Optimization method for optical storage system considering price and optical power uncertainty

Abstract

The application provides an optical storage system optimization method considering electricity price and light power uncertainty. The method firstly predicts the light power based on a long short-term memory network (LSTM), decomposes and predicts the electricity price signal by combining a variational mode decomposition (VMD) with an XGBoost algorithm, and obtains a light power and electricity price prediction sequence. Secondly, an empirical distribution is used to construct the marginal probability distribution of the light power and the electricity price, and a Copula function is used to construct a joint probability model to generate multiple groups of random scenes reflecting the uncertainty characteristics of the light power and the electricity price. Finally, a target function is established with the maximum system operation economy as the target, and a sequential quadratic programming method is used for solving to obtain a day-ahead power reference curve of the optical storage system, an electricity price response reference curve and an energy storage power allocation scheme. The obtained result can be used as an auxiliary decision basis for electric power market traders to improve the system economy under the premise of ensuring the safe operation of the power grid.

Description

Technical Field

[0001] This invention belongs to the field of photovoltaic-storage system collaborative optimization technology, and more specifically, it relates to a photovoltaic-storage system optimization method that considers the uncertainty of electricity price and optical power. Background Technology

[0002] With the continuous growth of installed capacity of new energy sources, photovoltaic power generation, due to its highly random and intermittent output, has become an important factor affecting the safe operation and economic dispatch of the power system. The integration of energy storage systems can, to some extent, smooth out photovoltaic power fluctuations, improve power quality, and provide flexible adjustment capabilities for new energy power plants. However, energy storage devices have power and capacity limitations, as well as aging issues. If not properly dispatched, they can easily lead to shortened equipment lifespan and increased system operating costs.

[0003] On the other hand, with the increasing sophistication of the electricity market mechanism, electricity price signals exhibit significant time-varying and uncertainties. When formulating day-ahead or intraday operation plans, photovoltaic (PV) power plants or operators must comprehensively consider the impact of electricity price fluctuations and PV power prediction errors to achieve a balance between economic efficiency and reliability. Traditional deterministic dispatch methods typically assume that PV power and electricity prices are known quantities, neglecting the impact of uncertainty on system economics and risks, which can easily lead to operational deviations, increased imbalance penalties, and reduced energy storage lifespan.

[0004] While existing research has introduced stochastic optimization or robust optimization methods, most of them only target a single source of uncertainty (such as optical power) or use simple historical averaging methods to describe changes in electricity prices, which cannot fully characterize the correlation between price and output.

[0005] In terms of electricity price forecasting, commonly used methods such as ARIMA or LSTM can reflect trends and cycles to a certain extent, but they are difficult to capture the multi-scale fluctuations and abrupt changes in price series. At the same time, there is a potential statistical correlation between optical power and electricity price signals. For example, electricity prices tend to fall when the light intensity is high, while they rise on cloudy days or when the optical power is low. How to accurately model this joint uncertainty is a key challenge to achieve the synergistic optimization of photovoltaic and energy storage systems.

[0006] Therefore, existing research has the following main shortcomings: (1) Insufficient modeling of uncertainty: Existing studies mostly focus on predicting and modeling optical power, lacking characterization of the time-varying and random nature of electricity price signals, making it difficult to reflect the statistical correlation between optical power and electricity price.

[0007] (2) Insufficient coordination and robustness of the optimization model: Most existing scheduling strategies adopt deterministic or single-scenario optimization methods, which fail to fully consider the impact of the joint uncertainty of power output and electricity price on the economic efficiency of system operation and energy storage life. The optimization results are prone to deviating from the actual optimal solution. Summary of the Invention

[0008] The purpose of this invention is to overcome the shortcomings of the prior art and provide an optimization method for photovoltaic-storage systems that considers the uncertainty of electricity price and optical power. This method solves the problems of insufficient modeling of uncertainty and insufficient synergy and robustness of optimization models in traditional optimization methods.

[0009] To achieve the above-mentioned objectives, the present invention provides an optimization method for a photovoltaic-storage system considering the uncertainties of electricity prices and optical power, comprising the following steps: (1) Optical power prediction; Historical data on solar irradiance, temperature, irradiance, and power output of photovoltaic power plants are collected and input into a Long Short-Term Memory (LSTM) network to predict the solar power output of the photovoltaic power plants in future periods. (2) Electricity price forecast; Variational Mode Decomposition (VMD) is used to decompose the historical electricity price time series signal into several intrinsic modes (IMFs) at multiple scales. Then, the boosting tree model (XGBoost) is used to predict each IMF, and the predicted values ​​of each IMF in the future period are obtained. Finally, the predicted values ​​of each IMF in the future period are superimposed to obtain the predicted value of electricity price in the future period. (3) Construct a standard uniform distribution model of optical power and electricity price; First, construct a marginal distribution model of optical power and electricity price, and then map the marginal distribution model to a standard uniform distribution model through probability integral transformation; (4) Joint uncertainty modeling; Based on the standard uniform distribution model, a joint probability distribution model between optical power and electricity price is established using the t-Copula function. Then, the optimal parameters of the joint probability distribution model are estimated using the maximum likelihood method, thereby obtaining the joint probability distribution model between optical power and electricity price. (5) Random scene generation; Based on the joint probability distribution model, multiple sets of random scenarios that can reflect the uncertainty characteristics of optical power and electricity price are generated, denoted as optical power-electricity price random scenarios; (6) Construct a collaborative optimization model for the photovoltaic-storage system; (7) Using the random scenario of optical power-electricity price as the input of the collaborative optimization model, the sequential quadratic programming algorithm is used to solve the collaborative optimization model to obtain the day-ahead power curve, the bidding plan curve and the energy storage power allocation scheme of the photovoltaic-storage system.

[0010] The objective of this invention is achieved as follows: This invention provides an optimization method for a photovoltaic-storage system considering the uncertainties of electricity price and photovoltaic power. First, it predicts photovoltaic power based on a long short-term memory network model and decomposes and predicts electricity price using variational mode decomposition combined with the XGBoost algorithm to obtain predicted sequences for photovoltaic power and electricity price. Second, it constructs marginal probability distributions for photovoltaic power and electricity price using empirical distributions and builds a joint probability model of photovoltaic power and electricity price using a Copula function. Finally, it establishes a comprehensive objective function that includes operational economics, maintenance costs, and unbalanced power penalty costs. With the goal of maximizing system operational economics, it solves the function using a sequential quadratic programming algorithm to obtain the day-ahead power reference curve, electricity price response reference curve, and energy storage power allocation scheme for the photovoltaic-storage system.

[0011] Meanwhile, the photovoltaic-storage system optimization method of the present invention, which considers the uncertainty of electricity price and optical power, also has the following beneficial effects: (1) Achieved joint uncertainty modeling of electricity price and optical power: This invention combines LSTM optical power prediction with VMD-XGBoost electricity price prediction and uses the Copula function to establish a joint probability model of the two, which can accurately characterize the stochastic correlation between electricity price and power output, and significantly improve the accuracy of prediction and the interpretability of the model.

[0012] (2) Improved optimization robustness and operation economy of photovoltaic-storage system: The collaborative optimization model constructed in this invention comprehensively considers operating costs, energy storage operation and maintenance costs and unbalanced power penalty terms in the objective function, so as to achieve a balance between profitability and stability of photovoltaic-storage system and effectively reduce operation deviation and scheduling risk.

[0013] (3) Improve the intelligence level of photovoltaic and energy storage system scheduling: This invention uses a collaborative modeling mechanism that combines deep learning and intelligent optimization algorithms to enable the operation decision-making process of photovoltaic and energy storage system to have stronger data-driven characteristics and adaptive capabilities, providing technical support for the market-oriented operation of new energy power plants. Attached Figure Description

[0014] Figure 1 This is a flowchart of an optimization method for a photovoltaic-storage system that considers the uncertainty of electricity price and optical power according to the present invention; Figure 2 This is a schematic diagram of the modal components after VMD decomposition; Figure 3 This is a line graph comparing the predicted and actual daily electricity prices on a certain day. Figure 4 It is a comparison curve of the daily predicted electricity price and the quotation plan for a typical scenario; Figure 5 These are the time-series curves of optical power, energy storage charging and discharging power, expected grid-connected power, and curtailed optical power. Figure 6This is the SOC change curve of the energy storage system within a day. Detailed Implementation

[0015] The specific embodiments of the present invention will now be described with reference to the accompanying drawings to enable those skilled in the art to better understand the invention. It should be particularly noted that in the following description, detailed descriptions of known functions and designs that might obscure the main content of the invention will be omitted here.

[0016] Example In this embodiment, as Figure 1 As shown, the present invention provides an optimization method for a photovoltaic-storage system considering the uncertainties of electricity prices and optical power, comprising the following steps: S1, Optical Power Prediction; Historical data on solar irradiance, temperature, irradiance, and power output of photovoltaic power plants are collected and input into a Long Short-Term Memory (LSTM) network to predict the solar power output of the photovoltaic power plants in future periods. S101. Data Preparation and Input Feature Construction: Select historical operating data of the target photovoltaic power plant for the past one to three years, including but not limited to photovoltaic power. Solar irradiance Ambient temperature Wind speed The data includes time information (hour, date, month, etc.), and is normalized and aligned to the specified time step. Construct a continuous input sample sequence within minutes; S102. LSTM Model Construction and Training: In this invention, the LSTM network is used to capture the temporal correlation between optical power and meteorological characteristics, and its structure is configured as follows: Input layer: The input dimensions include light intensity, irradiance, temperature, wind speed, and time features (hour, day of the week, month), totaling d dimensions; Hidden layers: Two LSTM units are set up, with 128 neurons in each layer; Output layer: Outputs the predicted optical power values ​​for the next 96 time steps (corresponding to 15-minute intervals and 24 hours); Activation function: The internal control unit uses a combination of tanh and sigmoid; Optimization algorithm: Adam optimizer is used, with a learning rate of 0.001 and a batch size of 32; Loss function: Mean Squared Error (MSE).

[0017] The network update calculation follows the standard LSTM state transition formula: ; in, These are the forget gate, input gate, and output gate, respectively. Candidate memory cell state, This represents the current state of the memory cell. Indicates the state of the memory unit at the previous moment; Output the hidden state at the current moment; This indicates the hidden state at the previous moment; This represents the Sigmoid activation function. Represents the Hadamard product. This represents the tanh activation function. This is the input vector at the current time t.

[0018] S103, Optical Power Prediction: After training, the model outputs optical power prediction values ​​for the next 96 time periods. ; S2, Electricity Price Forecast; Variational Mode Decomposition (VMD) is used to decompose the historical electricity price time series signal into several intrinsic modes (IMFs) at multiple scales. Then, the boosting tree model (XGBoost) is used to predict each IMF, and the predicted values ​​of each IMF in the future period are obtained. Finally, the predicted values ​​of each IMF in the future period are superimposed to obtain the predicted value of electricity price in the future period. In this embodiment, this step is used to predict electricity spot market prices: considering that electricity price time series exhibit both multi-scale fluctuation characteristics and nonlinear dynamic relationships, this invention adopts a combined prediction framework of "Variational Mode Decomposition (VMD) + XGBoost", the specific process of which is as follows: S201. Preparation of Historical Electricity Price Data: Collect historical electricity price series q(t) from the day-ahead electricity market and maintain consistency with the time resolution of the optical power prediction, setting the time step Δt = 15 minutes. The data sample includes, but is not limited to: Market clearing price (RMB / MWh); Time period characteristics (minutes, hours, dates, holidays, seasons, etc.); System load level and reserve ratio; Weather characteristics (temperature, irradiance, wind speed, etc., used to reflect the impact of the proportion of new energy sources); And perform normalization and time alignment processing; S202, Variational Mode Decomposition (VMD): Decomposing historical electricity price sequences in the electricity spot market. VMD decomposition is performed to obtain K non-overlapping intrinsic mode functions (IMFs), each IMF corresponding to a frequency band component of historical electricity prices. The objective function of VMD decomposition is: ; in, For the first k One modal component, The center frequency of the corresponding component. K The number of modes is taken in this embodiment. K=8 , Indicates a countdown of time. t Represents a time variable. j The imaginary unit, Represents the unit impulse function. Represents the convolution operator. The L2 norm is represented by the modal components after VMD decomposition, as shown below. Figure 2 As shown; S203, XGBoost Modal Prediction: For each modal component Each XGBoost model is constructed separately to predict future electricity prices. The input variables for each model include the following four features: 1. Short-term autoregressive characteristics: Take the modal hysteresis values ​​within the last 4 hours. It is used to reflect short-term fluctuation trends.

[0019] 2. Historical characteristics of the same period: Modal values ​​from the same time points of the previous day, two days, and the previous week are introduced. , , To capture intraday and weekly cyclical patterns.

[0020] 3. Sliding statistical characteristics: Calculate the mean of the historical window with standard deviation Used to describe the intensity and stability of fluctuations: ; in W Indicates the length of the history window. t Indexed by the current time. i For historical backtracking step size index, t i Indicates the current time t The first step of backtracking from the baseline i A historical point in time.

[0021] 4. Time cycle characteristics: In this embodiment, to reflect the cyclical pattern of electricity prices at different time scales, the time variable is encoded using a discretized identifier method, specifically including: Hourly characteristics: Indicates the hour number of the day; Weekly characteristics: This indicates the date within the week, where 1 represents Monday and 7 represents Sunday. Holiday characteristics: Setting Boolean variables If the day is a public holiday or a day off in lieu of a holiday, the value is 1; otherwise, the value is 0. Seasonal characteristics: based on months Based on this, the model is divided into four seasonal categories (spring, summer, autumn, and winter) and uses one-hot encoding as the input model.

[0022] All of the above time variables are input as discrete features into the XGBoost model, enabling the model to learn the intraday peak and valley, weekly patterns and seasonal variations of electricity prices through automatic splitting.

[0023] In this embodiment, the main hyperparameters of the XGBoost model are set to n_estimators=300, max_depth=6, learning_rate=0.05, subsample=0.8, and colsample_bytree=0.8. The model outputs modal predictions for the next 96 time periods (24 hours): ; final t The predicted future electricity price is obtained by superimposing the values: ; In this embodiment, as Figure 3 As shown in the figure, a line graph comparing the predicted and actual electricity prices on a certain day shows that the electricity price can be accurately predicted with a small error range.

[0024] S3. Construct a standard uniform distribution model of optical power and electricity price; S3.1 Constructing a marginal distribution model using an empirical distribution function: The edge distribution model of optical power is as follows: ; The marginal distribution model of electricity prices is as follows: ; Where 1() is an indicator function; if the condition in the parentheses is true, the function value is 1; if it is false, the function value is 0. for Predicted optical power for the time period for Electricity price forecast for the period Number of time periods; S3.2, Probability Integral Transformation; Based on optical power edge distribution model And the marginal distribution model of electricity prices The predicted optical power and electricity price are substituted into their respective edge cumulative distribution models, and the edge distribution models are mapped to the standard uniform distribution models through probability integral transformation. Standard uniform distribution model of optical power: ; Electricity price standard uniform distribution model: .

[0025] S4, Joint Uncertainty Modeling; S4.1 To capture the potential nonlinear correlation and tail dependence between optical power and electricity price, this embodiment uses t-Copula to construct a joint probability distribution model between optical power and electricity price: ; in, This represents the parameter set of the t-Copula function. , The correlation matrix represents the relationship between photovoltaic power and electricity prices. Indicates the degree of freedom parameter; Describing the degrees of freedom as of t Distribution inverse cumulative distribution function, Describing the degrees of freedom as Two-dimensional correlation matrix R t distributed; S4.2. Estimate the optimal parameters using the maximum likelihood method. ; ; in, The density function of the t-Copula function; S4.3 Construction of joint probability distribution; Based on Sklar's theorem, a joint probability distribution model of optical power and electricity price is constructed:

[0026] in, , These represent the marginal distributions of optical power and electricity price, respectively. Optical power is a random variable; Let be the random variable representing electricity price.

[0027] S5, Random Scene Generation; Multiple sets of random scenarios, denoted as optical power-electricity price random scenarios, are generated based on a joint probability distribution model to reflect the uncertainty characteristics of optical power and electricity price. The specific generation process is described below: S5.1, t-Copula spatial random sampling; Using a joint probability distribution model of optical power and electricity price, the following is generated from the t-Copula function: Two-dimensional standard uniform random samples: ; in, For the first s In each scenario t Uniform variable corresponding to optical power over time period; For the first s In each scenario t The uniform variable corresponding to the time-of-use electricity price; Number of scenes; T This represents the total number of time periods; S5.2, Inverse Edge Transformation; Marginal distribution of optical power and electricity price , Perform inverse cumulative distribution transformation: The optical power scenario is as follows: ; The electricity price scenario is as follows: ; in, and These are the marginal inverse distribution functions of photovoltaic power and electricity prices, respectively. Indicates the first s In each scenario t Optical power during the time period Indicates the first s In each scenario t Electricity prices during specific time periods; S5.3, Construct A random scenario of optical power versus electricity price: ; In this embodiment, based on the joint probability distribution model of optical power and electricity price obtained in step S3, multiple sets of random scenarios that can reflect the joint uncertainty characteristics of the two are generated by random sampling, providing input data for the subsequent day-ahead collaborative optimization of the photovoltaic-storage system.

[0028] S6. Construct a collaborative optimization model for the photovoltaic-storage system; The collaborative optimization model for photovoltaic and energy storage systems includes establishing an objective function and constraints to maximize expected operating benefits; The objective function, which aims to maximize the expected operating profit, is as follows:

[0029] in, Representing a scene The occurrence Probability , For the photovoltaic energy storage system in the firsts In each scenario t Electricity sales revenue during different time periods For the first s In each scenario t Time-based energy storage lifespan loss cost The interval between adjacent time periods is 15 minutes in this embodiment; For the first s In each scenario t The time-slot scheduling deviation penalty is used to balance economy and reliability; Grid-connected power indicates the power of the photovoltaic-storage system in the [number]th [period]. s In each scenario t The net output power to the grid during a given time period, with positive values ​​representing electricity sales and negative values ​​representing electricity purchases; This represents the charging and discharging power of the energy storage system; a positive value indicates discharging, and a negative value indicates charging. This indicates the deviation between actual and planned power. Representing a scene Down t The electricity price for a given time period satisfies: ; in, for t The price quote for a photovoltaic and energy storage power station during a specific time period. If the bid is deemed too high and the bid is not awarded, the electricity price is 0; otherwise, the bid is considered successful and the electricity price is equal to... ; The energy storage lifetime loss cost function, expressed using an empirical model, is as follows: ; in, The equivalent lifetime cost factor for energy storage per unit capacity is 10 yuan / MWh in this embodiment. The maximum charge and discharge power for energy storage is 50MW in this embodiment; The power imbalance penalty cost function, representing the negative impact of scheduling deviations on system economy, is defined as follows: ; in, The power imbalance penalty factor is 50 yuan / MWh in this embodiment; The constraints are: Power balance constraints of photovoltaic-storage systems:

[0030] in, The power balance constraint for photovoltaic-storage systems, which specifies the curtailed power, applies to both successful and unsuccessful bids. When a bid is successful, the following constraint must be met: curtailed power. The sum of energy storage power and grid-connected power equals the optical power: When the bid is not won: Grid-connected power Photovoltaic power is prioritized for energy storage charging; if the photovoltaic power exceeds the maximum charging power of the energy storage, then... This results in wasted light power. If the photovoltaic power is less than or equal to the maximum charging power of the energy storage Then all the photovoltaic power will be used for energy storage charging. Discarded light power ; Waste power constraint: ; Energy storage device charging and discharging power constraints: ; Remaining power constraints of energy storage devices: ; ; in, and They are respectively t +1 time period and t Remaining electricity stored during the time period; and These represent the energy storage charge and discharge efficiencies, both of which are 0.9 in this embodiment. The energy storage self-discharge coefficient is 0.05 in this embodiment; In this embodiment, the energy storage capacity is 100MWh; and These represent the maximum and minimum allowable remaining power of the energy storage system, respectively, and in this embodiment, they are both taken as 0.9. and 0.1 .

[0031] S7. Using the stochastic scenario of photovoltaic power-electricity price as the input of the collaborative optimization model, the sequential quadratic programming algorithm is used to solve the collaborative optimization model to obtain the day-ahead power curve, the bidding plan curve and the energy storage power allocation scheme of the photovoltaic-storage system.

[0032] In this embodiment, as Figure 4 As shown, the daily predicted electricity price curve and the optimized pricing plan curve are displayed for a typical scenario. From Figure 4 As can be seen, the bid price was lower than the corresponding predicted electricity price for most periods, indicating that the optimized model prioritizes ensuring the probability of winning the bid in order to obtain more tradable electricity.

[0033] Figure 5The figure illustrates the temporal changes in photovoltaic power, energy storage charging and discharging power, projected grid-connected power, and curtailed photovoltaic power. It shows that under the optimized strategy, the curtailed photovoltaic power in this embodiment remains zero, indicating that the photovoltaic power is fully utilized. The grid-connected power is composed of both projected photovoltaic power and energy storage power. During the midday period when electricity prices are low, the energy storage system is mostly in a charging state to absorb surplus photovoltaic power. During the nighttime period when electricity prices are higher, the energy storage system switches to a discharging state to increase grid-connected power, thereby improving economic benefits. This figure visually verifies that the model can reasonably allocate energy storage power under different electricity price levels, improving the utilization rate of photovoltaic resources.

[0034] Figure 6 The study demonstrates the SOC (State of Charge) changes of the energy storage system throughout the day. It shows that the SOC consistently fluctuates within the permissible range, without exhibiting adverse behaviors such as deep charging / discharging or high-frequency charging / discharging, indicating that the optimization model effectively avoids overuse of the energy storage system. The SOC rises during periods of low electricity prices and falls during periods of high electricity prices, consistent with the trend of electricity price changes, demonstrating a high degree of consistency between the energy storage dispatch strategy and the goal of maximizing revenue.

[0035] Calculations show that the predicted revenue in this scenario is 56,419.16 yuan. As shown in the attached figures, in the solution results of the collaborative optimization model of this invention, the output curve of the photovoltaic-storage system can track the changes in electricity price signals well, realizing an economical operation strategy of actively charging during low electricity price periods and discharging and selling electricity during high electricity price periods. Optimization results show that, while ensuring participation in bidding for most of the time, the system effectively reduces the curtailment of photovoltaic power and achieves full utilization of photovoltaic power. At the same time, the energy storage state of charge (SOC) curve is stable, with no frequent deep charging and discharging phenomena within the allowable range, reducing the cycle loss of energy storage equipment and extending the system life.

[0036] In summary, this invention organically combines optical power, electricity price signals, and energy storage operation characteristics to achieve intelligent and economical scheduling of photovoltaic-storage systems in complex market environments, and has significant engineering application value for the market-oriented operation and consumption optimization of new energy power plants.

[0037] Although the illustrative specific embodiments of the present invention have been described above to enable those skilled in the art to understand the invention, it should be understood that the invention is not limited to the scope of the specific embodiments. For those skilled in the art, various changes are obvious as long as they are within the spirit and scope of the invention as defined and determined by the appended claims, and all inventions utilizing the concept of the present invention are protected.

Claims

1. A method for optimizing a photovoltaic-storage system considering the uncertainties of electricity price and optical power, characterized in that, Includes the following steps: (1) Optical power prediction; Historical data on solar irradiance, temperature, irradiance, and power output of photovoltaic power plants are collected and input into a Long Short-Term Memory (LSTM) network to predict the solar power output of the photovoltaic power plants in future periods. (2) Electricity price forecast; Variational Mode Decomposition (VMD) is used to decompose the historical electricity price time series signal into several intrinsic modes (IMFs) at multiple scales. Then, the boosting tree model (XGBoost) is used to predict each IMF, and the predicted values ​​of each IMF in the future period are obtained. Finally, the predicted values ​​of each IMF in the future period are superimposed to obtain the predicted value of electricity price in the future period. (3) Construct a standard uniform distribution model of optical power and electricity price; First, construct a marginal distribution model of optical power and electricity price. Then, map the marginal distribution model to the standard uniform distribution space through probability integral transformation to obtain the standard uniform distribution model of optical power and electricity price. (4) Joint uncertainty modeling; Based on the standard uniform distribution model, the dependence structure model between optical power and electricity price is established using the t-Copula function. Then, the optimal parameters of the joint probability distribution model are estimated using the maximum likelihood method, and the joint probability distribution model of optical power and electricity price is obtained according to Sklar's theorem. (5) Random scene generation; Based on the joint probability distribution model, multiple sets of random scenarios that can reflect the uncertainty characteristics of optical power and electricity price are generated, denoted as optical power-electricity price random scenarios; (6) Construct a collaborative optimization model for the photovoltaic-storage system; (7) Using the random scenario of optical power-electricity price as the input of the collaborative optimization model, the sequential quadratic programming algorithm is used to solve the collaborative optimization model to obtain the day-ahead power curve, the bidding plan curve and the energy storage power allocation scheme of the photovoltaic-storage system.

2. The method for optimizing a photovoltaic-storage system considering the uncertainty of electricity price and optical power according to claim 1, characterized in that, The construction process of the standard uniform distribution model is as follows: (2.1) Constructing a marginal distribution model using an empirical distribution function: The edge distribution model of optical power is as follows: ; The marginal distribution model of electricity prices is as follows: ; Where 1() is an indicator function; if the condition in the parentheses is true, the function value is 1; if it is false, the function value is 0. for Predicted optical power for the time period for Electricity price forecast for the period Number of time periods; (2.2) Probability integral transformation; Based on optical power edge distribution model And the marginal distribution model of electricity prices Substituting the predicted optical power and electricity price into their respective edge cumulative distribution models, we obtain the standard uniform distribution model. Standard uniform distribution model of optical power: ; Electricity price standard uniform distribution model: .

3. The method for optimizing a photovoltaic-storage system considering the uncertainty of electricity price and optical power according to claim 1, characterized in that, The process of joint uncertainty modeling is as follows: (3.1) Establish a dependency structure model between optical power and electricity price using the t-Copula function: in, This represents the t-Copula function, used to describe the dependency structure between optical power and electricity price. This represents the parameter set of the t-Copula function. , The correlation matrix represents the relationship between photovoltaic power and electricity prices. Indicates the degree of freedom parameter; , represents the parameter set of the t-Copula function. The correlation matrix represents the relationship between photovoltaic power and electricity prices. Indicates the degree of freedom parameter; Describing the degrees of freedom as of t Distribution inverse cumulative distribution function, Describing the degrees of freedom as Two-dimensional correlation matrix R t distributed; (3.2) Estimate the optimal parameters using the maximum likelihood method. ; in, The density function of the t-Copula function; (3.3) Construction of joint probability distribution; Based on Sklar's theorem, a joint probability distribution model of optical power and electricity price is constructed: in, , These represent the marginal distributions of optical power and electricity price, respectively. Optical power is a random variable; Let be the random variable representing electricity price.

4. The method for optimizing a photovoltaic-storage system considering the uncertainty of electricity price and optical power according to claim 1, characterized in that, The random scene generation process is as follows: (4.1) Random sampling in t-Copula space; Using a joint probability distribution model of optical power and electricity price, the t-Copula function is generated. Two-dimensional standard uniform random samples: ； in, For the first s In each scenario t Uniform variable corresponding to optical power over time period; For the first s In each scenario t The uniform variable corresponding to the time-of-use electricity price; Number of scenes; (4.2) Edge inverse transformation; Marginal distribution of optical power and electricity price , Perform inverse cumulative distribution transformation: The optical power scenario is as follows: ; The electricity price scenario is as follows: ; in, and These are the marginal inverse distribution functions of photovoltaic power and electricity prices, respectively. Indicates the first s In each scenario t Optical power during the time period Indicates the first s In each scenario t Electricity prices during specific time periods; (4.3) Construct A random scenario of optical power versus electricity price: 。 5. The method for optimizing a photovoltaic-storage system considering the uncertainty of electricity price and optical power according to claim 1, characterized in that, The collaborative optimization model of the photovoltaic-storage system includes establishing an objective function and constraints to maximize the expected operating benefits. The objective function, which aims to maximize the expected operating profit, is as follows: in, Representing a scene The probability of occurrence, The interval between adjacent time periods. For the photovoltaic energy storage system in the first s In each scenario t Electricity sales revenue during different time periods For the first s In each scenario t Time-based energy storage lifespan loss cost For the first s In each scenario t The time-slot scheduling deviation penalty is used to balance economy and reliability; Grid-connected power indicates the power output of the photovoltaic-storage system in the [number]th [period]. s In each scenario t The net output power to the grid during a given time period, with positive values ​​representing electricity sales and negative values ​​representing electricity purchases; This represents the charging and discharging power of the energy storage system; a positive value indicates discharging, and a negative value indicates charging. This indicates the deviation between actual and planned power. Representing a scene Down t The electricity price for a given time period satisfies: ； in, for t The price quote for a photovoltaic and energy storage power station during a specific time period. If the bid is deemed too high and the bid is not awarded, the electricity price is 0; otherwise, the bid is considered successful and the electricity price is equal to... ; The energy storage lifetime loss cost function, expressed using an empirical model, is as follows: ； in, The equivalent lifetime cost coefficient for energy storage per unit capacity; This represents the maximum charging and discharging power of the energy storage. Let represent the power imbalance penalty cost function, and define it as follows: ； in, This is the power imbalance penalty coefficient; The constraints are: Power balance constraints of photovoltaic-storage systems: ; in, The power balance constraint for photovoltaic-storage systems, which specifies the curtailed power, applies to both successful and unsuccessful bids. When a bid is successful, the following constraint must be met: curtailed power. The sum of energy storage power and grid-connected power equals the optical power: When the bid is not won: Grid-connected power Photovoltaic power is prioritized for energy storage charging; if the photovoltaic power exceeds the maximum charging power of the energy storage, then... This results in wasted light power. If the photovoltaic power is less than or equal to the maximum charging power of the energy storage Then all the photovoltaic power will be used for energy storage charging. Discarded light power ; Waste power constraint: ; Energy storage device charging and discharging power constraints: ; Remaining power constraints of energy storage devices: ; ； in, and They are respectively t +1 time period and t Remaining electricity stored during the time period; and These are the energy storage charging and discharging efficiencies, respectively. The self-discharge coefficient of energy storage; For energy storage capacity; and These represent the maximum and minimum allowable remaining power of the energy storage system.

Images