A charging and discharging optimization scheduling method and device

Abstract

The application provides a charging and discharging optimization scheduling method and device, relates to the technical field of energy storage scheduling, and comprises the following steps: acquiring power multi-source heterogeneous data, constructing a price probability interval prediction model, generating price probability parameters based on the power multi-source heterogeneous data, generating a target price scenario set based on the price probability parameters, constructing a negative income risk threshold based on the target price scenario set, constructing a target optimization model based on the negative income risk threshold, and generating an optimal charging and discharging power sequence containing optimal charging and discharging state variables and optimal power continuous variables. Therefore, the robustness and reliability of the energy storage arbitrage strategy can be improved.

Description

Technical Field

[0001] This invention relates to the field of energy storage scheduling technology, and more specifically, to a charging and discharging optimization scheduling method and apparatus. Background Technology

[0002] With the development of the times, the price volatility of the electricity spot market has become increasingly significant. This volatility has created huge commercial opportunities for energy storage systems through the "low-charge, high-discharge" arbitrage model. The economic benefits of energy storage power stations are highly dependent on the precision of their charging and discharging timing, a characteristic that makes energy storage operation exhibit significant "time sensitivity".

[0003] Current energy storage dispatching technologies generally adopt a two-stage "prediction-optimization" paradigm: First, deep learning models (such as LSTM and Transformer) are used to predict electricity prices at 96 time points over the next 24 hours, generating a deterministic electricity price sequence; then, this predicted sequence is input into optimization models such as Mixed Integer Linear Programming (MILP) to solve for the theoretically optimal charging and discharging strategy. This approach can achieve the expected returns when the electricity price curve maintains the predicted shape, but when the actual peak and off-peak electricity prices deviate from the prediction, the preset "low charging, high discharging" strategy will fail. More seriously, this point prediction method cannot provide information on the probability distribution of electricity price fluctuations, making the optimization decision lack the ability to quantitatively assess time uncertainty, thereby reducing the robustness and return stability of the overall dispatching strategy. Summary of the Invention

[0004] In view of this, the purpose of the present invention is to provide a charging and discharging optimization scheduling method and apparatus, which can directly quantify and control the negative returns caused by the time shift of electricity price peaks and valleys by constructing a negative return risk threshold as a risk indicator, thereby significantly improving the robustness and reliability of energy storage arbitrage strategies.

[0005] To achieve the above objectives, the technical solutions adopted in the embodiments of the present invention are as follows: In a first aspect, the present invention provides a charging and discharging optimization scheduling method, applied to an energy storage system, comprising: Acquire multi-source heterogeneous power data, which should include at least meteorological data, historical power output data of renewable energy power plants, historical electricity prices, and transaction clearing data. A probability interval prediction model for electricity prices is constructed, and the probability parameters for electricity prices are generated based on multi-source heterogeneous power data using the electricity price probability interval prediction model. A target electricity price scenario set is generated based on the electricity price probability parameter. The target electricity price scenario set includes multiple target electricity price scenarios, and the sum of the probabilities of occurrence of each target electricity price scenario is 1. Construct a negative return risk threshold based on a target electricity price scenario set; A scheduling planning model is constructed based on the negative return risk threshold, and the scheduling planning model is used as the target optimization model. The optimal charging and discharging power sequence is determined based on the target optimization model. The optimal charging and discharging power sequence includes the optimal charging and discharging state variables and the optimal power continuous variables.

[0006] Optionally, the scheduling planning model includes an objective function, and the steps for determining the optimal charging and discharging power sequence based on the objective optimization model include: Obtain the physical constraints of the energy storage system. The physical constraints include at least the continuity constraints of the energy storage state of charge, the upper and lower limits of capacity, the upper and lower limits of charge and discharge power, and the mutual exclusion constraints of charge and discharge. Convert the negative return risk threshold into a linear constraint; The global optimal solution of the objective function is determined based on physical constraints and linear constraints, and the global optimal solution is used as the optimal charge and discharge power sequence.

[0007] Optionally, the charge / discharge optimization scheduling method also includes: Real-time electricity prices and the real-time state of charge of the energy storage system are obtained at preset time intervals. Update the scheduling planning model using real-time electricity prices and real-time state of charge. Using the updated scheduling planning model as the target optimization model, the steps of determining the optimal charging and discharging power sequence based on the target optimization model are executed.

[0008] Optionally, the scheduling planning model includes an objective function, which is expressed as follows: ; In the formula, The objective function value, This is the risk aversion coefficient. The threshold for negative return risk. For the benefit in the s-th target scenario, The total number of target scenarios, Let be the probability of occurrence corresponding to the s-th target scenario.

[0009] Optionally, the electricity price probability interval prediction model includes a deep learning network. The steps for generating electricity price probability parameters based on multi-source heterogeneous power data using the electricity price probability interval prediction model include: Candidate feature sets are extracted from multi-source heterogeneous power data; the candidate feature sets include at least: time dimension features, historical electricity price lag features, and meteorological and load coupling features; The Pearson method is used to screen target features from the candidate feature set, and the electricity price time series is obtained based on the target features; the target features are used to characterize features that are linearly correlated with electricity prices. Multiple quantile electricity price forecasts were generated based on electricity price time series using deep learning networks. Electricity price probability parameters are generated based on the predicted electricity prices for each quantile.

[0010] Optionally, the training function of the deep learning network includes minimizing the bouncing loss function, wherein, for a single quantile, the formula for minimizing the bouncing loss function is expressed as: ; In the formula, For a moment Next Minimize the bouncing loss function value at each quantile. For a moment Next The true value at each quantile; For a moment Next Predicted quantiles at each quantile.

[0011] Optionally, the steps for generating the target electricity price scenario set based on the electricity price probability parameters include: Monte Carlo simulation of electricity price probability parameters is performed to generate an initial set of electricity price scenarios; Calculate the Wasserstein distance between any two scenarios in the initial electricity price scenario set; The target electricity price scenario set is obtained from the initial electricity price scenario set using the synchronous back-substitution elimination method.

[0012] Optionally, the steps of obtaining the target electricity price scenario set from the initial electricity price scenario set using the synchronous back-substitution elimination method include: In each iteration, the scenario with the smallest Wasserstein distance in the current initial electricity price scenario set is removed, and the probability corresponding to the removed scenario is accumulated to the target scenario. The target scenario is used to represent the retained scenario that is closest to the removed scenario, until the number of retained scenarios in the current iteration number is equal to a preset value, and the target electricity price scenario set is constructed based on the retained scenarios corresponding to the current iteration process.

[0013] Optionally, the steps for constructing a negative return risk threshold based on the target electricity price scenario set include: Risk thresholds are determined based on the revenue of energy storage systems within a scheduling cycle. Calculate the expected loss when the risk threshold is exceeded, and use the expected loss as the negative return risk threshold. The formula for calculating the negative return risk threshold is as follows: ; In the formula, The threshold for negative return risk. As a risk threshold, For confidence level, For the profit / loss function Uncertainty in actual electricity prices The function value between, This is the expectation operator. Secondly, the present invention also provides a charge-discharge optimization scheduling device for use in an energy storage system, comprising: The data acquisition module is used to acquire multi-source heterogeneous power data, which includes at least meteorological data, historical power output data of new energy power plants, historical electricity prices and transaction clearing data. The electricity price probability parameter generation module is used to construct an electricity price probability interval prediction model and generate electricity price probability parameters based on multi-source heterogeneous power data using the electricity price probability interval prediction model. The target electricity price scenario set construction module is used to generate a target electricity price scenario set based on the electricity price probability parameter. The target electricity price scenario set includes multiple target electricity price scenarios, and the sum of the occurrence probabilities of each target electricity price scenario is 1. The negative return risk threshold construction module is used to construct negative return risk thresholds based on the target electricity price scenario set. The decision-making module is used to construct a scheduling planning model based on a negative return risk threshold, and to optimize the model with the scheduling planning model as the objective. Based on the objective optimization model, the optimal charging and discharging power sequence is determined. The optimal charging and discharging power sequence includes the optimal charging and discharging state variables and the optimal power continuous variables.

[0014] This invention provides a charging and discharging optimization scheduling method and apparatus. By acquiring multi-source heterogeneous power data and constructing a power price probability interval prediction model, a power price probability parameter is generated based on the multi-source heterogeneous power data. Subsequently, a target power price scenario set is generated based on the power price probability parameter, and a negative return risk threshold is constructed based on the target power price scenario set. Then, a target optimization model is constructed based on the negative return risk threshold to generate an optimal charging and discharging power sequence that includes the optimal charging and discharging state variable and the optimal power continuous variable. This guides the energy storage system in charging and discharging, thereby significantly improving the robustness and reliability of the energy storage arbitrage strategy.

[0015] To make the above-mentioned objects, features and advantages of the present invention more apparent and understandable, preferred embodiments are described below in detail with reference to the accompanying drawings. Attached Figure Description

[0016] To more clearly illustrate the technical solutions of the embodiments of the present invention, the accompanying drawings used in the embodiments will be briefly introduced below. It should be understood that the following drawings only show some embodiments of the present invention and should not be regarded as a limitation on the scope. For those skilled in the art, other related drawings can be obtained based on these drawings without creative effort.

[0017] Figure 1This invention provides a flowchart of one of the steps of a charge / discharge optimization scheduling method. Figure 2 A flowchart of step 200 provided in an embodiment of the present invention is shown; Figure 3 A flowchart of step 300 provided in an embodiment of the present invention is shown; Figure 4 A flowchart of step 400 provided in an embodiment of the present invention is shown; Figure 5 A flowchart of step 500 provided in an embodiment of the present invention is shown; Figure 6 This illustrates a second flowchart of the charging and discharging optimization scheduling method provided in this embodiment of the invention. Figure 7 A block diagram of the charge-discharge optimization scheduling device provided in an embodiment of the present invention is shown; Figure 8 A block diagram of a server provided in an embodiment of the present invention is shown.

[0018] Icons: 10-Charging and discharging optimization scheduling device; 11-Data acquisition module; 12-Electricity price probability parameter generation module; 13-Target electricity price scenario set construction module; 14-Negative return risk threshold construction module; 15-Decision construction module; 20-Server; 21-Memory; 22-Processor; 23-Communication module. Detailed Implementation

[0019] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. The components of the embodiments of the present invention described and shown in the accompanying drawings can generally be arranged and designed in various different configurations.

[0020] Therefore, the following detailed description of the embodiments of the invention provided in the accompanying drawings is not intended to limit the scope of the claimed invention, but merely to illustrate selected embodiments of the invention. All other embodiments obtained by those skilled in the art based on the embodiments of the invention without inventive effort are within the scope of protection of the invention.

[0021] It should be noted that relational terms such as "first" and "second" are used merely to distinguish one entity or operation from another, and do not necessarily require or imply any such actual relationship or order between these entities or operations. Furthermore, the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such a process, method, article, or apparatus. Without further limitations, an element defined by the phrase "comprising one..." does not exclude the presence of other identical elements in the process, method, article, or apparatus that includes said element.

[0022] The inventors discovered that most existing technologies are based on "point prediction" logic to achieve deterministic optimization scheduling. However, in actual operation, this model faces a severe challenge, such as the destructive nature of peak-valley shifts in electricity prices, the collapse of the profit base, and the disconnect between prediction and decision-making.

[0023] Based on this, the present invention establishes a prediction-decision integrated framework decision improvement scheme that can quantify "time offset risk" and deeply embed the probability characteristics of price fluctuations into the decision objective, so as to improve the robustness and reliability of energy storage arbitrage strategies.

[0024] Please refer to Figure 1 , Figure 1 The flowchart of the charge-discharge optimization scheduling method provided in the embodiment of the present invention is shown. The charge-discharge optimization scheduling method is applied to an energy storage system and includes steps 100 to 500.

[0025] Step 100: Obtain heterogeneous power data from multiple sources.

[0026] In this embodiment, the multi-source heterogeneous data includes at least meteorological data, historical power output data of new energy power plants, historical electricity prices, and transaction clearing electricity data.

[0027] Step 200: Construct an electricity price probability interval prediction model and use the electricity price probability interval prediction model to generate electricity price probability parameters based on multi-source heterogeneous power data; Step 300: Generate a target electricity price scenario set based on the electricity price probability parameters.

[0028] The target electricity price scenario set includes multiple target electricity price scenarios, and the sum of the probabilities of occurrence of each target electricity price scenario is 1.

[0029] Step 400: Construct a negative return risk threshold based on the target electricity price scenario set; Step 500: Construct a scheduling planning model based on the negative return risk threshold, optimize the model with the scheduling planning model as the objective, and determine the optimal charging and discharging power sequence based on the objective optimization model.

[0030] In this embodiment, the optimal charge / discharge power sequence includes the optimal charge / discharge state variable and the optimal power continuous variable.

[0031] This application constructs a negative return risk threshold based on a target electricity price scenario set and uses this negative return risk threshold as a risk indicator. This risk indicator is then used in the scheduling planning model to generate the optimal charging and discharging power sequence of the energy storage system for the next 24 hours, taking into account both risk and return. In turn, the electricity price probability prediction result is directly transformed into a controllable risk indicator CVaR, which is then used as part of the scheduling planning model. This achieves a seamless connection from the quantification of electricity price uncertainty to risk-averse charging and discharging decisions. It overcomes the problems of poor robustness of energy storage arbitrage decisions to electricity price uncertainty (especially peak-valley time shift), insufficient risk quantification, and the disconnect between prediction and decision-making in existing technologies, thereby significantly improving the reliability of energy storage arbitrage strategies.

[0032] To further improve the accuracy of the optimal charge and discharge power sequence, step 100 of acquiring multi-source heterogeneous power data in this embodiment includes: Initial meteorological data, initial historical power output data of renewable energy power plants, initial historical electricity prices, and initial cleared electricity volume data are obtained using an external multi-source data structure. Subsequently, data cleaning is performed on each initial data set to obtain the multi-source heterogeneous power data.

[0033] In one possible implementation, this embodiment can access meteorological data via an API interface or an IoT gateway, which includes at least humidity, horizontal irradiance, and wind speed.

[0034] In one possible implementation, this embodiment can access the historical power output data of new energy power plants through the power system. The historical power output data of new energy power plants includes at least photovoltaic active power and wind power active power.

[0035] In one possible implementation, this embodiment accesses historical electricity prices and traded clearing volume data through the power trading center interface.

[0036] After obtaining the above initial data, this embodiment can use the 3σ principle for cleaning. Specifically, this embodiment can identify abnormal data such as continuous and unchanging "dead values", "out-of-bounds values" that exceed physical extreme values ​​and mutation points under each initial data based on the 3σ principle, and then remove each abnormal data.

[0037] In addition to removing outlier data, some data also needs to be supplemented. One possible approach is to use linear interpolation to fill in short-term missing data (e.g., less than 1 hour), while for long-term missing data (e.g., more than 24 hours), a similar daily correction method is used, selecting data from days with similar historical meteorological conditions for weighted replacement. The cleaned data is then stored in the database to obtain the final multi-source heterogeneous power data.

[0038] To address the shortcomings of traditional point prediction in reflecting uncertainty, this embodiment can construct an electricity price probability interval prediction model to generate a continuous probability distribution, namely the electricity price probability parameter.

[0039] Please refer to Figure 2 , Figure 2 The flowchart of step 200 provided in this embodiment of the invention is shown. In this embodiment, step 200, which uses the electricity price probability interval prediction model to generate electricity price probability parameters based on multi-source heterogeneous power data, includes steps 201 to 204.

[0040] Step 201: Extract candidate feature sets from multi-source heterogeneous power data.

[0041] In this embodiment, the candidate feature set includes at least: time dimension features, historical electricity price lag features, and meteorological and load coupling features.

[0042] Step 202: Use the Pearson method to screen target features from the candidate feature set, and obtain the electricity price time series based on the target features.

[0043] In this embodiment, the target feature is used to characterize features that are linearly related to electricity prices.

[0044] Step 203: Use a deep learning network to generate multiple quantile electricity price prediction values ​​based on the electricity price time series.

[0045] Step 204: Generate electricity price probability parameters based on the predicted electricity prices for each quantile.

[0046] To improve data standardization and eliminate redundant noise, this embodiment can prioritize extracting candidate feature sets from multi-source heterogeneous power data. The time-dimensional features in step 201 can include hourly, weekly, monthly / seasonal, and holiday markers. Historical electricity price lag features can include short-term lag sequences, periodic lag sequences, and statistical aggregation features; among them, statistical aggregation features are used to characterize moving averages or maximum / minimum values ​​calculated based on a sliding window. Meteorological and load coupling features can include degree indicators, wind force indicators, solar radiation indicators, and system load forecasts.

[0047] Subsequently, in order to solve the "curse of dimensionality" and model overfitting problems caused by excessively high input feature dimensionality, this embodiment can screen key variables that have a significant impact on electricity prices from the above candidate feature set based on the Pearson correlation coefficient, namely the target features.

[0048] Specifically, set the target variable Electricity price at any time Any candidate feature variable is At this point, the Pearson correlation coefficient between the two is... The calculation formula can then be expressed as:

[0049] In the formula, For target variable Electricity price at any time The corresponding standard deviation For any candidate feature variable The corresponding standard deviation For target variable Electricity price at any time and any candidate feature variables covariance, For the first The mean of the candidate feature variable corresponding to the sample. For any candidate feature variable The corresponding mean, For the first The sample corresponds to the electricity price. Let be the mean of the electricity price series. This represents the number of samples.

[0050] After obtaining the Pearson correlation coefficients between each candidate feature in the candidate feature set and the target electricity price, a preset Pearson correlation coefficient value can be set. To filter target features, for example, a preset Pearson correlation coefficient value can be set. satisfy: Based on this, if the absolute value of the Pearson correlation coefficient corresponding to any candidate feature is greater than the preset Pearson correlation coefficient value... If the correlation between the candidate feature and the electricity price is significant, then the candidate feature is determined to be a target feature; otherwise, it is discarded. Furthermore, if the correlation frequency between two target features exceeds a preset correlation value, for example, if two input features... and If the correlation coefficient between the two is extremely high (e.g., >0.9), it indicates that there is information redundancy on the surface. Therefore, based on the correlation between the two and the target electricity price, the one with a stronger correlation with the target electricity price should be retained.

[0051] Based on this, this embodiment can determine and obtain the electricity price time series in the manner described above.

[0052] In one possible implementation, the electricity price probability interval prediction model can be constructed from a deep learning network. In this embodiment, the deep learning network can be constructed from a Long Short-Term Memory (LSTM) network or a Transformer model.

[0053] This deep learning network includes a feature extractor and a quantile regression output layer. Taking the aforementioned electricity price time series as input, it generates a context feature vector through its own nonlinear activation function and weight matrix operations. Then, based on the output layer constructed by quantile regression, multiple quantiles are set on the output layer, thereby outputting the electricity price prediction value corresponding to each quantile based on the fully connected layer, i.e., multiple quantile electricity price prediction values.

[0054] In the (future) moment The electricity price forecast at any quantile can be expressed as: In the formula, and It is a quantile The corresponding learning weights and bias parameters, This is the output vector at the current time step t.

[0055] To further improve the accuracy of deep learning networks, training can be performed based on minimizing the bouncing loss function. In this embodiment, the formula for minimizing the bouncing loss function for a single quantile is expressed as follows: ; In the formula, For a moment Next Minimize the bouncing loss function value at each quantile. For a moment Next The true value at each quantile; For a moment Next Predicted quantiles at each quantile.

[0056] Based on this, when N training samples are included, the loss objective of this deep learning network is the weighted sum of the losses of all quantiles, which can be expressed as: In the formula, T Let M be a set containing M quantiles, satisfying: ; For a moment i Next Minimize the bouncing loss function value at each quantile.

[0057] Specifically, in this embodiment, the weighted sum of the loss at this quantile can be minimized using the backpropagation algorithm. To update quantiles Corresponding learning weights and bias parameters After the above model is trained, it outputs a series of quantiles. At this time, time... At a confidence level of The electricity price forecast range can be expressed as: ; For a moment At a confidence level of The predicted quantiles below, For a moment At a confidence level of The predicted quantiles below.

[0058] In summary, the electricity price probability parameter in this embodiment is essentially a series of quantiles.

[0059] To further improve the shortcomings of traditional point forecasting in failing to reflect uncertainty, this embodiment can construct a continuous probability distribution based on the above series of quantiles, i.e., the electricity price probability parameter.

[0060] In one possible approach, discrete quantile data predicted for each time period can be obtained, and then the discrete quantile data can be smoothed based on kernel density estimation methods, such as the Gaussian kernel function, to obtain the continuous probability density function of the electricity price at each future time.

[0061] In this embodiment, the electricity price probability parameter can be expressed as: , The value used to characterize the electricity price at time t is... The probability density is given by One predicted quantile Centered on, with It is formed by superimposing small Gaussian distributions with standard deviations; For a moment i Next Predicted quantiles at each quantile.

[0062] This embodiment can convert a continuous probability distribution into a discrete set of scenes that can be processed by a computer. At the same time, by reducing the number of scenes, the computational complexity is reduced while retaining the statistical features.

[0063] Please refer to Figure 3 , Figure 3 The flowchart of step 300 provided in the embodiment of the present invention is shown. In this embodiment, step 300, which generates a target electricity price scenario set based on the electricity price probability parameter, includes steps 301 to 303.

[0064] Step 301: Perform Monte Carlo simulation on the electricity price probability parameters to generate an initial electricity price scenario set.

[0065] Step 302: Calculate the Wasserstein distance between any two scenarios in the initial electricity price scenario set.

[0066] Step 303: Obtain the target electricity price scenario set from the initial electricity price scenario set using the synchronous back-substitution elimination method.

[0067] In this embodiment, the target electricity price scenario set is used to characterize the discrete scenario set that is most similar to the initial electricity price scenario set in terms of Wasserstein distance.

[0068] Specifically, Monte Carlo simulation is performed using methods such as inverse transform sampling to generate N distinct initial electricity price scenarios for the next 24 hours. In this embodiment, each initial electricity price scenario can be defined as a 96-dimensional vector. , represented as: ;in , represents the scene number, and N is the total number of scenes generated in the Monte Carlo simulation. Indicates the first In the first scenario, the... Predicted electricity value per hour (unit: yuan / kWh or $ / MWh).

[0069] Among them, the initial electricity price scenario set This can be understood as a set of binary pairs consisting of a scene vector and its corresponding probability of occurrence. in: For the first The probability of each scenario occurring.

[0070] The initial electricity price scenario set directly generated by Monte Carlo simulation We can assume that each scenario has an equal probability, that is... .

[0071] Subsequently, Wasserstein distance was used to measure the similarity between different electricity price curves.

[0072] Specifically, first define any two scenarios and Geometric distance between , can be represented as: , It is the Euclidean norm.

[0073] And in each iteration, the current original scene set is calculated. For each scenario in the initial electricity price scenario set, if the scenario is removed and its probability is transformed to the Wasserstein distance generated by the nearest neighbor field, it can be expressed as: Simultaneously, scenarios with the minimum Wasserstein distance in the current initial electricity price scenario set will be removed, and these removed scenarios will be discarded. probability Accumulate to the nearest preserved scene In the target scenario, that is: Repeat the above iterative process until the number of retained scenarios reaches the preset value. The target electricity price scenario set obtained at this time is the discrete scenario set that is most similar to the original distribution in the sense of Wasserstein distance.

[0074] To quantify the potential economic losses from "buying high and selling low" in energy storage due to deviations in electricity price forecasts, this embodiment can convert uncertain risks into mathematically calculable indicators.

[0075] Please refer to Figure 4 , Figure 4 The flowchart of step 400 provided in the embodiment of the present invention is shown. In this embodiment, step 400, which constructs a negative return risk threshold based on the target electricity price scenario set, includes steps 401 to 402.

[0076] Step 401: Determine the risk threshold based on the revenue of the energy storage system within a scheduling cycle.

[0077] Step 402: Calculate the expected loss when the risk threshold is exceeded, and use the expected loss as the negative return risk threshold.

[0078] This embodiment can construct a benefit / loss function based on the charging and discharging behavior of the energy storage system at various times. It can be represented as: ; This refers to discharge behavior; For charging behavior.

[0079] Subsequently, the electricity prices for each scenario within the quantifiable target electricity price scenario set can be determined. , can be represented as: , This refers to the electricity price at any given time in any given scenario.

[0080] Based on this, the benefits of an energy storage system within a scheduling cycle It can be represented as:

[0081] In the formula, For time difference.

[0082] Furthermore, to align with the objective optimization model, negative returns need to be defined as losses. In this case, when predicted electricity prices cause energy storage to charge (buy high) during periods of high electricity prices or discharge (sell low) during periods of low electricity prices, the returns... It may decrease or even become negative, leading to losses. Increase, loss The expression is: .

[0083] Based on this, risk threshold It can be represented as: ; In the formula, This risk threshold represents the confidence level. The confirmation method can be understood as: at this confidence level The minimum value below This means that the maximum loss of the system will not exceed this threshold. (with electricity price) (Same numerical value).

[0084] Furthermore, this embodiment can determine the negative return risk threshold based on the risk threshold to further characterize the potential depth of loss under extremely unfavorable scenarios.

[0085] Specifically, regarding losses exceeding [amount] due to drastic fluctuations in electricity prices. By modeling the tail risk, a negative return risk threshold is obtained, which represents the risk level at which losses exceed a certain threshold. Under these conditions, the expected value of the loss is... In the formula, The threshold for negative return risk. For the profit / loss function With Scene The function value between, This is the expectation operator.

[0086] This embodiment can solve the game problem between "pursuing high returns" and "avoiding high risks" by constructing physical constraints based on the energy storage system and constructing the objective function of the scheduling planning model based on the negative return risk threshold.

[0087] Please refer to Figure 5 , Figure 5 A flowchart of step 500 provided in an embodiment of the present invention is shown. In this embodiment, step 500 includes steps 501 to 503.

[0088] Step 501: Obtain the physical constraints of the energy storage system.

[0089] In this embodiment, the physical constraints include at least the continuity constraints of the energy storage state of charge, the upper and lower limits of capacity, the upper and lower limits of charge and discharge power, and the mutual exclusion constraints of charge and discharge.

[0090] Step 502: Convert the negative return risk threshold into a linear constraint.

[0091] Step 503: Determine the global optimal solution of the objective function based on physical constraints and linear constraints, and use the global optimal solution as the optimal charge and discharge power sequence.

[0092] Taking the continuity constraint of energy storage state of charge as an example, this t Continuity constraints of energy storage state of charge at any given time It can be represented as: In the formula, For charging efficiency, for t Charging power at any given time For discharge efficiency, for t Discharge power at time t, for t-1 Continuity constraints on the state of charge of the energy storage at any given time.

[0093] In addition, 0-1 binary variables can be introduced to determine the mutual exclusion constraint between charging and discharging, so as to prevent simultaneous charging and discharging.

[0094] The aforementioned physical constraints can be determined by the energy storage system, and the upper and lower limits of capacity and charging / discharging power will not be elaborated here.

[0095] In this embodiment, the objective function is expressed as follows: ; In the formula, The objective function value, For risk aversion coefficient, satisfying , The threshold for negative return risk. For the first s Benefits under a target scenario The total number of target scenarios, For the first s The probability of occurrence corresponding to each target scenario.

[0096] Among them, the s Benefits under each target scenario That is, it is characterized as: ; For the first s In each target scenario Electricity price at any given moment.

[0097] It should be noted that the risk aversion coefficient in this embodiment... The specific value can be adjusted by the user according to their own risk tolerance, thereby improving the flexibility of the energy storage strategy, so as to maximize expected returns while effectively controlling risks, and thus obtain more stable and reliable arbitrage returns.

[0098] Since the negative return risk threshold is nonlinear, this embodiment can first perform a linear expansion of the negative return risk threshold under the objective function before solving for the global optimal solution of the objective function.

[0099] Specifically, this embodiment may introduce auxiliary variables. This auxiliary variable Representing the s Exceeding the risk threshold in each target scenario The tail loss portion will exceed the risk threshold. The terms are transformed into a set of linear constraints, which can be expressed as: ,in Approximately value, For the first s The cost of the target scenario. Then the cost of the... s Benefits under each target scenario Expanding and substituting, we obtain the following set of linear constraints, for... All of them are: ;and .

[0100] Based on this, this embodiment can obtain the optimal charge and discharge power sequence more quickly based on the above-mentioned linear constraint set.

[0101] In addition, to eliminate the impact of accumulated prediction errors, please refer to... Figure 6 , Figure 6 The flowchart of another step of the charge-discharge optimization scheduling method provided in this embodiment of the invention is shown. The charge-discharge optimization scheduling method in this embodiment also includes steps 600 and 700.

[0102] Step 600: Obtain the real-time electricity price and the real-time state of charge of the energy storage system at preset time intervals.

[0103] Step 700: Update the scheduling planning model using real-time electricity price and real-time state of charge; and use the updated scheduling planning model as the target optimization model to return to the step of determining the optimal charging and discharging power sequence based on the target optimization model.

[0104] In this embodiment, real-time electricity prices can be used to optimize the electricity price probability parameters in step 200, and then optimize the scheduling planning model by adjusting the negative return risk threshold, especially the objective function of the scheduling planning model. Real-time state of charge is used to optimize the physical constraints under the scheduling planning model, especially the continuity constraint of the energy storage state of charge.

[0105] In one possible implementation, this embodiment can acquire the electricity price and the state of charge of the energy storage system every 15 minutes in real time. Then, based on the latest electricity price and the latest state of charge, the scheduling planning model is optimized to obtain a more accurate optimal charging and discharging power sequence to guide the energy storage system in charging and discharging.

[0106] The same idea applies as the previous embodiment; please refer to [the previous embodiment]. Figure 7 , Figure 7 A block diagram of a charge-discharge optimization scheduling device 10 provided in an embodiment of the present invention is shown. This charge-discharge optimization scheduling device 10 is applied to an energy storage system and includes: Data acquisition module 11 is used to acquire multi-source heterogeneous power data, which includes at least meteorological data, historical power output data of new energy power plants, historical electricity prices and transaction clearing power data; Electricity price probability parameter generation module 12 is used to construct an electricity price probability interval prediction model and generate electricity price probability parameters based on multi-source heterogeneous power data using the electricity price probability interval prediction model; The target electricity price scenario set construction module 13 is used to generate a target electricity price scenario set based on the electricity price probability parameter. The target electricity price scenario set includes multiple target electricity price scenarios, and the sum of the occurrence probabilities of each target electricity price scenario is 1. Negative return risk threshold construction module 14 is used to construct a negative return risk threshold based on the target electricity price scenario set; The decision building module 15 is used to build a scheduling planning model based on the negative return risk threshold, and optimize the model with the scheduling planning model as the objective. Based on the objective optimization model, the optimal charging and discharging power sequence is determined. The optimal charging and discharging power sequence includes the optimal charging and discharging state variables and the optimal power continuous variables.

[0107] In summary, this embodiment acquires multi-source heterogeneous power data and constructs a probability interval prediction model for electricity prices. Based on this multi-source heterogeneous power data, it generates electricity price probability parameters. Subsequently, it generates a target electricity price scenario set based on the electricity price probability parameters, constructs a negative return risk threshold based on the target electricity price scenario set, and then constructs a target optimization model based on the negative return risk threshold to generate an optimal charge and discharge power sequence that includes the optimal charge and discharge state variables and the optimal power continuous variables. This significantly improves the robustness and reliability of the energy storage arbitrage strategy.

[0108] The same idea applies as the previous embodiment; please refer to [the previous embodiment]. Figure 8 , Figure 8A block diagram of a server provided in an embodiment of the present invention is shown. The server 20 includes a memory 21, a processor 22, and a communication module 23. The memory 21, processor 22, and communication module 23 are electrically connected to each other directly or indirectly to realize data transmission or interaction. For example, these components can be electrically connected to each other through one or more communication buses or signal lines.

[0109] The memory 21 is used to store programs or data. The aforementioned memory may be, but is not limited to, random access memory (RAM), read-only memory (ROM), programmable read-only memory (PROM), erasable programmable read-only memory (EPROM), electrically erasable programmable read-only memory (EEPROM), etc.

[0110] The processor 22 is used to read / write data or programs stored in the memory and execute corresponding functions, namely, to acquire multi-source heterogeneous power data and construct a power price probability interval prediction model to generate power price probability parameters based on the multi-source heterogeneous power data, then generate a target power price scenario set based on the power price probability parameters, construct a negative return risk threshold based on the target power price scenario set, and then construct a target optimization model based on the negative return risk threshold to generate an optimal charging and discharging power sequence including the optimal charging and discharging state variables and the optimal power continuous variables.

[0111] The communication module 23 is used to establish a communication connection between the server and other communication terminals through the network, and to send and receive data through the network.

[0112] It should be understood that, Figure 8 The structure shown is only a schematic diagram of the server structure; the server may also include components such as... Figure 8 The more or fewer components shown, or having the same Figure 8 The different configurations shown. Figure 8 The components shown can be implemented using hardware, software, or a combination thereof.

[0113] In the several embodiments provided in this application, it should be understood that the disclosed apparatus and methods can also be implemented in other ways. The apparatus embodiments described above are merely illustrative; for example, the flowcharts and block diagrams in the accompanying drawings illustrate the architecture, functionality, and operation of possible implementations of apparatus, methods, and computer program products according to various embodiments of the present invention. In this regard, each block in a flowchart or block diagram may represent a module, segment, or portion of code containing one or more executable instructions for implementing a specified logical function. It should also be noted that in some alternative implementations, the functions marked in the blocks may occur in a different order than those marked in the drawings. For example, two consecutive blocks may actually be executed substantially in parallel, and they may sometimes be executed in reverse order, depending on the functions involved. It should also be noted that each block in a block diagram and / or flowchart, and combinations of blocks in block diagrams and / or flowcharts, can be implemented using a dedicated hardware-based system that performs the specified function or action, or using a combination of dedicated hardware and computer instructions.

[0114] In addition, the functional modules in the various embodiments of the present invention can be integrated together to form an independent part, or each module can exist independently, or two or more modules can be integrated to form an independent part.

[0115] If the aforementioned functions are implemented as software functional modules and sold or used as independent products, they can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of this invention, or the part that contributes to the prior art, or a portion of the technical solution, can be embodied in the form of a software product. This computer software product is stored in a storage medium and includes several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute all or part of the steps of the methods described in the various embodiments of this invention. The aforementioned storage medium includes various media capable of storing program code, such as USB flash drives, portable hard drives, read-only memory (ROM), random access memory (RAM), magnetic disks, or optical disks.

[0116] The above description is merely a preferred embodiment of the present invention and is not intended to limit the invention. Various modifications and variations can be made to the present invention by those skilled in the art. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the scope of protection of the present invention.

Claims

1. A charging and discharging optimization scheduling method, applied to an energy storage system, characterized in that, include: Acquire multi-source heterogeneous power data, which includes at least meteorological data, historical power output data of new energy power plants, historical electricity prices and transaction clearing power data; A probability interval prediction model for electricity prices is constructed, and the electricity price probability parameters are generated based on multi-source heterogeneous power data using the electricity price probability interval prediction model. A target electricity price scenario set is generated based on the electricity price probability parameters, wherein the target electricity price scenario set includes multiple target electricity price scenarios, and the sum of the occurrence probabilities of each target electricity price scenario is 1; Construct a negative return risk threshold based on the target electricity price scenario set; A scheduling planning model is constructed based on the negative return risk threshold, and the scheduling planning model is used as the target optimization model. The optimal charging and discharging power sequence is determined based on the target optimization model. The optimal charging and discharging power sequence includes the optimal charging and discharging state variables and the optimal power continuous variables.

2. The charging and discharging optimization scheduling method according to claim 1, characterized in that, The scheduling planning model includes an objective function, and the steps for determining the optimal charging and discharging power sequence based on the objective optimization model include: Obtain the physical constraints of the energy storage system, which include at least the continuity constraints of the energy storage state of charge, the upper and lower limits constraints of capacity, the upper and lower limits constraints of charge and discharge power, and the mutual exclusion constraints of charge and discharge. Convert the negative return risk threshold into a linear constraint; The global optimal solution of the objective function is determined based on the physical constraints and the linear constraints, and the global optimal solution is used as the optimal charge and discharge power sequence.

3. The charging and discharging optimization scheduling method according to claim 2, characterized in that, The charge / discharge optimization scheduling method further includes: The real-time electricity price and the real-time state of charge of the energy storage system are acquired at preset time intervals. The scheduling planning model is updated using the real-time electricity price and the real-time state of charge. Using the updated scheduling planning model as the target optimization model, the step of determining the optimal charging and discharging power sequence based on the target optimization model is executed.

4. The charge / discharge optimization scheduling method according to claim 1 or 2, characterized in that, The scheduling planning model includes an objective function, which is expressed as follows: ； In the formula, The objective function value, This is the risk aversion coefficient. The threshold for negative return risk. For the benefit in the s-th target scenario, The total number of target scenarios Let be the probability of occurrence corresponding to the s-th target scenario.

5. The charge / discharge optimization scheduling method according to claim 1 or 2, characterized in that, The electricity price probability interval prediction model includes a deep learning network. The steps of generating electricity price probability parameters based on the multi-source heterogeneous power data using the electricity price probability interval prediction model include: Candidate feature sets are extracted from the multi-source heterogeneous power data; wherein, the candidate feature sets include at least: time dimension features, historical electricity price lag features, and meteorological and load coupling features; The Pearson method is used to screen target features from the candidate feature set, and the electricity price time series is obtained based on the target features; the target features are used to characterize features that are linearly correlated with electricity prices. The deep learning network is used to generate multiple quantile electricity price predictions based on the electricity price time series. Electricity price probability parameters are generated based on the predicted quantile electricity prices.

6. The charge / discharge optimization scheduling method according to claim 5, characterized in that, The training function of the deep learning network includes a bouncing loss function, wherein, for a single quantile, the formula for calculating the bouncing loss function is expressed as: ； In the formula, For a moment Next Minimize the bouncing loss function value at each quantile. For a moment Next The true value at each quantile; For a moment Next Predicted quantiles at each quantile.

7. The charge / discharge optimization scheduling method according to claim 1 or 2, characterized in that, The steps for generating a target electricity price scenario set based on the electricity price probability parameters include: A Monte Carlo simulation is performed on the electricity price probability parameters to generate an initial electricity price scenario set; Calculate the Wasserstein distance between any two scenarios in the initial electricity price scenario set; The target electricity price scenario set is obtained from the initial electricity price scenario set using the synchronous back-substitution elimination method.

8. The charge / discharge optimization scheduling method according to claim 7, characterized in that, The steps for obtaining the target electricity price scenario set from the initial electricity price scenario set using the synchronous back-substitution elimination method include: In each iteration, the scenario with the smallest Wasserstein distance in the current initial electricity price scenario set is removed, and the probability corresponding to the removed scenario is accumulated to the target scenario. The target scenario is used to represent the retained scenario that is closest to the removed scenario. This process continues until the number of retained scenarios in the current iteration is equal to a preset value, and the target electricity price scenario set is constructed based on the retained scenarios corresponding to the current iteration.

9. The charge / discharge optimization scheduling method according to claim 1 or 2, characterized in that, The steps for constructing a negative return risk threshold based on the target electricity price scenario set include: The risk threshold is determined based on the revenue of the energy storage system within a scheduling cycle. Calculate the expected loss when the risk threshold is exceeded, and use the expected loss as the negative return risk threshold. The formula for calculating the negative return risk threshold is as follows: ； In the formula, The threshold for negative return risk. As a risk threshold, For confidence level, For the profit / loss function Uncertainty in actual electricity prices The function value between, This is the expectation operator.

10. A charge / discharge optimization scheduling device, applied to an energy storage system, characterized in that, include: The data acquisition module is used to acquire multi-source heterogeneous power data, which includes at least meteorological data, historical power output data of new energy power plants, historical electricity prices and transaction clearing power data; The electricity price probability parameter generation module is used to construct an electricity price probability interval prediction model and generate electricity price probability parameters based on multi-source heterogeneous power data using the electricity price probability interval prediction model. The target electricity price scenario set construction module is used to generate a target electricity price scenario set based on the electricity price probability parameter, wherein the target electricity price scenario set includes multiple target electricity price scenarios, and the sum of the occurrence probabilities of each target electricity price scenario is 1; A negative return risk threshold construction module is used to construct a negative return risk threshold based on the target electricity price scenario set. The decision-making construction module is used to construct a scheduling planning model based on the negative return risk threshold, and to optimize the scheduling planning model as the target model. Based on the target optimization model, the optimal charging and discharging power sequence is determined. The optimal charging and discharging power sequence includes the optimal charging and discharging state variable and the optimal power continuous variable.

Images