day power verification", using methods such as average distribution, according to load ratio or power station capacity ratio, etc. However, considering the inherent output volatility of new - energy power stations, if the dispatching agency further subdivides the contract electricity quantities, it will bring deviations. It has become a more common exploration method for new - energy power stations to participate in the power market declaration in a self - planned mode.

[0004] Therefore, how to make full use of new - energy power generation, design a reasonable power trading mode to ensure that new energy participates in market trading fairly has become an important issue that needs to be considered by the formulators, implementers, and managers of new - energy trading rules.

[0005] The information disclosed in this background art is only used to understand the background of the inventive concept of the present invention, so it may include information that does not constitute prior art. Summary of the Invention

[0006] In view of the above problems or one of the above problems, the first object of the present invention is to provide a medium - and long - term trading declaration decision - making method and system for a photovoltaic power station. By constructing a weather prediction simulation model, an output characteristic extraction model, an index generation model, a declaration scenario simulation model, and an optimal output decision model, the weather condition data within the trading cycle is predicted, and the typical characteristics in the photovoltaic output and market electricity price data are extracted to obtain a joint sample of output - electricity - price indexes; then the joint sample of output - electricity - price indexes is clustered and sampled to obtain photovoltaic output and electricity - price scenarios; finally, with the goal of maximizing revenue, the declaration decision - making information of the photovoltaic power station is obtained, and the medium - and long - term trading declaration decision of the photovoltaic power station is completed, so as to fully consider the characteristics of new - energy power generation, form a reasonable power trading mode, ensure that new energy participates in market trading fairly, and the solution is scientific, reasonable, and practical.

[0007] In view of the above problems or one of the above problems, the second object of the present invention is to provide a medium - and long - term trading declaration decision - making method and system for a photovoltaic power station. An optimal output decision model considering conditional value - at - risk and transaction probability is established for the output characteristics of the photovoltaic power station, so as to avoid risks for the photovoltaic power station and ensure that the power station has sustainable revenue.

[0008] To achieve the first object, the first technical solution of the present invention is as follows:

[0009] A medium- and long-term trading declaration decision-making method for a photovoltaic power station, comprising the following steps:

[0010] Step 1: Through a pre-constructed weather prediction simulation model, obtain a certain trading period in the medium- and long-term trading of the photovoltaic power station to be decided, and predict the weather condition data within this trading period;

[0011] Step 2: Using a pre-constructed output feature extraction model, based on the weather condition data, extract typical features from the photovoltaic output and market electricity price data to obtain a joint sample of output electricity price indicators;

[0012] Step 3: Adopt a pre-constructed index generation model to perform clustering processing on the joint sample of output electricity price indicators to obtain joint output electricity price indicators;

[0013] Step 4: Use a pre-constructed declaration scenario simulation model to sample the joint sample of output electricity price indicators based on the joint output electricity price indicators to obtain photovoltaic output and electricity price scenarios;

[0014] Step 5: Based on a pre-constructed optimal output decision-making model, with the goal of maximizing the profit, process the photovoltaic output and electricity price scenarios to obtain the declaration decision-making information of the photovoltaic power station, and complete the declaration decision-making of the medium- and long-term trading of the photovoltaic power station.

[0015] Through continuous exploration and experiments, the present invention constructs a weather prediction simulation model, an output feature extraction model, an index generation model, a declaration scenario simulation model, and an optimal output decision-making model, predicts the weather condition data within the trading period, and extracts typical features from the photovoltaic output and market electricity price data to obtain a joint sample of output electricity price indicators; then performs clustering processing on the joint sample of output electricity price indicators and conducts sampling to obtain photovoltaic output and electricity price scenarios; finally, with the goal of maximizing the profit, obtains the declaration decision-making information of the photovoltaic power station and completes the declaration decision-making of the medium- and long-term trading of the photovoltaic power station, so as to fully consider the characteristics of new energy power generation, form a reasonably designed power trading mode, ensure the fair participation of new energy in market trading, enable the photovoltaic power station to avoid risks, ensure sustainable profits of the photovoltaic power station, and the solution is scientific, reasonable and practical.

[0016] As a preferred technical measure:

[0017] Step 1: The method for obtaining a certain trading period in the medium- and long-term trading of the photovoltaic power station to be decided and predicting the weather condition data within this trading period through a pre-constructed weather prediction simulation model is as follows:

[0018] Obtain a certain trading period in the medium- and long-term trading of the photovoltaic power station to be decided;

[0019] Obtain historical weather information at corresponding times according to a certain trading cycle, and obtain the weather conditions for several dates based on the historical weather information;

[0020] The weather conditions are divided into sunny, cloudy or rainy, numbered 1, 2, and 3 respectively;

[0021] Statistically analyze the conversions between the weather conditions on different dates to obtain the transition frequencies between each state;

[0022] Utilize the previously constructed multi - order weighted Markov chain model, and based on the transition frequencies between each state, calculate the s - order state transition probability matrix Its calculation formula is as follows:

[0023]

[0024] In the formula, x i and x i+s represent the weather conditions on the i - th and (i + s) - th days respectively, and represent the frequency and probability of transitioning from state a to state b respectively, where a, b ∈ {1, 2, 3};

[0025] After obtaining the s - order state probability matrices with different step lengths, confirm the corresponding weights according to their contribution degrees to the prediction result accuracy to obtain the weight matrix;

[0026] Couple the weight matrix with the s - order state probability matrix to calculate the weighted transition probability matrix P W ;

[0027] To obtain the future medium - and - long - term weather distribution, based on the transition probability matrix P W , determine the stationary distribution of the Markov chain;

[0028] According to the ergodic theorem, utilize the stationary distribution of the Markov chain to calculate the recurrence period of each state, and obtain the average recurrence period and stable probability of each weather condition within this trading cycle; the specific calculation formula is as follows:

[0029]

[0030] In the formula, π a is the stable probability of the weather condition numbered a, where a = 1, 2, 3;

[0031] Based on the average recurrence period and stable probability of each weather condition within this trading cycle, predict the weather condition data within this trading cycle.

[0032] As an optimized technical measure:

[0033] The method for constructing the multi - order weighted Markov chain model is as follows:

[0034] Set a stochastic process {ξ(t), t ∈ T}. If for any integer {ξ(t), t ∈ T}, the state space is I, the parameter is a non - negative integer, and ξ(n) satisfies the probability distribution of a Markov chain. The expression of the probability distribution of the Markov chain is as follows:

[0035]

[0036] Among them, P{ξ(n + 1) = j|ξ(n) = i} is the conditional probability, which includes several one - step transition probabilities p ij ;

[0037] Arrange the one - step transition probabilities p ij in sequence to obtain the one - step transition probability matrix, and its expression is as follows:

[0038]

[0039] Among them, p ij represents the probability at the next moment n + 1 under the condition that ξ(n) = i at moment n;

[0040] Based on the transition probability matrix, determine the probability distribution {π, i ∈ I};

[0041] Take the probability distribution {π, i ∈ I} as the stationary distribution of the Markov chain, where I is the state space, and it satisfies the following relationship:

[0042]

[0043] ∑ i∈I π i = 1 (i ∈ I).

[0044] As a preferred technical measure:

[0045] After obtaining the s - order state probability matrices with different step sizes, confirm the corresponding weights according to their contribution degrees to the prediction result accuracy. The method for obtaining the weight matrix is as follows:

[0046] First, calculate the correlation between weather states to obtain the autocorrelation coefficient, which is used to judge the strength of the relationship between historical and current weather conditions. Its calculation formula is as follows:

[0047]

[0048] After normalizing the autocorrelation coefficient, use it as the weight value of each - order transition probability matrix. Its calculation formula is as follows:

[0049]

[0050] Sum up the weight values to obtain the weight matrix.

[0051] As a preferred technical measure:

[0052] In step two, by using a pre-constructed output feature extraction model, based on weather condition data, typical features in photovoltaic output and market electricity price data are extracted, and the method for obtaining a joint sample of output electricity price indicators is as follows:

[0053] Obtain weather condition data, as well as corresponding historical photovoltaic output and market electricity price data;

[0054] Regarding historical photovoltaic output, output features are measured from three dimensions: single point, overall, and fluctuation, which include single-point evaluation indicators, overall evaluation indicators, and fluctuation evaluation indicators;

[0055] The single-point evaluation indicator of historical photovoltaic output is the full-day maximum photovoltaic output P max and the time t max corresponding to the maximum output, and its expression is as follows:

[0056] P max = max{P1, P2, …, P n}

[0057] In the formula, P i (i = 1, 2, …, n) represents the output value of the photovoltaic power station at the i-th moment;

[0058] The overall evaluation indicator of historical photovoltaic output is the expected value of photovoltaic output P av and the half-load output probability p 0.5～1 ; The half-load output probability represents the ratio of the number of times the output of the photovoltaic power station reaches half of the rated output throughout the day to the total frequency; its expression is as follows:

[0059]

[0060] p 0.5～1 = n 0.5～1 / n

[0061] Since the photovoltaic output is affected by meteorological conditions and has randomness, corresponding volatility indicators are needed to measure the random fluctuation of the output. Therefore, the maximum fluctuation ratio ΔP max is selected as the representative volatility indicator; its expression is as follows:

[0062]

[0063] Regarding market electricity price data, electricity price features are measured from three dimensions: single point, overall, and price difference, which include single-point evaluation indicators, overall evaluation indicators, and price difference evaluation indicators;

[0064] Due to the bimodal characteristic of the electricity price, it reaches one of the peaks in the morning, and the electricity price tends to decrease during the peak photovoltaic generation period at noon. Therefore, the single-point evaluation index of the market electricity price data is the maximum daily electricity price p max and the minimum electricity price p min,noon at noon; its expression is as follows:

[0065] p max = max{p1, p2, …, p n}

[0066] p min,noon = min{p 1,noon , p 2,noon , …, p n,noon}

[0067] In the formula, p i (i = 1, 2, …, n) represents the market electricity price value at the i-th moment;

[0068] The overall evaluation index of the market electricity price data for the whole day is selected as the expected value p av of the daily electricity price; its expression is as follows:

[0069]

[0070] Since the daily electricity price has obvious peak-valley characteristics, the peak-valley electricity price difference Δp noon is used as the price difference characteristic of the daily electricity price; its expression is as follows:

[0071] Δp noon = p max - p min,noon

[0072] The single-point evaluation index of historical photovoltaic output, the overall evaluation index of historical photovoltaic output, the fluctuation evaluation index, the single-point evaluation index of market electricity price data, the overall evaluation index of market electricity price data, and the price difference evaluation index are respectively standardized to form a joint sample of output electricity price indexes in historical photovoltaic output and market electricity price data;

[0073] The calculation formula for standardization is as follows:

[0074]

[0075] In the formula, X represents each evaluation index.

[0076] As an optimal technical measure:

[0077] Step 3, using a pre-constructed index generation model, cluster the joint sample of output electricity price indexes to obtain the method of output electricity price joint indexes as follows:

[0078] Obtain the joint sample of output electricity price indexes;

[0079] Based on the FCM clustering algorithm, the clustering centers of the joint samples of the output electricity price indicators are calculated;

[0080] According to the clustering centers, the joint samples of the output electricity price indicators are clustered by FCM to obtain a fuzzy membership matrix, which is used to describe the degree to which each group of joint samples of the output electricity price belongs to a certain clustering center;

[0081] FCM clustering is an iterative optimization algorithm with the goal of minimizing the sum of the distances between each joint sample of the output electricity price index and each clustering center weighted by the corresponding fuzzy membership degree. The fuzzy membership degree is determined according to the distance between the sample point and the clustering center. The larger the membership degree, the closer the sample point is to the clustering center;

[0082] According to the fuzzy membership matrix, the joint index of the output electricity price is obtained;

[0083] Step 4, using the pre-constructed declaration scenario simulation model, based on the joint index of the output electricity price, sampling the joint samples of the output electricity price indicators, and the method for obtaining the photovoltaic output and electricity price scenarios is as follows:

[0084] Obtain the joint index of the output electricity price and the joint samples of the output electricity price indicators;

[0085] Calculate the proportion of the joint index of the output electricity price in the joint samples of the output electricity price indicators to obtain the sample proportion;

[0086] According to the sample proportion, perform Latin hypercube sampling on the joint samples of the output electricity price indicators to simulate the photovoltaic output and electricity price data during the future trading period;

[0087] Generate photovoltaic output and electricity price scenarios based on the photovoltaic output and electricity price data.

[0088] As an optimal technical measure:

[0089] Step 5, based on the pre-constructed optimal output decision model, with the goal of maximizing the profit, process the photovoltaic output and electricity price scenarios, and the method for obtaining the declaration decision information of the photovoltaic power station is as follows:

[0090] Obtain the photovoltaic output and electricity price scenarios;

[0091] Based on the photovoltaic output and electricity price scenarios, determine the declared output value, electricity price at time t of the power station, and the actual output value of the power station at time t;

[0092] According to the declared output value, electricity price at time t of the power station, and the actual output value of the power station at time t, calculate the profit R i The calculation formula is as follows:

[0093]

[0094] Among them, R i is the income under the corresponding scenario, which is the difference between the declared output income and the deviation settlement; t = 1, 2, ... T, T is the total number of time periods; p t , t is the declared output value and electricity price of the power station at time t; P t,i is the actual output value of the power station at time t; For excess power generation; The amount of power generation is insufficient;

[0095] When the actual output of a power station is greater than the declared output, the difference between the two is the excess power generation. The photovoltaic power station obtains electricity sales fees, and the unit electricity sales fee is λ + When the power station's declared output is greater than its actual output, the difference between the two is the under-generated power. The photovoltaic power station pays the electricity purchase fee, and the unit electricity purchase fee is λ - ;

[0096] According to the income under the corresponding scenario, the probability of each output scenario, the number of output scenarios and the decision quotation λ t The corresponding transaction probability is used to calculate the expected revenue of the power plant under different scenarios;

[0097]

[0098] Among them, π(i) is the probability of each output scenario, N is the number of output scenarios, R i is the profit in the corresponding scenario, p R The decision quote is λ t The corresponding transaction probability when

[0099] Based on the expected returns of power plants under different scenarios, and using conditional value at risk (CVaR) as a risk metric, and with the goal of maximizing returns and minimizing risks, an objective function is constructed to quantify the risks brought about by uncertain factors when power plants participate in transactions, thereby providing the optimal decision-making plan for output curves and quotations.

[0100] The expression of the objective function is as follows:

[0101] max{E(R)-μ·CVaR β}

[0102] Where E(R) is the expected revenue of the power station under different scenarios, CVaR β is the optimal CVaR value under the confidence level β; μ is the risk preference factor, which indicates the decision maker’s risk aversion. The larger μ is, the more conservative the decision maker is, and the smaller μ is, the more inclined to high risk the decision maker is.

[0103] Based on the objective function and constraints, the initial declared output value of the first stage is obtained

[0104] According to the initial declared output value of the first phase Calculate the probability distribution function of returns and the initial bid value of the second stage

[0105] The initial quoted value Substitute λ t , and update the declared output value to obtain the new declared output value

[0106] Iterative solution of P t , t Until the convergence conditions are met, or the given number of iterations or calculation time is reached, the photovoltaic power station declaration decision information is finally obtained.

[0107] As preferred technical measures:

[0108] Set constraints, including upper and lower limits on power plant output, ramp rate, and upper and lower limits on bid. The specific expressions are as follows:

[0109] 0≤P t ≤P max,t

[0110]

[0111] Where P max,t is the maximum output limit of the power station, are the limits of the maximum power increase and decrease allowed by the power station per unit time, λ min,t , max,t The minimum and maximum values of the declared prices;

[0112] Or / and, the optimal CVaR value CVaR under confidence level β β The calculation process is as follows:

[0113] CVaR β The meaning of the optimal CVaR value is to minimize the CVaR loss value under the confidence level β, set the loss function f(x,y) to -E(R), and set the confidence level to β, that is:

[0114]

[0115] Introducing the dummy variable z j , let z i =max[-R i -α,0] represents the loss exceeding VaR, and CVaR based on CVaR and transaction probability is obtained βThe calculation formula is as follows:

[0116]

[0117] Or / and, the method for obtaining the revenue probability distribution function is as follows:

[0118] Select the Gaussian kernel function as the kernel density function. The expression of the Gaussian kernel density function is:

[0119]

[0120] In the formula, is the distribution function value of the revenue R;

[0121] According to the Gaussian kernel density function, use the non-parametric kernel density method that can effectively fit any distribution to fit the revenue probability density function. The expression is as follows:

[0122]

[0123] In the formula, R j is the market revenue, N j is the number of samples, is the kernel density estimate of the revenue, that is, the revenue probability distribution function, K() is the kernel density function, and h is the bandwidth.

[0124] To achieve one of the above purposes, the second technical solution of the present invention is:

[0125] A medium- and long-term trading declaration decision-making method for a photovoltaic power station, including the following steps:

[0126] Step 1, according to the weather state distribution predicted by the weighted Markov chain, obtain the joint sample of the output power price index, and perform Latin hypercube sampling to generate the photovoltaic output power and price scenarios, and then use the optimal output power decision-making model based on CVaR to obtain the initial declared output power value.

[0127] Step 2, calculate the revenue probability distribution function according to the initial declared output power value in the first stage to obtain the initial declared value.

[0128] Step 3, use the initial declared value as the power price of the power station at the next moment, and update the declared output power value according to the optimal output power decision-making model to obtain the new declared output power value.

[0129] Step 4, iteratively solve the new declared output power value and the power price of the power station at the next moment until the convergence condition is met, or the given number of iterations or calculation time is reached. Otherwise, recalculate until the decision-making output power of the photovoltaic power station is finally obtained.

[0130] In view of the output characteristics of a photovoltaic power station, the present invention establishes an optimal output decision-making model considering conditional value at risk and transaction probability, so as to avoid risks for the photovoltaic power station and ensure sustainable profits for the power station.

[0131] To achieve one of the above objects, the third technical solution of the present invention is as follows:

[0132] A medium- and long-term trading declaration decision-making system for a photovoltaic power station, comprising:

[0133] One or more processors;

[0134] A storage device for storing one or more programs;

[0135] When the one or more programs are executed by the one or more processors, the one or more processors implement the above-mentioned medium- and long-term trading declaration decision-making method for a photovoltaic power station.

[0136] Compared with the prior art solutions, the present invention has the following beneficial effects:

[0137] Through continuous exploration and experiments, the present invention predicts the weather condition data within the trading period by constructing a weather prediction simulation model, an output characteristic extraction model, an index generation model, a declaration scenario simulation model, and an optimal output decision-making model, and extracts the typical characteristics in the photovoltaic output and market electricity price data to obtain a combined sample of output electricity price indexes; then performs clustering processing on the combined sample of output electricity price indexes and conducts sampling to obtain photovoltaic output and electricity price scenarios; finally, with the goal of maximizing profits, obtains the declaration decision-making information of the photovoltaic power station and completes the medium- and long-term trading declaration decision-making of the photovoltaic power station, so that the characteristics of new energy power generation can be fully considered, a reasonably designed power trading mode can be formed, and it can ensure that new energy participates in market trading fairly, and the solution is scientific, reasonable, and practical.

[0138] Furthermore, in view of the output characteristics of a photovoltaic power station, the present invention establishes an optimal output decision-making model considering conditional value at risk and transaction probability, so as to avoid risks for the photovoltaic power station and ensure sustainable profits for the power station. BRIEF DESCRIPTION OF THE DRAWINGS

[0139] Figure 1 is a flowchart of a medium- and long-term trading declaration decision-making method for a photovoltaic power station according to the present invention;

[0140] Figure 2 is a schematic diagram of a weather prediction Markov chain according to the present invention;

[0141] Figure 3 is a schematic diagram of an FCM clustering algorithm according to the present invention;

[0142] Figure 4 is a schematic diagram of a medium- and long-term curve decision-making model according to the present invention. Detailed implementation manners

[0143] In order to make the objectives, technical solutions and advantages of the present invention clearer and more understandable, the present invention will be further described in detail below with reference to 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 used to limit the present invention.

[0144] On the contrary, the present invention covers any alternatives, modifications, equivalent methods and solutions made within the spirit and scope of the present invention defined by the claims. Further, in order to enable the public to better understand the present invention, some specific details are described in detail in the following detailed description of the present invention. Those skilled in the art can fully understand the present invention without the description of these details.

[0145] Unless otherwise defined, all technical and scientific terms used in this application have the same meaning as commonly understood by those skilled in the technical field to which the present invention belongs. The terms used in this application are only for the purpose of describing specific embodiments and are not intended to limit the present invention. The term "or / and" used in this application includes any and all combinations of one or more of the related listed items.

[0146] As Figure 1 shown, the first specific embodiment of the medium- and long-term trading declaration decision-making method for a photovoltaic power station of the present invention:

[0147] A medium- and long-term trading declaration decision-making method for a photovoltaic power station, comprising the following steps:

[0148] Step 1, obtain a certain trading period in the medium- and long-term trading of the photovoltaic power station to be decided through a pre-constructed weather prediction simulation model, and predict the weather condition data within the certain trading period;

[0149] Step 2, use a pre-constructed output feature extraction model to extract typical features from the photovoltaic output and market electricity price data based on the weather condition data to obtain a combined sample of output electricity price indicators;

[0150] Step 3, adopt a pre-constructed index generation model to perform clustering processing on the combined sample of output electricity price indicators to obtain combined output electricity price indicators;

[0151] Step 4, use a pre-constructed declaration scenario simulation model to sample the combined sample of output electricity price indicators based on the combined output electricity price indicators to obtain photovoltaic output and electricity price scenarios;

[0152] Step 5, based on a pre-constructed optimal output decision model, with the goal of maximizing the profit, process the photovoltaic output and electricity price scenarios to obtain the declaration decision information of the photovoltaic power station, and complete the declaration decision of the medium- and long-term trading of the photovoltaic power station.

[0153] The second specific embodiment of the medium- and long-term trading declaration decision-making method for the photovoltaic power station of the present invention:

[0154] A medium- and long-term trading declaration decision-making method for a photovoltaic power station, comprising the following steps:

[0155] Step 1: According to the weather state distribution predicted by the weighted Markov chain, obtain the joint sample of the output power price index, and perform Latin hypercube sampling to generate the photovoltaic output power and price scenarios, and then use the optimal output power decision-making model based on CVaR to obtain the initial declared output power value.

[0156] Step 2: Calculate the revenue probability distribution function according to the initial declared output power value in the first stage to obtain the initial declared value.

[0157] Step 3: Take the initial declared value as the power price of the power station at the next moment, and update the declared output power value according to the optimal output power decision-making model to obtain the new declared output power value.

[0158] Step 4: Iteratively solve the new declared output power value and the power price of the power station at the next moment until the convergence condition is met, or the given number of iterations or calculation time is reached. Otherwise, recalculate until finally obtaining the decision-making output power of the photovoltaic power station.

[0159] The third specific embodiment of the medium- and long-term trading declaration decision-making method for the photovoltaic power station of the present invention:

[0160] A medium- and long-term trading declaration decision-making method for a photovoltaic power station, comprising the following:

[0161] Firstly, use the multi-order weighted Markov chain model to predict the weather condition distribution during the trading period. After processing the output power data and market power price data of the photovoltaic power station, obtain the joint index of output power and price. After performing FCM clustering according to the index and then performing Latin hypercube sampling, obtain the output power scenario and price scenario of the photovoltaic power station, and establish a photovoltaic power station output power curve and quotation decision-making model based on conditional value at risk (CVaR) and transaction probability. With the goal of maximizing revenue under the condition of avoiding the revenue risk of extreme situations, complete the declaration decision-making for the medium- and long-term trading of the photovoltaic power station.

[0162] The fourth specific embodiment of the medium- and long-term trading declaration decision-making method for the photovoltaic power station of the present invention:

[0163] A medium- and long-term trading declaration decision-making method for a photovoltaic power station, comprising the following:

[0164] The first step is to construct a multi-order weighted Markov chain model, which includes the following:

[0165] A Markov chain is a concept closely related to a Markov process. A process that satisfies a Markov chain has the following three characteristics: a. Discreteness of the process. The development of things can be discretized into a finite or countable number of states in terms of time. b. Randomness of the process. The transition from one state to another within the system is random, and the probability of the transition is represented by the probability values of the previous historical conditions within the system. c. Lack of aftereffect of the process. The transition probability within the system depends only on the current state and not on the previous states.

[0166] Given a stochastic process {ξ(t), t ∈ T}, if for any integer {ξ(t), t ∈ T} (the state space is I) and the parameter is a non - negative integer, and ξ(n) satisfies:

[0167]

[0168] Then this type of process is called a Markov chain. A Markov chain indicates that the state of a thing system changes from the past to the present and then from the present to the future, like a chain link by link. For a dynamic system that is a Markov chain, what state and value it will take in the future only depend on the current state and value, and not on its previous state and value. To describe the (n + 1)-dimensional probability distribution of a Markov chain, the most important is the conditional probability P{ξ(n + 1) = j|ξ(n) = i}, and this conditional probability is called the one - step transition probability p at time n. ij , which represents the probability p at the next moment n + 1 under the condition that ξ(n) = i at time n. ij . By sorting p ij in sequence, the one - step transition probability matrix can be obtained:

[0169]

[0170] The probability distribution {π, i ∈ i} is called the stationary distribution of the Markov chain, where I is the state space and it satisfies the following relationship:

[0171]

[0172] Second, conduct weather forecasting within the trading cycle, which includes the following content:

[0173] The output of a photovoltaic power station has obvious meteorological characteristics, mainly manifested in that the output of the photovoltaic power station changes with the meteorological conditions. Therefore, when formulating the output curve of the photovoltaic power station, it is necessary to judge the weather distribution within the trading cycle.

[0174] Predict the distribution of weather conditions during the trading cycle using Markov chains. A Markov chain is a discrete-time stochastic process model with the Markov property, characterized by discreteness, randomness, and lack of aftereffect, that is, the past (i.e., the historical states before the current one) is irrelevant for predicting the future (i.e., the future states after the current state). Since the change of weather conditions is related to the meteorological conditions in the past several days, or a certain climate phenomenon lasts for a long time, it is necessary to consider the influence of the continuous multi-step historical situation on the prediction accuracy, and adopt a multi-order weighted Markov chain model to predict the weather in order to make full use of historical weather information.

[0175] The transition between weather states on different days can be represented by a state transition probability matrix Weather types are divided into sunny, cloudy, and rainy, numbered 1, 2, and 3 respectively. After obtaining the weather types of each day according to historical data, count the transition frequencies between each state and then calculate the s-order state transition probability matrix The calculation formula is as follows:

[0176]

[0177] In the formula, is the s-order transition probability from state a to state b, x i and x i+s represent the weather states on the i-th and i + s-th days respectively, and represent the frequency and probability of the transition from state a to state b respectively, a, b ∈ {1, 2, 3}. The schematic diagram of the Markov chain for weather prediction is as Figure 2 shown.

[0178] After obtaining the state probability matrices with different step lengths, it is necessary to confirm the corresponding weights according to their contribution degrees to the prediction result accuracy, and calculate the weighted transition probability matrix P W . If the change trend between the historical weather of a certain step length and the current weather is stronger, then the influence of the historical weather of this step length on the current weather is greater, which means that the prediction result accuracy of the Markov chain model of this order is higher, and the assigned weight coefficient should be larger. Therefore, use the autocorrelation coefficient r s of each order, that is, the correlation between the weather states from the 1st to the T - s-th day and from the 1 + s-th to the T-th day, to judge the correlation degree between the historical situation of each step length and the change trend of the current weather state and determine the weight, which includes the following content:

[0179] First, calculate the autocorrelation coefficient to judge the strength of the relationship between the historical and current weather conditions, and after normalizing it, use it as the weight value of each order transition probability matrix. Finally, sum the weighted values to obtain P W . The relevant calculation formulas are as follows:

[0180]

[0181] wherein, u s is the normalized autocorrelation coefficient, and is the average value of weather conditions.

[0182] To obtain the weather distribution in a relatively long future period, the recurrence period of each state can be calculated by using the stationary distribution characteristics of the Markov chain. For a finite-state aperiodic irreducible Markov chain, the limiting distribution, i.e., the stationary distribution, of this chain can be obtained according to the ergodicity theorem. Therefore, this application utilizes this characteristic to calculate the average recurrence period and stable probability of each weather state within the trading period. The formula is as follows:

[0183]

[0184] In the formula, π a is the stable probability of the weather state numbered a.

[0185] Thirdly, based on stochastic optimization, generate the joint scenarios of power output and electricity price, which include the following contents:

[0186] To generate the power output curve and electricity price scenarios of the photovoltaic power station, by extracting the typical characteristics from the historical photovoltaic power output and market electricity price data under different weather conditions, and performing FCM clustering analysis on the joint index of power output and electricity price, and then conducting Latin hypercube sampling according to the distribution of the clustering results.

[0187] For the photovoltaic power output data, measure the power output characteristics from three dimensions: single point, overall, and fluctuation.

[0188] The single-point evaluation index of the photovoltaic power output data is selected as the full-day maximum power output P max and the time t max corresponding to the maximum power output. Its expression is as follows:

[0189] P max = max{P1, P2, …, P n}

[0190] In the formula, P i (i = 1, 2, …, n) represents the power output value of the photovoltaic power station at the i-th moment.

[0191] The full-day overall evaluation index is selected as the expected value of the photovoltaic power output P av and the half-load power output probability p 0.5～1 . The half-load power output probability represents the ratio of the number of times the power output of the photovoltaic power station reaches half of the rated power output throughout the day to the total frequency. Its expression is as follows:

[0192]

[0193] p 0.5～1 = n 0.5～1 / n

[0194] Since the output of photovoltaic power is affected by meteorological conditions such as solar irradiance and temperature, and has obvious randomness, corresponding volatility indicators are needed to measure the random fluctuation of the output. Therefore, the maximum fluctuation ratio ΔP max is selected as the representative index of fluctuation, and its expression is as follows:

[0195]

[0196] For electricity price data, the electricity price characteristics are measured from three dimensions: single point, overall and price difference.

[0197] Since the electricity price shows a bimodal characteristic, reaching one peak in the morning and the electricity price tending to decrease during the peak photovoltaic power generation period at noon, the single-point evaluation index of photovoltaic output data is selected as the maximum daytime electricity price p max and the minimum electricity price p min,noon .

[0198] p max =max{p1,p2,…,p n}

[0199] p min,noon =min{p 1,noon ,p 2,noon ,…,p n,noon}

[0200] In the formula, p i (i = 1,2,…,n) represents the market electricity price value at the i-th moment.

[0201] The overall evaluation index for the whole day is selected as the expected value p av of the daytime electricity price, and its expression is as follows:

[0202]

[0203] Since the daytime electricity price has obvious peak-valley characteristics, the peak-valley electricity price difference Δp noon is used as the price difference characteristic of the daytime electricity price, and its expression is as follows:

[0204] Δp noon =p max -p min,noon

[0205] After calculating all the indicators based on the output and electricity price data, according to Equation Perform standardization, where X represents each output price index. Then, the output price index joint samples are clustered using FCM. FCM clustering is an iterative optimization algorithm that aims to minimize the sum of the distances between each output price joint index sample and each cluster center with the corresponding fuzzy membership as the weight. The fuzzy membership index is used to characterize the degree to which each group of output price joint samples belongs to a certain cluster center. The fuzzy membership is determined based on the distance between the sample point and the cluster center. The larger the membership, the closer the sample point is to the cluster center. The basic flow chart of the FCM clustering algorithm is as follows: Figure 3 As shown, Figure 3 In the FCM clustering, u ij is the membership degree of the output price sample to cluster k, x i is the value of the output electricity price sample, c j is the eigenvector of cluster center j, and m is the fuzzy factor.

[0206] Finally, based on the weather condition distribution during the trading period predicted by the Markov chain model above, Latin hypercube sampling was performed on the output and electricity price samples clustered under different weather types to obtain output and electricity price data. That is, stratification was performed according to the clustering results, and random sampling was performed according to the sample ratio based on the number of output and electricity price samples of each type to simulate the photovoltaic output and electricity price scenarios during the future trading period.

[0207] The fourth step is to build a medium- and long-term trading curve decision model based on conditional value at risk, which includes the following:

[0208] Mainstream optimization methods can be divided into two categories: stochastic optimization and robust optimization. Stochastic optimization transforms uncertain optimization problems into deterministic optimization problems by generating random scenarios or characterizing probability density functions. The conditional value at risk (CVaR) optimization method is a typical stochastic optimization method that can take into account the influence of decision makers' risk preferences. It is a consistent risk measurement model proposed by Rockafeller and Uryase, which means that the loss amount of the decision result exceeds the conditional mean of the value at risk (VaR). CVaR can reflect all potential losses of the decision result, and the CVaR model does not rely on the decision payoff to conform to a symmetric distribution such as the normal distribution, which has great advantages, especially in the case of fat-tailed distributions.

[0209] VaR reflects the maximum potential loss of a portfolio at a given confidence level β∈(0,1). Let f(x,y) be the loss function, x be the decision variable, and y be the random variable. Assuming the joint probability density of the random variable y is p(y), the probability that the loss function f(x,y) does not exceed a certain loss level α is:

[0210] ψ(x,α)=∫f(x,y)≤α p(y)dy

[0211] Where ψ(x,α) is the cumulative distribution function of the loss of the decision variable x. Then for a given confidence level β, it is non-decreasing and right-continuous with respect to α. The calculation formulas for VaR and CVAR are as follows:

[0212] V VaR,β (x) = min{α ∈ R; ψ(x,α) ≥ β}

[0213]

[0214] Where V VaR,β (x) and V CVaR,β (x) are VAR and CVAR at the confidence level β.

[0215] At the same confidence level, the value of CVaR ≥ the value of VaR. Thus, minimizing CVaR is also minimizing VaR. Since it is difficult to directly solve V CVaR,β (x) which contains V VaR,β (x), the transformation function CVaR β is used to replace V CVaR,β (x):

[0216]

[0217] Where α is the value of VaR. The 0 integral term is discretized to obtain the expected value, and after simplification, it is as follows:

[0218]

[0219] Where: n = 1, 2,..., N represents the number of discrete samples under a certain probability distribution, and p n (y) is the occurrence probability of the discrete sample y.

[0220] (2) Medium- and long-term trading declaration decision-making of photovoltaic power stations based on conditional value at risk

[0221] When a photovoltaic power station participates in trading, the uncertainty of its output makes the power station face the risk of deviation settlement, and the uncertainty of the market electricity price makes the power station's bidding decision face the risk of whether it can be traded. Therefore, the objective function of new energy power stations participating in medium- and long-term trading comprehensively considers revenue and risk, uses CVaR as the risk measurement index, the model aims to maximize revenue and minimize risk, quantifies the risks brought by uncertain factors when the power station participates in trading, and provides the optimal decision-making scheme for the output curve and bidding. The specific framework is as Figure 4 shown.

[0222] The objective function can be expressed as:

[0223] max{E(R)-μ·CBaR β}

[0224]

[0225] Where E(R) is the expected revenue of the power station under different scenarios, CVaR β is the CVaR value under the confidence level β. μ is the risk preference factor, which indicates the risk aversion of the decision maker. The larger μ is, the more conservative the decision maker is, and the smaller μ is, the more inclined to high risk the decision maker is. π(i) is the probability of each output scenario, N is the number of output scenarios, and R i is the profit in the corresponding scenario, p R The decision quote is λ t The corresponding transaction probability is expressed as follows:

[0226]

[0227]

[0228] Revenue of new energy power station R i The difference between the declared output income and the deviation settlement. t=1,2,…T, where T is the total number of time periods. t , t P is the declared output value and electricity price of the power station at time t. t,i is the actual output value of the power station at time t. When the actual output of the power station is greater than the declared output, the difference between the two is the excess power generation The photovoltaic power station obtains electricity sales fees, and the unit electricity sales fee is λ + When the power station's declared output is greater than its actual output, the difference between the two is the under-generated power. The new energy power station pays for the electricity purchase, and the unit electricity purchase cost is λ - .

[0229] Transaction probability p R Related to market electricity prices and revenue distribution, the market electricity price level during the trading cycle will affect market users' willingness to trade. The transaction probability varies under different electricity price backgrounds, which brings uncertainty to photovoltaic power plants. This application uses a non-parametric kernel density method that can effectively fit any distribution to fit the revenue probability density function. The expression is as follows:

[0230]

[0231] In the formula, R j is the market return, N j is the number of samples, is the kernel density estimate of the return, K() is the kernel density function, and h is the bandwidth. This application uses the Gaussian kernel function as the kernel density function, and the Gaussian kernel function expression is:

[0232]

[0233] Wherein, is the distribution function value of the revenue R.

[0234] In addition, the constraint conditions also include the upper and lower limits of the declared output of the power station, the ramping rate constraint, and the upper and lower limits of the bid price. The expressions are as follows:

[0235] 0 ≤ P t ≤ P max,t

[0236]

[0237] λ min,t ≤ λ t ≤ λ max,t

[0238] Wherein, P max,t is the maximum output limit of the power station, are respectively the limits of the maximum power increase and decrease allowed per unit time for the power station, and λ min,t , λ max,t are the minimum and maximum values of the declared price.

[0239] The meaning of the optimal CVaR is to minimize the CVaR loss value at the confidence level β. The loss function f(x, y) is defined as -E(R). Set the confidence level to β, that is:

[0240]

[0241] For the convenience of solving, introduce a dummy variable z i , and let z i = max[-R i -α, 0] represents the loss exceeding VaR, and obtain the bidding optimization model based on CVaR and the transaction probability. The calculation formula is as follows:

[0242]

[0243] Since the transaction probability in the bidding optimization model based on CVaR and the transaction probability is related to the declared output, the model is changed into a two-stage problem and solved by an alternating iteration method. The first-stage problem is the declared output decision model considering CVaR, and the second-stage problem is the optimal output decision model considering revenue and the probability of being unsealed.

[0244] The first-stage problem only considers the uncertainty of the PV power plant output, and obtains the optimal declared output based on CVaR. The second-stage problem considers the impact of market electricity price uncertainty on the quotation. The probability of the power plant making a deal with users is related to the market revenue distribution. When the cost of users participating in medium- and long-term transactions is lower than the market revenue, the PV power plant can be successfully unblocked. To obtain a quotation decision aiming at successful transactions and maximizing revenue, it is necessary to consider both revenue and the probability of being unblocked.

[0245] Furthermore, the process of decision-making on output and quotation is as follows:

[0246] 1) First, according to the weather state distribution predicted by the weighted Markov chain, obtain the joint sample of output and electricity price indicators, and generate scenarios through Latin hypercube sampling. Use the first-stage model, that is, the optimal output decision model based on CVaR, to obtain the initial declared output value.

[0247] 2) According to the decision result of the first stage Calculate the revenue probability distribution function, and calculate the initial quotation value using the second-stage model.

[0248] 3) Substitute the initial quotation value into λ t , and update the declared output value according to the PV power plant output decision model based on CVaR.

[0249] 4) Iteratively solve P t , λ t until the convergence condition is met, that is, the power plant revenue changes little in the previous and subsequent iterations, or the given number of iteration steps or calculation time is reached, otherwise recalculate.

[0250] Therefore, a method for a PV power plant to declare an output curve and electricity price to participate in medium- and long-term transactions. In view of the output characteristics of the PV power plant, an optimization decision model for the medium- and long-term transaction output curve and quotation of the PV power plant considering conditional value at risk and transaction probability is established, so as to avoid risks for the PV power plant and ensure sustainable revenue.

[0251] An equipment embodiment applying the method of the present invention:

[0252] An electronic device, which includes:

[0253] One or more processors;

[0254] A storage device for storing one or more programs;

[0255] When the one or more programs are executed by the one or more processors, the one or more processors implement the above-mentioned medium- and long-term transaction declaration decision method for a PV power plant.

[0256] An embodiment of a computer medium applying the method of the present invention:

[0257] A computer-readable storage medium having a computer program stored thereon, which when executed by a processor implements the above-mentioned medium-term and long-term trading declaration decision-making method for a photovoltaic power station.

[0258] Those skilled in the art should understand that the embodiments of the present application can be provided as methods, systems, or computer program products. Therefore, the present application can take the form of a complete hardware embodiment, a complete software embodiment, or an embodiment combining software and hardware aspects. Moreover, the present application can take the form of a computer program product implemented on one or more computer-usable storage media (including but not limited to disk storage, optical storage, etc.) containing computer-usable program code.

[0259] The present application is described according to the flowcharts and / or block diagrams of the methods, devices (systems), and computer program products of the embodiments of the present application. It should be understood that each process and / or block in the flowchart and / or block diagram, as well as the combination of processes and / or blocks in the flowchart and / or block diagram, can be implemented by computer program instructions. These computer program instructions can be provided to the processor of a general-purpose computer, a special-purpose computer, an embedded processor, or other programmable data processing devices to generate a machine, such that the instructions executed by the processor of the computer or other programmable data processing devices generate a device for implementing the specified functions in Figure 1 one process or multiple processes or / and blocks Figure 1 one block or multiple blocks.

[0260] These computer program instructions can also be stored in a computer-readable memory that can direct a computer or other programmable data processing device to work in a specific manner, such that the instructions stored in the computer-readable memory generate a manufactured article including an instruction device that implements the specified functions in Figure 1 one process or multiple processes or / and blocks Figure 1 one block or multiple blocks.

[0261] These computer program instructions can also be loaded onto a computer or other programmable data processing device, such that a series of operation steps are executed on the computer or other programmable device to generate a computer-implemented process, and thus the instructions executed on the computer or other programmable device provide steps for implementing the specified functions in Figure 1 one process or multiple processes or / and blocks Figure 1 one block or multiple blocks.

[0262] The model in this application is an object that constitutes an objective descriptive morphological structure by means of physical or virtual representations. The object is not equal to an object and is not limited to physical and virtual. It can be a data processing function, a software program, a processing mode, a usage method, an operation method, a work process, an application process, electronic hardware, a circuit module, a processing system, a system imitation, or a simulation object.

[0263] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and are not intended to limit them. Although the present invention has been described in detail with reference to the above embodiments, those of ordinary skill in the art can still modify or equivalently replace the specific implementation manners of the present invention. Any modification or equivalent replacement without departing from the spirit and scope of the present invention shall be covered by the protection scope of the claims of the present invention.

Claims

1. A medium- and long-term trading declaration decision-making method for a photovoltaic power station, characterized in that: It includes the following steps: Step 1, through a pre-constructed weather prediction simulation model, obtain a certain trading cycle in the medium- and long-term trading of the photovoltaic power station to be decided, and predict the weather condition data within this trading cycle; Step 2, using a pre-constructed output characteristic extraction model, based on the weather condition data, extract the typical characteristics in the photovoltaic output and market electricity price data to obtain a combined sample of output electricity price indicators; Step 3, adopt a pre-constructed index generation model to perform clustering processing on the combined sample of output electricity price indicators to obtain combined output electricity price indicators; Step 4, use a pre-constructed declaration scenario simulation model, based on the combined output electricity price indicators, sample the combined sample of output electricity price indicators to obtain photovoltaic output and electricity price scenarios; Step 5, based on a pre-constructed optimal output decision-making model, with the goal of maximizing the profit, process the photovoltaic output and electricity price scenarios to obtain the declaration decision-making information of the photovoltaic power station, and complete the declaration decision-making of the medium- and long-term trading of the photovoltaic power station.

2. A medium- and long-term trading declaration decision-making method for a photovoltaic power station as described in claim 1, characterized in that: The method for obtaining a certain trading cycle in the medium- and long-term trading of the photovoltaic power station to be decided and predicting the weather condition data within this trading cycle through a pre-constructed weather prediction simulation model is as follows: Obtain a certain trading cycle in the medium- and long-term trading of the photovoltaic power station to be decided; According to a certain trading cycle, obtain the historical weather information at the corresponding moment, and obtain the weather states of several dates based on the historical weather information; The weather states are divided into sunny, cloudy or rainy, and are numbered 1, 2, and 3 respectively; Statistically analyze the transitions between the weather states of different dates to obtain the transition frequencies between each state; Using the pre-constructed multi-order weighted Markov chain model and calculating the s-order state transition probability matrix based on the transition frequencies between states The calculation formula is as follows: where x i and x i+s represent the weather conditions on the i-th and (i + s)-th days respectively, and represent the frequency and probability of the state transition from state a to state b respectively, where a, b ∈ {1, 2, 3}; After obtaining the s-order state probability matrices of different step lengths, confirm the corresponding weights according to their contribution degrees to the prediction result accuracy to obtain a weight matrix; Couple the weight matrix with the s-order state probability matrix to calculate the weighted transition probability matrix P W ; To obtain the weather distribution in the medium and long term in the future, based on the transition probability matrix P W , the stationary distribution of the Markov chain is determined; According to the ergodicity theorem, use the stationary distribution of the Markov chain to calculate the recurrence period of each state, and obtain the average recurrence period and stable probability of each weather state within this trading cycle; the specific calculation formula is as follows: where, π a is the stable probability with the weather state number a, where a = 1, 2, 3; Based on the average recurrence period and stable probability of each weather state within this trading cycle, predict the weather condition data within this trading cycle.

3. A medium- and long-term trading declaration decision-making method for a photovoltaic power station as described in claim 2, characterized in that: The method for constructing a multi-order weighted Markov chain model is as follows: Set a stochastic process {ξ(t), t ∈ T}, if for any integer {ξ(t), t ∈ T}, the state space is I, the parameter is a non-negative integer, and ξ(n) satisfies the probability distribution of the Markov chain, the probability distribution expression of the Markov chain is as follows: where P{ξ(n + 1) = j|ξ(n) = i} is a conditional probability, which includes a number of one-step transition probabilities p ij ; The one-step transition probability p ij is sorted in sequence to obtain a one-step transition probability matrix, and its expression is as follows: where p ij represents the probability at the next moment n + 1 under the condition that ξ(n) = i at moment n; Based on the transition probability matrix, determine the probability distribution {π, i ∈ I}; Take the probability distribution {π, i ∈ I} as the stationary distribution of the Markov chain, where I is the state space, and it satisfies the following relationship: That is, πP = π ∑ i∈I π i = 1 (i ∈ I).

4. A medium- and long-term trading declaration decision-making method for a photovoltaic power station as described in claim 3, characterized in that: After obtaining the s-order state probability matrices with different step sizes, the corresponding weights are confirmed according to their contribution degrees to the prediction result accuracy, and the method for obtaining the weight matrix is as follows: First, calculate the correlation between weather states to obtain the autocorrelation coefficient, which is used to judge the strength of the relationship between historical and current weather conditions. The calculation formula is as follows: After normalizing the autocorrelation coefficient, it is used as the weight value of each-order transition probability matrix. The calculation formula is as follows: Sum up the weight values to obtain the weight matrix.

5. A medium- and long-term trading declaration decision-making method for a photovoltaic power station as described in claim 1, characterized in that: In step two, using a pre-constructed output feature extraction model, based on weather condition data, extract typical features from photovoltaic output and market electricity price data. The method for obtaining the output electricity price index joint sample is as follows: Obtain weather condition data, as well as corresponding historical photovoltaic output and market electricity price data; For historical photovoltaic output, measure the output features from three dimensions: single point, overall, and fluctuation, including single-point evaluation indicators, overall evaluation indicators, and fluctuation evaluation indicators; The single-point evaluation index of historical PV output is the maximum daily PV output P max and the corresponding time t of the maximum output max , and its expression is as follows: P max = max{P1, P2, …, P n} Where P i (i = 1, 2, …, n) represents the output value of the PV power station at time i; The overall evaluation index of historical PV output is the expected value of PV output, P av and the probability of half-load output, p 0.5～1 ; The probability of half-load output represents the ratio of the number of times the output of the PV power station reaches half of the rated output throughout the day to the total frequency; The expression is as follows: p 0.5～1 = n 0.5～1 / n Since the output of photovoltaic power is affected by meteorological conditions and is random, corresponding volatility indicators are needed to measure the random fluctuations of the output. Therefore, the maximum fluctuation ratio ΔP max is selected as the representative volatility indicator; The expression is as follows: For market electricity price data, measure the electricity price features from three dimensions: single point, overall, and price difference, including single-point evaluation indicators, overall evaluation indicators, and price difference evaluation indicators; Due to the bimodal characteristic of the electricity price, it reaches one of the peaks in the morning, and the electricity price tends to decrease during the peak photovoltaic generation period at noon. Therefore, the single-point evaluation index of the market electricity price data is the maximum daytime electricity price p max and the minimum electricity price p min,noon at noon; its expression is as follows: p max = max{p1, p2, …, p n} p min,noon = min{p 1,noon , p 2,noon , …, p n,noon} where p i (i = 1, 2, …, n) represents the market electricity value at time i; The overall evaluation index of the market electricity price data for the whole day is selected as the expected value p of the daytime electricity price av ; Its expression is as follows: Since the daytime electricity price has obvious peak-valley characteristics, the peak-valley electricity price difference Δp noon is used as the price difference characteristic of the daytime electricity price; The expression is as follows: Δp noon = p max - p min,noon Normalize the single-point evaluation indicator of historical photovoltaic output, the overall evaluation indicator of historical photovoltaic output, the fluctuation evaluation indicator, the single-point evaluation indicator of market electricity price data, the overall evaluation indicator of market electricity price data, and the price difference evaluation indicator respectively to form the output electricity price index joint sample in historical photovoltaic output and market electricity price data; The calculation formula for normalization is as follows: In the formula, X represents each evaluation indicator.

6. A medium- and long-term trading declaration decision-making method for a photovoltaic power station as described in claim 1, characterized in that: In step three, use a pre-constructed index generation model to perform clustering processing on the output electricity price index joint sample to obtain the output electricity price joint index. The method is as follows: Obtain the output electricity price index joint sample; Based on the FCM clustering algorithm, calculate the clustering centers of the output electricity price index joint sample; According to the clustering centers, use FCM clustering on the output electricity price index joint sample to obtain the fuzzy membership matrix, which is used to describe the degree to which each group of output electricity price joint samples belongs to a certain clustering center; FCM clustering is an iterative optimization algorithm with the goal of minimizing the sum of the distances between each output electricity price joint index sample and each clustering center weighted by the corresponding fuzzy membership. The fuzzy membership is determined according to the distance between the sample point and the clustering center. The larger the membership, the closer the sample point is to the clustering center; According to the fuzzy membership matrix, obtain the output electricity price joint index; In step four, use a pre-constructed declaration scenario simulation model to sample the output electricity price index joint sample based on the output electricity price joint index to obtain the photovoltaic output and electricity price scenarios. The method is as follows: Obtain the output electricity price joint index and the output electricity price index joint sample; Calculate the proportion of the output electricity price joint index in the output electricity price index joint sample to obtain the sample proportion; According to the sample ratio, conduct Latin - Hypercube sampling on the weather condition data to simulate the photovoltaic power output and electricity price data during future trading periods; Generate photovoltaic power output and electricity price scenarios based on the photovoltaic power output and electricity price data.

7. A medium - and long - term trading declaration decision - making method for a photovoltaic power station as claimed in claim 1, wherein: Step Five, based on the pre - constructed optimal power output decision - making model, with the goal of maximizing the profit, process the photovoltaic power output and electricity price scenarios to obtain the declaration decision - making information of the photovoltaic power station as follows: Obtain the photovoltaic power output and electricity price scenarios; Based on the photovoltaic power output and electricity price scenarios, determine the declared power output value, electricity price at time t of the power station, and the actual power output value of the power station at time t; Calculate the revenue R for the corresponding scenario based on the declared output value, electricity price at time t of the power station, and the actual output value of the power station at time t i The calculation formula is as follows: Among them, R i is the revenue under the corresponding scenario, which is the difference between the declared output revenue and the deviation settlement; t = 1, 2, … T, where T is the total number of time segments; P t , λ t are the declared output value and electricity price of the power station at time t; P t,i is the actual output value of the power station at time t; is the over-generated electricity; is the under-generated electricity; When the actual power output of the power station is greater than the declared power output, the difference between the two is the excess power generation The PV power station obtains electricity sales revenue, and the unit electricity sales price is λ + , when the declared power output of the power station is greater than the actual power output, the difference between the two is the under-generation The PV power station pays for the purchased electricity, and the unit purchase price is λ - ; According to the revenue in the corresponding scenario, the probabilities of each output scenario, the number of output scenarios, and the decision-making bid price of λ t and the corresponding transaction probability, calculate the expected revenue of the power station under different scenarios; Among them, π(i) is the probability of each output scenario, N is the number of output scenarios, R i is the profit in the corresponding scenario, p R The decision quote is λ t The corresponding transaction probability when Based on the expected profit of the power station under different scenarios, and using Conditional Value at Risk (CVaR) as the risk measurement index and the expected profit of the power station under different scenarios, construct an objective function with the goal of maximizing the profit and minimizing the risk, to quantify the risk brought by uncertain factors when the power station participates in trading, so as to provide an optimal decision - making scheme for the power output curve and quotation; The expression of the objective function is as follows: max{E(R)-μ·CVaR β} where E(R) is the expected revenue of the power station under different scenarios, and CVaR β is the optimal CVaR value at the confidence level β; μ is the risk preference factor, representing the degree of risk aversion of the decision maker. The larger μ is, the more conservative the decision maker is, and the smaller μ is, the more inclined the decision maker is to high risks; Based on the objective function and constraints, obtain the initial declared output value in the first stage According to the initial declared output value in the first stage Calculate the revenue probability distribution function and the initial declared value in the second stage Substitute the initial declared value into λ t , and update the declared output value to obtain a new declared output value Iteratively solve for P t , λ t until the convergence condition is met, or the given number of iteration steps or calculation time is reached, and finally obtain the PV power station application decision information.

8. A medium - and long - term trading declaration decision - making method for a photovoltaic power station as claimed in claim 7, wherein: Set constraint conditions, which include upper and lower limits constraints on the declared power output of the power station, ramp rate constraints, and upper and lower limits constraints on the quotation. The specific expressions are as follows: 0 ≤ P t ≤ P max,t λ min,t ≤ λ t ≤ λ max,t where P max,t is the maximum output limit of the power station, are respectively the limits of the maximum power increase and decrease allowed per unit time in the power station, and λ min,t , λ max,t are the minimum and maximum values of the declared price; Or / and, the optimal CVaR value CVaR at the confidence level β β The calculation process is as follows: CVaR β The meaning of the optimal CVaR value is to minimize the CVaR loss value at the confidence level β. Set the loss function f(x, y) to -E(R) and the confidence level to β, that is: Introduce a dummy variable z i , let z i = max[-R i -α, 0] represents the loss exceeding VaR, and obtain the calculation formula of CVaR based on CVaR and the transaction probability, which is specifically as follows: β The calculation formula is as follows: Or / and, the method for obtaining the profit probability distribution function is as follows: Select the Gaussian kernel function as the kernel density function. The expression of the Gaussian kernel density function is: In the formula, is the distribution function value of the revenue R; According to the Gaussian kernel density function, use the non - parametric kernel density method that can effectively fit any distribution to fit the profit probability density function. The expression is as follows: where R j is the market return, N j is the number of samples, is the kernel density estimate of the return, i.e., the return probability distribution function, K() is the kernel density function, and h is the bandwidth.

9. A medium - and long - term trading declaration decision - making method for a photovoltaic power station, wherein: It includes the following steps: Step 1, according to the weather state distribution predicted by the weighted Markov chain, obtain the joint sample of the power output - electricity price index, conduct Latin - Hypercube sampling to generate photovoltaic power output and electricity price scenarios, and then use the optimal power output decision - making model based on CVaR to obtain the initial declared power output value. Step 2, calculate the profit probability distribution function based on the initial declared power output value in the first stage to obtain the initial quotation value. Step 3, use the initial quotation value as the electricity price of the power station at the next moment, and update the declared power output value according to the optimal power output decision - making model to obtain a new declared power output value. Step 4, iteratively solve the new declared power output value and the electricity price of the power station at the next moment until the convergence condition is met, or the given number of iterations or calculation time is reached. Otherwise, recalculate until finally obtaining the decision - making power output of the photovoltaic power station.

10. A medium - and long - term trading declaration decision - making system for a photovoltaic power station, wherein: It includes: One or more processors; A storage device for storing one or more programs; When the one or more programs are executed by the one or more processors, the one or more processors implement a medium - and long - term trading declaration decision - making method for a photovoltaic power station as claimed in any one of claims 1 - 9.

Images