Independent energy storage station charge-discharge control method and system based on spot transaction similar day

Abstract

The application discloses an independent energy storage station charge-discharge control method and system based on spot transaction similar days, obtains power boundary, meteorological and energy storage change data sets, calculates energy storage sensitivity coefficients through the energy storage change data sets, and combines the annual growth rate of new energy to allocate time-varying characteristic weights to the power boundary data to screen preliminary similar days; the independent influence of the energy storage scale on the price is isolated through causal intervention processing, a dynamic correction coefficient is obtained through time-varying exploration noise online optimization in combination with a meteorological deviation factor, a charge-discharge two-side price prediction sequence is generated by correcting a historical price sequence, an independent energy storage power station spot transaction charge-discharge bidding strategy is formulated accordingly, and the independent energy storage power station executes charge-discharge operation based on the charge-discharge bidding strategy. The application can improve the accuracy of the independent energy storage power station spot transaction bidding and the precision of the physical charge-discharge control.

Description

Technical Field

[0001] This invention relates to the field of data processing technology, and in particular to a charging and discharging control method and system for independent energy storage stations based on similar spot trading days. Background Technology

[0002] With the rapid growth of new energy installed capacity, independent energy storage power stations are playing an increasingly important role in the power system. In the electricity spot market, independent energy storage power stations profit through arbitrage strategies of buying low and selling high, which relies on accurate prediction of future electricity price curves.

[0003] Currently, the common method for price forecasting is the similarity day screening method. This involves acquiring historical and forecast power system boundary data (such as grid load and generation data), calculating Euclidean distance and cosine similarity, and selecting several days from historical data that most closely match the boundary conditions of the forecast day as similar days. Then, the historical price series corresponding to these similar days are directly weighted and averaged or pattern matched to generate the price curve for the forecast day. Finally, a pricing strategy for energy storage charging and discharging is formulated based on this predicted price. However, existing similarity day screening methods have several drawbacks: First, the screening weights are usually static or empirically set, failing to adapt to the rapidly changing time-varying characteristics of new energy penetration and energy storage capacity. Second, they treat energy storage as a pure price taker, ignoring how its installed capacity growth alters the market supply and demand structure, thus having an independent and time-varying impact on the clearing price. Structural price discrepancies caused by changes in energy storage scale (such as the concentrated commissioning of large-scale regional energy storage power stations) cannot be effectively identified and corrected under the traditional similarity day framework, resulting in systematic biases in forecast results and affecting the trading decisions and economic benefits of energy storage power stations.

[0004] Therefore, the existing technology has low accuracy in screening similar days for price impacts, resulting in large prediction deviations in spot trading prices of independent energy storage power stations. This makes it difficult to improve the accuracy of spot trading quotations for independent energy storage power stations and to optimize the control of the physical charging and discharging process of independent energy storage power stations. Summary of the Invention

[0005] This invention provides a method and system for controlling the charging and discharging of an independent energy storage station based on similar spot trading days, which can achieve optimized control of the physical charging and discharging process of an independent energy storage station.

[0006] An embodiment of the present invention provides a charging and discharging control method for an independent energy storage station based on similar spot trading dates, comprising the following steps: Obtain the power boundary dataset, meteorological dataset, and energy storage change dataset required for spot trading of independent energy storage power stations; The energy storage sensitivity coefficient is calculated based on the energy storage change dataset, and the power boundary dataset is processed by time-varying feature weight allocation to calculate the similarity between the current trading day and historical days and to obtain a preliminary set of similar days; the weight coefficient of the time-varying feature weight allocation is determined by the energy storage sensitivity coefficient and the annual growth rate of new energy. Causal intervention processing is performed on the historical spot price series corresponding to the preliminary similar day set to generate a causal intervention vector for isolating the impact of changes in energy storage scale on spot prices; The historical spot price series corresponding to the causal intervention vector, the meteorological deviation factor, and the preliminary similar day set are subjected to online optimization processing based on time-varying exploration noise to obtain the dynamic correction coefficients for each historical day; The historical spot price series is corrected using the dynamic correction coefficient to obtain a corrected price reference series. Based on the corrected price reference series, a user-side charging price prediction series that characterizes charging costs and a generator-side discharging price prediction series that characterizes discharging revenue are calculated. Based on the user-side charging price prediction sequence and the generator-side discharging price prediction sequence, a charging and discharging pricing strategy for independent energy storage power stations in the spot market is generated. Based on the charging and discharging pricing strategy, the independent energy storage power station is controlled to perform charging and discharging operations.

[0007] Another embodiment of the present invention provides a charging and discharging control system for an independent energy storage station based on similar spot trading days, comprising: The acquisition module is used to acquire the power boundary dataset, meteorological dataset, and energy storage change dataset required for spot trading of independent energy storage power stations; The calculation module is used to calculate the energy storage sensitivity coefficient based on the energy storage change dataset, and to perform time-varying feature weight allocation processing on the power boundary dataset, calculate the similarity between the current trading day and historical days, and filter to obtain a preliminary set of similar days; the weight coefficient of the time-varying feature weight allocation processing is determined by the energy storage sensitivity coefficient and the annual growth rate of new energy. The vector generation module is used to perform causal intervention processing on the historical spot price series corresponding to the preliminary similar day set, and generate a causal intervention vector to isolate the impact of changes in energy storage scale on spot prices; The optimization module is used to perform online optimization processing based on time-varying exploratory noise on the historical spot price series corresponding to the causal intervention vector, the meteorological deviation factor, and the preliminary similar day set to obtain the dynamic correction coefficient for each historical day. The correction module is used to correct the historical spot price series using the dynamic correction coefficient to obtain a corrected price reference series, and based on the corrected price reference series, to calculate a user-side charging price prediction series that characterizes charging costs and a generator-side discharging price prediction series that characterizes discharging revenue. The strategy generation module is used to generate a charging and discharging pricing strategy for independent energy storage power stations in the spot market based on the user-side charging price prediction sequence and the generator-side discharging price prediction sequence. The control module is used to control the independent energy storage power station to perform charging and discharging operations based on the charging and discharging pricing strategy.

[0008] Compared with the prior art, the embodiments of the present invention have the following beneficial effects: First, acquire power boundary datasets, meteorological datasets, and energy storage change datasets. Calculate the energy storage sensitivity coefficient using the energy storage change dataset. Combine this with the annual growth rate of new energy sources to perform time-varying feature weighting on the power boundary dataset. Calculate the similarity between the current trading day and historical days to filter a preliminary set of similar days. Then, use causal intervention processing to isolate the independent impact of energy storage scale changes on spot prices, generating a causal intervention vector. Input this vector, a meteorological deviation factor, and the historical spot price series of the preliminary similar days into an online optimization process based on time-varying exploration noise to obtain dynamic correction coefficients. Use these coefficients to correct the historical spot price series to obtain a corrected price reference series, thereby generating user-side charging price and generator-side discharging price prediction series. Finally, formulate charging and discharging pricing strategies based on these predictions. Specifically, the time-varying feature weighting is determined by the energy storage sensitivity coefficient and the annual growth rate of new energy sources, adapting to the rapidly changing time-varying characteristics of new energy penetration and energy storage capacity, thus improving the accuracy of similar day selection. Causal intervention processing effectively identifies structural price deviations caused by changes in energy storage scale. Online optimization based on time-varying exploration noise combined with the meteorological deviation factor accurately corrects the price series. Separate price predictions for charging and discharging can specifically match trading demands. The embodiments of this invention can improve the accuracy of similar-day screening, reduce the prediction deviation of spot trading prices for independent energy storage power stations, and improve the accuracy of spot trading quotations, thereby achieving optimized control of the physical charging and discharging process of independent energy storage power stations. As can be seen from the above analysis, the embodiments of this invention improve the accuracy of similar-day screening and reduce the prediction deviation of spot trading prices for independent energy storage power stations by dynamically allocating feature weights based on energy storage sensitivity and using causal intervention vectors to isolate the impact of changes in energy storage scale, thereby improving the accuracy of spot trading quotations and achieving precise control of the charging and discharging of independent energy storage power stations. Attached Figure Description

[0009] Figure 1 This is a flowchart illustrating a charging and discharging control method for an independent energy storage station based on similar spot trading days, provided by an embodiment of the present invention. Figure 2 This is a schematic diagram of the structure of an independent energy storage station charging and discharging control system based on similar spot trading days, provided by an embodiment of the present invention. Detailed Implementation

[0010] 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. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0011] See Figure 1 This is a flowchart illustrating a method for controlling the charging and discharging of an independent energy storage station based on similar spot trading dates, according to an embodiment of the present invention. The method includes the following steps: S10: Obtain the power boundary dataset, meteorological dataset, and energy storage change dataset required for spot trading of independent energy storage power stations; S11, calculate the energy storage sensitivity coefficient based on the energy storage change dataset, and perform time-varying feature weight allocation processing on the power boundary dataset to calculate the similarity between the current trading day and historical days and filter to obtain a preliminary set of similar days; the weight coefficient of the time-varying feature weight allocation processing is determined by the energy storage sensitivity coefficient and the annual growth rate of new energy. S12, perform causal intervention processing on the historical spot price series corresponding to the preliminary similar day set to generate a causal intervention vector for isolating the impact of changes in energy storage scale on spot prices; S13, perform online optimization processing based on time-varying exploration noise on the historical spot price series corresponding to the causal intervention vector, the meteorological deviation factor and the preliminary similar day set to obtain the dynamic correction coefficient of each historical day; S14, the historical spot price series is corrected using the dynamic correction coefficient to obtain the corrected price reference series, and based on the corrected price reference series, a user-side charging price prediction series for characterizing charging costs and a generator-side discharging price prediction series for characterizing discharging revenue are calculated. S15, Based on the user-side charging price prediction sequence and the generator-side discharging price prediction sequence, generate a charging and discharging pricing strategy for independent energy storage power stations in the spot market; S16, Based on the charging and discharging pricing strategy, control the independent energy storage power station to perform charging and discharging operations.

[0012] Compared with the prior art, the embodiments of the present invention have the following beneficial effects: First, acquire power boundary datasets, meteorological datasets, and energy storage change datasets. Calculate the energy storage sensitivity coefficient using the energy storage change dataset. Combine this with the annual growth rate of new energy sources to perform time-varying feature weighting on the power boundary dataset. Calculate the similarity between the current trading day and historical days to filter a preliminary set of similar days. Then, use causal intervention processing to isolate the independent impact of energy storage scale changes on spot prices, generating a causal intervention vector. Input this vector, a meteorological deviation factor, and the historical spot price series of the preliminary similar days into an online optimization process based on time-varying exploration noise to obtain dynamic correction coefficients. Use these coefficients to correct the historical spot price series to obtain a corrected price reference series, thereby generating user-side charging price and generator-side discharging price prediction series. Finally, formulate charging and discharging pricing strategies based on these predictions. Specifically, the time-varying feature weighting is determined by the energy storage sensitivity coefficient and the annual growth rate of new energy sources, adapting to the rapidly changing time-varying characteristics of new energy penetration and energy storage capacity, thus improving the accuracy of similar day selection. Causal intervention processing effectively identifies structural price deviations caused by changes in energy storage scale. Online optimization based on time-varying exploration noise combined with the meteorological deviation factor accurately corrects the price series. Separate price predictions for charging and discharging can specifically match trading demands. The embodiments of this invention can improve the accuracy of similar-day screening, reduce the prediction deviation of spot trading prices for independent energy storage power stations, and improve the accuracy of spot trading quotations, thereby achieving optimized control of the physical charging and discharging process of independent energy storage power stations. As can be seen from the above analysis, the embodiments of this invention improve the accuracy of similar-day screening and reduce the prediction deviation of spot trading prices for independent energy storage power stations by dynamically allocating feature weights based on energy storage sensitivity and using causal intervention vectors to isolate the impact of changes in energy storage scale, thereby improving the accuracy of spot trading quotations and achieving precise control of the charging and discharging of independent energy storage power stations.

[0013] As an example, by connecting with authoritative data sources in the power system and energy storage industry management platforms, standardized data interfaces are used to accurately collect and time-series align three types of datasets, ensuring data timeliness and format compatibility. The power boundary dataset is obtained through the real-time data sharing platform of the power dispatch center. Its core includes three types of core forecast data closely related to spot trading: load forecast data (one data point every 15 minutes within the 24 hours of the current day), thermal power generation forecast data, and new energy (wind power, photovoltaic) generation forecast data, forming a 96-point time-series dataset consistent with subsequent processing dimensions. For example, the regional load forecast sequence for the current trading day is obtained from the provincial power dispatch center, while corresponding time-series thermal power and new energy power generation forecast data are aggregated from the operation and maintenance systems of various thermal power plants and new energy power plants. The meteorological dataset is collected through a refined meteorological forecast platform issued by the meteorological department, focusing on obtaining 15-minute meteorological data within the candidate interval of the current trading day and historically similar days, including temperature, solar irradiance, wind speed, etc., which are strongly correlated with changes in new energy output and load. For example, relevant indicators include obtaining temperature forecast data for the current trading day at 96:00 AM and historical meteorological records for the past 6 months at 96:00 AM daily through the API interface of the local meteorological service center; the energy storage change dataset is collected from the energy storage power station registration system of the regional energy regulatory department and the operation and maintenance management platform of the energy storage power station itself, including data such as the new commissioning timestamps of all energy storage power stations in the region, the monthly growth rate of energy storage capacity, and the growth rate of new energy installed capacity. For example, the commissioning time information of new energy storage power stations in the past 12 months is extracted from the energy storage operation and maintenance management platform, and the statistical data of regional new energy installed capacity growth during the same period is obtained from the energy regulatory department. At the same time, it is ensured that the time granularity of all collected data is uniformly 15 minutes, the time series dimension matches the 96-point sequence requirement, and the fields of the three types of datasets are standardized through data format conversion to avoid deviations in subsequent data processing due to format differences.

[0014] As one example, the process involves calculating the energy storage sensitivity coefficient based on the energy storage change dataset, performing time-varying feature weight allocation on the power boundary dataset, calculating the similarity between the current trading day and historical days, and filtering to obtain a preliminary set of similar days, including: Extract the energy storage capacity growth rate and the new energy installed capacity growth rate from the energy storage change dataset; The ratio of the energy storage capacity growth rate to the new energy installed capacity growth rate is calculated to obtain the energy storage sensitivity coefficient; The load forecast, thermal power generation forecast, and new energy generation forecast contained in the power boundary dataset are obtained. The time-varying feature weight allocation processing is performed on the load forecast, thermal power generation forecast, and new energy generation forecast using the energy storage sensitivity coefficient and the annual growth rate of new energy, so as to obtain the weight coefficients corresponding to each of the three. Based on the respective weight coefficients of the three, the load forecast, the thermal power generation forecast, and the new energy power generation forecast are weighted and calculated to obtain the weighted similarity between the current trading day and historical days; The number of screening days is determined based on the historical price dispersion, and a corresponding number of historical days are selected from high to low based on the weighted similarity to generate the preliminary set of similar days.

[0015] In this embodiment, the energy storage capacity growth rate and the new energy installed capacity growth rate are first extracted from the energy storage change dataset. The ratio of these two is used to obtain the energy storage sensitivity coefficient, which reflects the sensitivity of energy storage growth relative to new energy growth. Next, the load forecast, thermal power generation forecast, and new energy power generation forecast included in the power boundary dataset are obtained. The energy storage sensitivity coefficient and the annual growth rate of new energy are used to perform time-varying feature weighting on these three types of features to obtain their respective weight coefficients, thereby dynamically adjusting the importance of different features in similarity calculation. Then, based on the weight coefficients, the features are weighted to obtain the weighted similarity between the current trading day and historical days. The number of screening days is determined by combining the historical price dispersion, and historical days are selected from high to low to generate a preliminary set of similar days. Therefore, this embodiment, by using the energy storage sensitivity coefficient to dynamically allocate feature weights, can adapt to the time-varying characteristics of new energy penetration rate and energy storage scale, and improve the matching degree between the screened similar days and the current trading day.

[0016] As an example, the energy storage capacity growth rate and the new energy installed capacity growth rate are first extracted from the energy storage change dataset. A unified statistical period is defined first; to ensure the data accurately reflects the current market size trend, the previous full calendar month is typically selected as the statistical period. For the energy storage capacity growth rate, the commissioning records and corresponding capacity data of all newly added energy storage power stations within the statistical period are selected from the energy storage change dataset. This is then aggregated to obtain the total newly added energy storage capacity within the period, and the ratio is calculated with the total regional energy storage capacity at the beginning of the statistical period to obtain the energy storage capacity growth rate. For the new energy installed capacity growth rate, the same statistical period and calculation logic as the energy storage capacity growth rate are used. Records of newly added installed capacity for various new energy sources such as wind power and photovoltaics within the same period are extracted from the energy storage change dataset, and the total newly added new energy installed capacity is aggregated. This ratio is then calculated with the total regional new energy installed capacity at the beginning of the statistical period to obtain the new energy installed capacity growth rate.

[0017] Secondly, the energy storage sensitivity coefficient is calculated. This coefficient quantifies the unique impact of changes in energy storage scale on spot prices and is a core improvement in this step. The algorithm is an improved ratio calculation model, and the specific implementation process is explained below with reference to the formula: The calculation formula used is... S represents the energy storage sensitivity coefficient, which characterizes the relative strength of energy storage capacity growth and new energy installation growth, thereby quantifying the degree of independent impact of changes in energy storage scale on spot prices. r1 represents the energy storage capacity growth rate extracted above, reflecting the relative growth level of energy storage scale within the statistical period. r2 represents the new energy installation growth rate extracted above, reflecting the relative growth level of new energy scale within the same period. If r2 is zero, to avoid abnormal division operations, r2 is set to a very small positive number close to zero to ensure that the calculation process can proceed normally. If r1 is zero, S is directly set to zero, which indicates that there is no change in energy storage scale within the statistical period, and its unique impact on spot prices can be ignored, so it does not need to be considered in subsequent weight allocation.

[0018] Next, load forecasts, thermal power generation forecasts, and renewable energy generation forecasts from the power boundary dataset are obtained, and data preprocessing and time-varying feature weight allocation are completed. First, three types of core forecast data are extracted from the power boundary dataset. Each type consists of a 96-point time series, corresponding to one data collection point every 15 minutes for 24 hours per day. To eliminate interference from differences in the units of measurement between different data points in subsequent calculations, standardization is performed on the three types of time series data using the conventional Z-score standardization algorithm. By calculating the mean and standard deviation of each type of data, the original data is transformed into standardized data with a mean of 0 and a standard deviation of 1, completing the data preprocessing. Subsequently, time-varying feature weight allocation is performed. The core of this process is to dynamically adjust the weights of the three types of forecast data based on the energy storage sensitivity coefficient and the annual growth rate of renewable energy. The algorithm formula is as follows: , , , This represents the weighting coefficient corresponding to the load forecast. This represents the weighting coefficient corresponding to the thermal power generation forecast. The three weighting coefficients, α, β, γ, and δ, represent the weighting coefficients corresponding to the new energy power generation forecast. The sum of these three weighting coefficients is 1 to ensure the rationality of the weighting allocation. S represents the energy storage sensitivity coefficient calculated above. r3 represents the pre-obtained regional annual growth rate of new energy, used to reflect the impact of the long-term development trend of new energy on market characteristics. α, β, γ, and δ are preset proportional coefficients, satisfying α+β=1 and γ+δ=1. These proportional coefficients can be determined through calibration using historical transaction data to ensure that the weighting allocation conforms to actual market rules. Specifically, by substituting the specific values ​​of S and r3, and combining them with the calibrated proportional coefficients, the following calculations are performed: , , This enables dynamic allocation of weights for the three types of forecast data, overcoming the shortcomings of traditional methods that use fixed weights and cannot adapt to the rapid changes in the penetration rate of new energy and the scale of energy storage.

[0019] Next, based on the weight coefficients corresponding to the three types of forecast data, the weighted similarity between the current trading day and historical days is calculated. First, the standardized load forecast, thermal power generation forecast, and renewable energy generation forecast data for the current trading day are constructed as feature vectors X1, X2, and X3, respectively. Simultaneously, the standardized forecast data of the three types for each historical day within the initial similarity day selection range are constructed as historical feature vectors X1h, X2h, and X3h. Then, the weighted similarity is calculated using the following formula. Sim represents the weighted similarity between the current trading day and a single historical day, with a value range of [0,1]. The closer the value is to 1, the higher the matching degree of the market boundary features between the two. , , Here, X1, X2, and X3 are the weight coefficients of the three types of predicted data obtained above; X1, X2, and X3 are the feature vectors of the three types of predicted data for the current trading day, and X1h, X2h, and X3h are the feature vectors of the three types of predicted data for historical days; "·" represents the dot product operation of vectors, and the superscript T of X1h, X2h, and X3h represents the transpose of the corresponding historical feature vectors X1h, X2h, and X3h, respectively; |·| represents the L2 norm of the vector, used to normalize the data. Specifically, the dot product operation is performed point-by-point on the 96-point time series to ensure that the similarity calculation accurately reflects the feature matching degree of each time period within the day. Then, the influence of vector dimension differences is eliminated through denominator normalization, finally obtaining the weighted similarity between the current trading day and each historical day.

[0020] Finally, the number of days to be screened is determined based on historical price dispersion, and a corresponding number of historical days are selected from high to low based on weighted similarity to generate a preliminary set of similar days. First, the statistical range of historical prices is determined, typically selecting all historical trading days within the past 6 months as the statistical sample, and extracting the spot price series data for these historical days. The coefficient of variation is used to calculate historical price dispersion. By calculating the ratio of the standard deviation to the mean of the spot price series for all historical days within the statistical sample, the volatility of historical prices is quantified; the larger the ratio, the more volatile the historical prices. The number of days to be screened, K, is adaptively determined based on historical price dispersion: if the dispersion is large, it indicates volatile historical prices, and the K value needs to be increased to ensure that the selected similar days are sufficiently representative; if the dispersion is small, it indicates stable historical price fluctuations, and the K value can be decreased to improve screening efficiency. After determining the K value, all historical days are sorted from high to low according to their weighted similarity to the current trading day, and the top K historical days are selected to form the preliminary set of similar days. This screening process ensures that the preliminary set of similar days accurately matches the market boundary characteristics and price volatility characteristics of the current trading day.

[0021] As one example, the causal intervention processing of the historical spot price series corresponding to the preliminary similar day set to generate a causal intervention vector for isolating the impact of changes in energy storage scale on spot prices includes: Based on the energy storage change dataset, determine the energy storage scale status for the current trading day and the energy storage scale status for historical trading days; Calculate the energy storage capacity difference between the current trading day's energy storage capacity status and the historical trading day's energy storage capacity status; The causal intervention process is applied to the historical spot price series, and the impact of changes in energy storage scale on prices is assessed based on the differences in energy storage scale to obtain the price adjustment component. Based on the price adjustment component, a causal intervention vector is generated to isolate the impact of changes in energy storage scale on spot prices.

[0022] In this embodiment, the energy storage scale status of the current and historical trading days is first determined based on the energy storage change dataset, which clarifies the spatiotemporal differences in energy storage scale. Next, the difference in energy storage scale between the two is calculated, providing a quantitative basis for assessing the impact of energy storage scale on prices. Then, causal intervention processing is performed on the historical spot price series, and the price impact is assessed based on the energy storage scale difference to obtain a price adjustment component, effectively isolating the coupled impact of energy storage scale changes on prices from other factors. Finally, a causal intervention vector is generated based on the price adjustment component, which can accurately isolate the independent impact of energy storage scale changes on spot prices. Therefore, this embodiment, through the quantification of energy storage scale differences and causal intervention processing, achieves effective isolation of the impact of energy storage scale changes on prices, solving the problem of the inability to identify structural price deviations in traditional methods.

[0023] Specifically, in this embodiment, firstly, the energy storage change dataset contains core data such as the monthly growth rate of energy storage capacity and the commissioning timestamps of newly added power stations. The core logic for determining the energy storage scale status is to accurately count the total capacity of all commissioned energy storage power stations before the target trading day (current or historical trading day). For the current trading day, firstly, all records of newly added energy storage power stations with commissioning timestamps less than or equal to the current trading day are selected from the energy storage change dataset. The newly added energy storage capacity corresponding to each record is extracted and summarized to obtain the total newly added energy storage capacity that meets the conditions. Then, the initial total energy storage capacity of the region at the start of the statistical period is retrieved, and the total newly added energy storage capacity is added to the initial total energy storage capacity to obtain the energy storage scale status of the current trading day. For each historical trading day in the preliminary similar day set, the statistical logic is completely consistent with that of the current trading day, that is, the newly added energy storage capacity with commissioning timestamps less than or equal to that historical trading day is selected and summarized, and then added to the initial total energy storage capacity to obtain the energy storage scale status of the historical trading day. Throughout the process, it is necessary to ensure that the statistical caliber of the initial total energy storage capacity is consistent and that the criteria for judging the commissioning timestamp are uniform, so as to avoid distortion of the energy storage scale status due to differences in statistical benchmarks.

[0024] Secondly, to avoid assessment bias caused by the absolute difference being influenced by the scale of basic energy storage, a relative difference is used to quantify the degree of scale change between the two. The calculation uses the energy storage scale status of historical trading days as a benchmark. The absolute difference value is obtained by subtracting the energy storage scale status of historical trading days from the energy storage scale status of the current trading day. This absolute difference value is then compared with the energy storage scale status of historical trading days to obtain the energy storage scale difference. Special handling is required for extreme data scenarios: if the energy storage scale status of a historical trading day is zero, meaning no energy storage power station was put into operation on that historical day, the energy storage scale difference is set to a maximum value, indicating a significant difference in energy storage scale. If the current energy storage scale status is equal to that of a historical trading day, the energy storage scale difference is zero, indicating no change in energy storage scale, and no price adjustment is needed for this factor. This relative difference calculation directly reflects the relative degree of change in energy storage scale, providing a precise quantitative basis for subsequent price assessment.

[0025] The core purpose of causal intervention is to isolate the coupled impact of changes in energy storage scale with other factors (such as load and weather) on spot prices, ensuring that price adjustments only reflect the independent effect of changes in energy storage scale. The causal intervention algorithm consists of two stages: parameter calibration and counterfactual inference. The parameter calibration stage uses a conventional least squares regression algorithm. Specifically, it uses the difference in energy storage scale as the core independent variable, introduces load forecasts and meteorological data (temperature, irradiance, wind speed) as control variables, and uses historical actual spot prices as the dependent variable to construct a regression model and calibrate the model parameters, obtaining the influence coefficient of energy storage scale differences on prices. The algorithm formula for the counterfactual inference stage is as follows: , This indicates the price adjustment component, i.e., the spot price deviation caused solely by changes in energy storage scale; This represents the impact coefficient of the energy storage scale difference on the price obtained after parameter calibration, and is used to quantify the price change corresponding to the relative change in unit energy storage scale. This represents the difference in energy storage capacity calculated above. In practical implementation, the calibrated value will be... and Substituting into the formula, the price adjustment component is calculated. This component accurately isolates the interference from other factors such as load and weather, and only corresponds to the independent impact of changes in energy storage scale on spot prices, overcoming the shortcomings of traditional methods that cannot distinguish the coupled effects of multiple factors.

[0026] Since the historical spot price series is a 96-point time series (corresponding to 24 hours per day, with one data point every 15 minutes), to achieve accurate price correction at each time point, the causal intervention vector needs to adopt a 96-dimensional vector structure that is completely consistent with the price series dimension. During construction, the price adjustment components calculated above are used... This is done by matching each data point in the historical spot price series along the time dimension to form a causal intervention vector. If the impact of energy storage capacity on prices differs between peak and off-peak periods in actual scenarios, the impact coefficient can be calibrated through time-sharing adjustments. Get the corresponding time period Then, construct time-differentiated causal intervention vectors; if the influence coefficients do not show significant time-specific differences, then a uniform approach is adopted. Constructing vectors. The generated causal intervention vectors can characterize the independent impact of changes in energy storage scale on historical prices point by point, providing a precise basis for subsequent online optimization processing based on time-varying exploration noise, and ensuring that the subsequent correction process only compensates for price deviations caused by changes in energy storage scale.

[0027] As one example, the online optimization process based on time-varying exploratory noise is performed on the historical spot price series corresponding to the causal intervention vector, the meteorological deviation factor, and the preliminary similar day set to obtain the dynamic correction coefficient for each historical day; including: The causal intervention vector, the meteorological deviation factor, and the historical spot price series are input into the online optimization process based on time-varying exploration noise; In the online optimization process based on time-varying exploration noise, an optimization objective including a correction bias penalty term is constructed; The time-varying exploration noise is generated, and the noise amplitude of the time-varying exploration noise is dynamically adjusted according to the energy storage sensitivity coefficient; The time-varying exploration noise is used to perturb the current correction coefficients, and the corrected correction coefficients are iteratively updated based on the optimization objective; When the optimization objective satisfies the convergence condition, the iteratively updated correction coefficient is output as the dynamic correction coefficient for each historical day.

[0028] In this embodiment, the causal intervention vector, meteorological deviation factor, and historical spot price series are first input into the optimization process. Next, an optimization objective including a correction deviation penalty term is constructed within the optimization process. Subsequently, time-varying exploration noise is generated, and its amplitude is dynamically adjusted based on the energy storage sensitivity coefficient to adapt to exploration needs under different energy storage growth environments. Then, the adjusted time-varying exploration noise is used to perturb the current correction coefficient, and the corrected correction coefficient is iteratively updated based on the optimization objective. When the convergence condition is met, the iteratively updated correction coefficient is output as the dynamic correction coefficient for each historical day. Therefore, this embodiment, through online optimization processing that dynamically adjusts the noise amplitude based on the energy storage sensitivity coefficient, can adaptively search for the optimal correction coefficient in complex market environments, effectively correcting price deviations caused by non-energy storage factors and random meteorological factors, and improving the accuracy and adaptability of the correction coefficient.

[0029] In this embodiment, the input causal intervention vector is a 96-dimensional vector generated by the causal intervention processing above, used to characterize the independent impact of changes in energy storage scale on historical prices. The meteorological deviation factor needs to be matched first, extracting meteorological data (temperature, irradiance, wind speed) corresponding to the preliminary similar days from the meteorological dataset, and calculating the deviation value between it and the meteorological data corresponding to the current trading day, forming a 96-dimensional meteorological deviation factor vector consistent with the price series dimension, quantifying the potential impact of meteorological differences on prices. The historical spot price series is a 96-point time series sequence of each historical day in the preliminary similar day set (corresponding to one data point every 15 minutes for 24 hours each day). To ensure that the data adapts to the input requirements of the optimization model, standardized preprocessing is uniformly performed on the three types of input data. The conventional Z-score standardization algorithm is used to convert the original data into standardized data with a mean of 0 and a standard deviation of 1 by calculating the mean and standard deviation of each type of data, avoiding optimization bias caused by differences in dimensions, and ensuring that the input data completely matches the input dimension of the subsequent optimization model.

[0030] Secondly, conventional optimization objectives only focus on correction accuracy and do not consider the stability of correction coefficients. This embodiment balances correction accuracy and coefficient stability by introducing a correction deviation penalty term. The optimization objective formula is as follows: J represents the objective function value, minimizing J achieves the optimal correction effect; Pc represents the reasonable market price representation value for the current trading day (constructed using market supply and demand fundamentals data for the same period, used as a reference benchmark for price correction); Ph represents the standardized value of the historical spot price series; V represents the standardized value of the causal intervention vector; k represents the dynamic correction coefficient to be optimized, which is the core parameter of the solution required in this step; M represents the standardized value of the weather deviation factor; λ represents the influence coefficient of the weather deviation factor on the price (obtained through offline calibration of historical data, a fixed value); μ represents the correction deviation penalty term coefficient, used to adjust the penalty intensity and avoid overfitting due to excessively large correction coefficients; ||k|| represents the L1 regularization term of the dynamic correction coefficient k, used to constrain the range of values ​​of the correction coefficient and improve the stability of the coefficient. First, the standardized Ph, V, and M, and the calibrated λ and μ are substituted to construct the complete objective function, clarifying the optimization direction as minimizing the deviation between the corrected price and the reference benchmark Pc, while controlling the rationality of the correction coefficient through the penalty term.

[0031] Next, this embodiment dynamically adjusts the magnitude of the energy storage sensitivity coefficient, and the algorithm formula is as follows: Nt represents the time-varying exploration noise in the t-th iteration; A represents the base noise amplitude (a fixed value obtained through historical optimization experience calibration); S represents the energy storage sensitivity coefficient calculated above, used to characterize the impact of changes in energy storage scale on price; η represents the noise adjustment coefficient, used to adjust the influence weight of S on the noise amplitude, satisfying 0 < η < 1; rand() represents the generated random number uniformly distributed in the interval [-1, 1]. In specific implementation, the current energy storage sensitivity coefficient S is first obtained, substituted into the formula to calculate the noise amplitude of this iteration, and then a random number is generated to obtain the time-varying exploration noise Nt. The specific adjustment logic is as follows: when S is large, it indicates that the energy storage scale has a significant impact on price, and the noise amplitude needs to be reduced to avoid excessive disturbance leading to optimization instability; when S is small, the energy storage impact is weak, and the noise amplitude can be increased to expand the exploration range, improve the global search capability of optimization, and ensure efficient exploration under different energy storage impact scenarios.

[0032] Then, the correction coefficient kt for the current iteration is obtained, and the time-varying exploration noise Nt is superimposed on kt to obtain the perturbed correction coefficient kt' = kt + Nt, completing the perturbed exploration. Subsequently, kt' is substituted into the optimization objective function J constructed above to calculate the current objective function value Jt'. The gradient of the objective function J with respect to kt' is calculated using a conventional gradient descent algorithm. The value of kt' is adjusted according to the gradient direction to obtain the updated correction coefficient k{t+1}. The adjustment logic is to move along the negative gradient direction, and the step size is determined through historical data calibration to ensure stable convergence of the iteration process. The above process of perturbation, objective function calculation, and gradient descent update is repeated to achieve iterative optimization of the correction coefficient in each iteration. Each iteration uses new time-varying exploration noise to improve the adaptability and global optimum of the optimization.

[0033] Finally, the convergence conditions are preset, including two criteria: first, the difference between the optimization objective function values ​​of two adjacent iterations is less than a preset threshold (this threshold is calibrated using historical optimization data to ensure the correction accuracy meets the standard); second, the number of iterations reaches a preset maximum number of iterations (to avoid iteration falling into an infinite loop). During the iteration process, it is checked in real time whether either convergence condition is met. If it is met, the iteration stops, and the correction coefficient k of the current iteration round is output as the final dynamic correction coefficient; if it is not met, the next round of perturbation and update continues. The output dynamic correction coefficient corresponds one-to-one with each historical day in the preliminary similar day set. Subsequently, this coefficient can be used to accurately correct the spot price series of each historical day, while ensuring that the correction coefficient adapts to the dynamic changes in energy storage scale and meteorological conditions.

[0034] As one example, the step of using the dynamic correction coefficient to correct the historical spot price series to obtain a corrected price reference series, and based on the corrected price reference series, calculating a user-side charging price prediction series to characterize charging costs and a generator-side discharging price prediction series to characterize discharging revenue, includes: The historical spot price series is corrected using the dynamic correction coefficient to obtain the corrected price reference series; Adaptive time-series fusion processing is performed on the corrected price reference sequence to extract fused time-series features containing information from multiple time scales; Based on the charging and discharging roles of independent energy storage power stations, the fused timing features are classified and mapped into user-side charging timing features and generator-side discharging timing features. The energy storage conversion efficiency of the independent energy storage power station is obtained, and the energy storage conversion efficiency is used to numerically calibrate the user-side charging timing characteristics and the generator-side discharging timing characteristics, respectively. Based on the numerically calibrated features, the user-side charging price prediction sequence and the generator-side discharging price prediction sequence are generated.

[0035] In this embodiment, the historical spot price series is first corrected using a dynamic correction coefficient to obtain a corrected price reference series. Next, adaptive time-series fusion processing is performed on the corrected price reference series to extract fused time-series features containing multi-time-scale information. Then, based on the charging and discharging roles of independent energy storage power stations, the fused time-series features are classified and mapped into user-side charging time-series features and generator-side discharging time-series features. Subsequently, the energy storage conversion efficiency of the independent energy storage power stations is obtained, and the energy storage conversion efficiency is used to numerically calibrate the two types of time-series features. Finally, user-side charging price prediction series and generator-side discharging price prediction series are generated based on the calibrated data. Therefore, this embodiment, through adaptive time-series fusion processing and numerical calibration based on energy storage conversion efficiency, can fully utilize the multi-scale information in the corrected price reference series to accurately calculate price prediction series reflecting actual charging costs and discharging revenues.

[0036] The specific workflow of this embodiment is as follows: The dynamic correction coefficients obtained above correspond one-to-one with the historical days in the preliminary similarity day set and perfectly match the 96-point time series dimension of the historical spot price series. The core logic of the correction calculation is to compensate for the deviations in historical prices caused by differences in energy storage scale and weather conditions through dynamic correction coefficients. In practice, the historical spot price series of each historical day is weighted and corrected point by point with the corresponding dynamic correction coefficient, while compensation terms for causal intervention vectors and weather deviation factors are superimposed to ensure that the corrected price is completely free from the interference of energy storage scale and weather conditions. Through this correction process, the obtained corrected price reference series can accurately reflect a reasonable price level that matches the market environment of the current trading day, providing a reliable basis for the calculation of subsequent price prediction series.

[0037] Secondly, conventional time-series feature extraction only uses a single time scale, which cannot fully capture short-term price fluctuations and long-term trends. This embodiment achieves comprehensive feature extraction through adaptive multi-scale fusion. The algorithm formula is as follows: F represents the fused time-series feature; ωi represents the adaptive weight coefficient of the i-th time-scale feature, satisfying Σωi=1, and its value is adaptively adjusted by the intensity of the time-series changes in the price series. The greater the intensity of change, the higher the weight of the short-term scale, and vice versa; Fi represents the time-series feature extracted at the i-th time scale. In this embodiment, three typical time scales are selected: short-term scale (corresponding to the fluctuation features of adjacent data points), medium-term scale (corresponding to the trend features at the hourly level), and long-term scale (corresponding to the periodic features at the intraday time period level). In specific implementation, conventional time-series feature extraction algorithms (such as wavelet transform) are first used to extract the time-series features F1, F2, and F3 at the three scales respectively; then, the adaptive weights ω1, ω2, and ω3 are determined by calculating the correlation between each scale feature and the price series. The higher the correlation, the greater the weight; finally, the weighted sum of each scale feature and its corresponding weight is obtained to obtain the fused time-series feature F that integrates information from multiple time scales, ensuring that the extracted features include both fine-grained price fluctuation information and macro trend information.

[0038] Next, the differences in charging and discharging roles of independent energy storage power stations are clarified. The core requirement for the user-side charging role is to accurately represent charging costs, corresponding to the characteristics of off-peak load periods in the price series. The core requirement for the generator-side discharging role is to accurately represent discharging revenue, corresponding to the characteristics of peak load periods in the price series. Based on this difference, feature mapping rules are constructed: features related to price fluctuations and load levels during off-peak periods are selected from the fused time-series features and mapped to user-side charging time-series features. These features should reflect the price stability and cost controllability during charging periods. Features related to price fluctuations and renewable energy output during peak periods are selected from the fused time-series features and mapped to generator-side discharging time-series features. These features should reflect the price peak and revenue potential during discharging periods. During the mapping process, it is necessary to ensure that there is no redundancy between the two types of features and that they fully cover the core requirements of the charging and discharging roles, providing targeted feature support for subsequent cost and revenue calculations.

[0039] Then, the energy storage conversion efficiency is obtained from the equipment parameter data of the energy storage power station. This efficiency includes two dimensions: charging conversion efficiency and discharging conversion efficiency, which respectively characterize the energy storage loss during charging and the energy release loss during discharging. The core logic of numerical calibration is to correct the efficiency loss deviation in the features. In specific operation, for the user-side charging time-series features, the charging conversion efficiency is used for calibration. That is, the efficiency value is used to correct the indicators related to charging power and energy consumption in the features, ensuring that the calibrated features can accurately reflect the effective energy cost in the actual charging process. For the generator-side discharging time-series features, the discharging conversion efficiency is used for calibration. The efficiency value is used to correct the indicators related to discharging power and energy output in the features, ensuring that the calibrated features can accurately reflect the effective energy gain in the actual discharging process. The calibration process must strictly distinguish between charging and discharging efficiencies to avoid confusion that could lead to calibration deviations.

[0040] Finally, the calibrated user-side charging time-series characteristics are input into a conventional time-series prediction model (such as linear regression or time-series smoothing algorithms). The model outputs a 96-point time-series sequence with the same dimensions as the corrected price reference sequence, which is the user-side charging price prediction sequence. This sequence represents the charging cost level for each time period within the day. Similarly, the calibrated generator-side discharging time-series characteristics are input into the same type of time-series prediction model, outputting a 96-point generator-side discharging price prediction sequence, representing the discharging revenue level for each time period within the day. During the generation process, it is necessary to ensure that the time dimension of the prediction sequence completely matches the time granularity of the actual spot transaction. At the same time, the stability and accuracy of the prediction sequence are ensured through model parameter calibration, providing direct cost and revenue quantification basis for the subsequent generation of charging and discharging pricing strategies.

[0041] As one example, the step of generating a charging and discharging pricing strategy for an independent energy storage power station in the spot market based on the user-side charging price prediction sequence and the generator-side discharging price prediction sequence includes: The charging cost is calculated based on the user-side charging price prediction sequence, and the discharge revenue is calculated based on the generator-side discharge price prediction sequence. Combining the charging cost and the discharging benefit, an optimization objective is constructed with the goal of maximizing transaction revenue; The operational constraints of the independent energy storage power station are incorporated into the optimization objective, and the strategy optimization process is executed to obtain the optimal charging and discharging power plan that satisfies the spot trading rules. Based on the optimal charge and discharge power plan, a charge and discharge pricing strategy for the independent energy storage power station in the spot market is generated.

[0042] In this embodiment, the charging cost is first calculated based on the user-side charging price prediction sequence, and the discharging revenue is calculated based on the generator-side discharging price prediction sequence. Next, combining the charging cost and discharging revenue, an optimization objective is constructed to maximize transaction revenue. Then, the operational constraints of the independent energy storage power station are incorporated into the optimization objective, and strategy optimization is performed to obtain the optimal charging and discharging power plan that satisfies the spot trading rules. Finally, a charging and discharging pricing strategy for the independent energy storage power station in the spot market is generated based on the optimal charging and discharging power plan. Therefore, this embodiment, by combining cost-benefit analysis with strategy optimization based on operational constraints, can formulate a charging and discharging pricing strategy that maximizes transaction revenue while meeting system operational limitations.

[0043] Specifically, both the user-side charging price prediction sequence and the generator-side discharging price prediction sequence are 96-point time series sequences generated above, corresponding to 24 hours per day with a time granularity of 15 minutes. The calculation needs to be completed in conjunction with the charging and discharging power time series characteristics of the energy storage power station. When calculating charging costs, the point-by-point price of the user-side charging price prediction sequence is used as a basis, and the planned charging power of the energy storage power station is associated with it. The charging volume for each time period is calculated through time integration (volume = power × time interval, where the time interval corresponds to a 15-minute duration conversion). Then, the volume for each time period is multiplied by the corresponding charging price for that time period, and the sum is obtained to get the total charging cost for the day. When calculating discharging revenue, the same time series matching logic is used. The point-by-point price of the generator-side discharging price prediction sequence is used as a basis, and the planned discharging power is associated with it to calculate the discharging volume for each time period. The volume for each time period is multiplied by the corresponding discharging price for that time period, and the sum is obtained to get the total discharging revenue for the day.

[0044] Secondly, conventional optimization objectives are simply represented as "revenue minus cost," without considering the timing coupling characteristics of charging and discharging. This embodiment introduces a timing integral term to accurately represent the total revenue over the entire time period. The optimization objective formula is as follows: R represents the total daily spot trading revenue of an independent energy storage power station, and maximizing R achieves the optimal trading effect; Pd represents the point-by-point price value of the generation-side discharge price prediction sequence; Pd,p represents the point-by-point discharge power (the decision variable to be optimized); Δt represents the time interval (corresponding to a 15-minute duration); Σ represents the integral summation over 96 time periods; Pc represents the point-by-point price value of the user-side charging price prediction sequence; Pc,p represents the point-by-point charging power (the decision variable to be optimized). In specific implementation, the time-series data of Pd and Pc obtained above are substituted in, and the fixed value of Δt is defined. The total revenue R is represented as a function of Pd,p and Pc,p. The optimal charging and discharging power time sequence allocation is determined by maximizing R. This formula accurately describes the coupling relationship between charging and discharging revenue and cost throughout the entire time period, overcoming the fuzzy defects of conventional target time sequence representation.

[0045] Next, the core operational constraints of the independent energy storage power station are defined, including energy storage capacity constraints (real-time energy storage capacity must be between the maximum and minimum capacity), charging and discharging power constraints (point-by-point charging and discharging power must not exceed the upper and lower limits of the equipment's rated power), mutual exclusion constraints of charging and discharging states (charging and discharging operations cannot be performed simultaneously at the same time), and daily charging and discharging energy balance constraints (the total charging energy and total discharging energy on a given day must meet the balance relationship under the energy storage conversion efficiency). Simultaneously, constraints from spot market trading rules must be incorporated (e.g., the quoted power must meet the minimum power step size requirement of the market). These constraints are transformed into mathematical constraint expressions, which, together with the optimization objective constructed above, constitute a complete optimization model. A conventional linear programming algorithm is used to solve this model. The algorithm iteratively searches for the Pd,p and Pc,p time series that satisfy all constraints and maximize the objective function R, obtaining an optimal charging and discharging power plan of 96 points. This plan ensures the safe and stable operation of the energy storage power station while complying with spot market trading rules.

[0046] Finally, the strategy transformation is completed based on the optimal charging and discharging power plan and the pricing rules of the spot market. For discharging quotations, the optimal discharging power is matched to the corresponding time period one by one according to 96 time granularities. The minimum acceptable discharging quotation for each time period is determined according to the generation-side discharging price prediction sequence (ensuring that the discharging revenue is not lower than expected). At the same time, the optimal discharging power is used as the discharging quotation power for each time period, forming a discharging quotation time sequence list. For charging quotations, the optimal charging power is matched to the corresponding time period. The maximum acceptable charging quotation for each time period is determined according to the user-side charging price prediction sequence (ensuring that the charging cost is not higher than expected). The optimal charging power is used as the charging quotation power for each time period, forming a charging quotation time sequence list. The two types of quotation lists are integrated, and the quotation information (including time period, quotation power, and quotation price) is standardized according to the market-specified format to generate a complete charging and discharging quotation strategy. This ensures that the strategy can be directly submitted to the spot market trading system, while ensuring that the quotation is fully matched with the optimal charging and discharging power plan, achieving the expected goal of maximizing trading revenue.

[0047] Furthermore, based on the aforementioned charging and discharging pricing strategy, the specific process of controlling the independent energy storage power station to perform charging and discharging operations is as follows: the optimal charging and discharging power plan in the charging and discharging pricing strategy is converted into a sequence of physical control commands and sent to the energy management system of the independent energy storage power station; the energy management system sends control commands for the corresponding time period to the energy storage converter according to the timing rhythm of the 96-point time sequence (every 15-minute interval); in specific operation, during the time period corresponding to the charging pricing time sequence list, the energy storage converter absorbs electrical energy from the grid according to the point-by-point charging power (Pc,p) in the command, and based on the preset charging... The electrical conversion efficiency stores electrical energy in the energy storage battery, performing a physical charging operation. During the corresponding discharge quotation time slot, the energy storage converter releases the electrical energy in the energy storage battery to the grid according to the point-by-point discharge power (Pd,p) in the instruction combined with the discharge conversion efficiency, performing a physical discharge operation. During the execution process, the energy management system monitors in real time and strictly meets the energy storage capacity constraints and charge / discharge power constraints, while ensuring that the charge / discharge states are mutually exclusive within the same minute time slot, and that no charge / discharge state switching conflict occurs, thereby achieving precise and automated control of the physical operation status of the independent energy storage power station based on the optimal charge / discharge power plan.

[0048] See Figure 2 This is a schematic diagram of a charging and discharging control system for an independent energy storage station based on similar spot trading dates, provided by an embodiment of the present invention. The charging and discharging control system for an independent energy storage station based on similar spot trading dates includes: Module 10 is used to acquire the power boundary dataset, meteorological dataset, and energy storage change dataset required for spot trading of independent energy storage power stations; The calculation module 11 is used to calculate the energy storage sensitivity coefficient based on the energy storage change dataset, and to perform time-varying feature weight allocation processing on the power boundary dataset to calculate the similarity between the current trading day and historical days and to filter out a preliminary set of similar days; the weight coefficient of the time-varying feature weight allocation processing is determined by the energy storage sensitivity coefficient and the annual growth rate of new energy. The vector generation module 12 is used to perform causal intervention processing on the historical spot price series corresponding to the preliminary similar day set, and generate a causal intervention vector to isolate the impact of changes in energy storage scale on spot prices; Optimization module 13 is used to perform online optimization processing based on time-varying exploration noise on the historical spot price series corresponding to the causal intervention vector, the meteorological deviation factor and the preliminary similar day set, to obtain the dynamic correction coefficient of each historical day; The correction module 14 is used to correct the historical spot price series using the dynamic correction coefficient to obtain a corrected price reference series, and based on the corrected price reference series, to calculate a user-side charging price prediction series that characterizes charging costs and a generator-side discharging price prediction series that characterizes discharging revenue. Strategy generation module 15 is used to generate a charging and discharging pricing strategy for an independent energy storage power station in the spot market based on the user-side charging price prediction sequence and the generator-side discharging price prediction sequence. Control module 16 is used to control the independent energy storage power station to perform charging and discharging operations based on the charging and discharging pricing strategy.

[0049] Compared with the prior art, the embodiments of the present invention have the following beneficial effects: First, acquire power boundary datasets, meteorological datasets, and energy storage change datasets. Calculate the energy storage sensitivity coefficient using the energy storage change dataset. Combine this with the annual growth rate of new energy sources to perform time-varying feature weighting on the power boundary dataset. Calculate the similarity between the current trading day and historical days to filter a preliminary set of similar days. Then, use causal intervention processing to isolate the independent impact of energy storage scale changes on spot prices, generating a causal intervention vector. Input this vector, a meteorological deviation factor, and the historical spot price series of the preliminary similar days into an online optimization process based on time-varying exploration noise to obtain dynamic correction coefficients. Use these coefficients to correct the historical spot price series to obtain a corrected price reference series, thereby generating user-side charging price and generator-side discharging price prediction series. Finally, formulate charging and discharging pricing strategies based on these predictions. Specifically, the time-varying feature weighting is determined by the energy storage sensitivity coefficient and the annual growth rate of new energy sources, adapting to the rapidly changing time-varying characteristics of new energy penetration and energy storage capacity, thus improving the accuracy of similar day selection. Causal intervention processing effectively identifies structural price deviations caused by changes in energy storage scale. Online optimization based on time-varying exploration noise combined with the meteorological deviation factor accurately corrects the price series. Separate price predictions for charging and discharging can specifically match trading demands. The embodiments of this invention can improve the accuracy of similar-day screening, reduce the prediction deviation of spot trading prices for independent energy storage power stations, and improve the accuracy of spot trading quotations, thereby achieving optimized control of the physical charging and discharging process of independent energy storage power stations. As can be seen from the above analysis, the embodiments of this invention improve the accuracy of similar-day screening and reduce the prediction deviation of spot trading prices for independent energy storage power stations by dynamically allocating feature weights based on energy storage sensitivity and using causal intervention vectors to isolate the impact of changes in energy storage scale, thereby improving the accuracy of spot trading quotations and achieving precise control of the charging and discharging of independent energy storage power stations.

[0050] As one example, the computing module is specifically used for: Extract the energy storage capacity growth rate and the new energy installed capacity growth rate from the energy storage change dataset; The ratio of the energy storage capacity growth rate to the new energy installed capacity growth rate is calculated to obtain the energy storage sensitivity coefficient; The load forecast, thermal power generation forecast, and new energy generation forecast contained in the power boundary dataset are obtained. The time-varying feature weight allocation processing is performed on the load forecast, thermal power generation forecast, and new energy generation forecast using the energy storage sensitivity coefficient and the annual growth rate of new energy, so as to obtain the weight coefficients corresponding to each of the three. Based on the respective weight coefficients of the three, the load forecast, the thermal power generation forecast, and the new energy power generation forecast are weighted and calculated to obtain the weighted similarity between the current trading day and historical days; The number of screening days is determined based on the historical price dispersion, and a corresponding number of historical days are selected from high to low based on the weighted similarity to generate the preliminary set of similar days.

[0051] As one example, the vector generation module is specifically used for: Based on the energy storage change dataset, determine the energy storage scale status for the current trading day and the energy storage scale status for historical trading days; Calculate the energy storage capacity difference between the current trading day's energy storage capacity status and the historical trading day's energy storage capacity status; The causal intervention process is applied to the historical spot price series, and the impact of changes in energy storage scale on prices is assessed based on the differences in energy storage scale to obtain the price adjustment component. Based on the price adjustment component, a causal intervention vector is generated to isolate the impact of changes in energy storage scale on spot prices.

[0052] As one example, the optimization module is specifically used for: The causal intervention vector, the meteorological deviation factor, and the historical spot price series are input into the online optimization process based on time-varying exploration noise; In the online optimization process based on time-varying exploration noise, an optimization objective including a correction bias penalty term is constructed; The time-varying exploration noise is generated, and the noise amplitude of the time-varying exploration noise is dynamically adjusted according to the energy storage sensitivity coefficient; The time-varying exploration noise is used to perturb the current correction coefficients, and the corrected correction coefficients are iteratively updated based on the optimization objective; When the optimization objective satisfies the convergence condition, the iteratively updated correction coefficient is output as the dynamic correction coefficient for each historical day.

[0053] It should be noted that the device embodiments described above are merely illustrative. The units described as separate components may or may not be physically separate, and the components shown as units may or may not be physical units; that is, they may be located in one place or distributed across multiple network units. Some or all of the modules can be selected to achieve the purpose of this embodiment according to actual needs. Furthermore, in the accompanying drawings of the device embodiments provided by this invention, the connection relationships between modules indicate that they have communication connections, which can be specifically implemented as one or more communication buses or signal lines. Those skilled in the art can understand and implement this without any creative effort.

[0054] The above description represents the preferred embodiments of the present invention. It should be noted that those skilled in the art can make various improvements and modifications without departing from the principles of the present invention, and these improvements and modifications are also considered to be within the scope of protection of the present invention.

Claims

1. A charging and discharging control method for an independent energy storage station based on similar spot trading days, characterized in that, Includes the following steps: Obtain the power boundary dataset, meteorological dataset, and energy storage change dataset required for spot trading of independent energy storage power stations; The energy storage sensitivity coefficient is calculated based on the energy storage change dataset, and the power boundary dataset is processed by time-varying feature weight allocation to calculate the similarity between the current trading day and historical days and to obtain a preliminary set of similar days; the weight coefficient of the time-varying feature weight allocation is determined by the energy storage sensitivity coefficient and the annual growth rate of new energy. Causal intervention processing is performed on the historical spot price series corresponding to the preliminary similar day set to generate a causal intervention vector for isolating the impact of changes in energy storage scale on spot prices; The historical spot price series corresponding to the causal intervention vector, the meteorological deviation factor, and the preliminary similar day set are subjected to online optimization processing based on time-varying exploration noise to obtain the dynamic correction coefficients for each historical day; The historical spot price series is corrected using the dynamic correction coefficient to obtain a corrected price reference series. Based on the corrected price reference series, a user-side charging price prediction series that characterizes charging costs and a generator-side discharging price prediction series that characterizes discharging revenue are calculated. Based on the user-side charging price prediction sequence and the generator-side discharging price prediction sequence, a charging and discharging pricing strategy for independent energy storage power stations in the spot market is generated; Based on the charging and discharging pricing strategy, the independent energy storage power station is controlled to perform charging and discharging operations.

2. The charging and discharging control method for independent energy storage stations based on similar spot trading days as described in claim 1, characterized in that, The process involves calculating the energy storage sensitivity coefficient based on the energy storage change dataset, performing time-varying feature weight allocation on the power boundary dataset, calculating the similarity between the current trading day and historical days, and filtering to obtain a preliminary set of similar days, including: Extract the energy storage capacity growth rate and the new energy installed capacity growth rate from the energy storage change dataset; The ratio of the energy storage capacity growth rate to the new energy installed capacity growth rate is calculated to obtain the energy storage sensitivity coefficient; The load forecast, thermal power generation forecast, and new energy generation forecast contained in the power boundary dataset are obtained. The time-varying feature weight allocation processing is performed on the load forecast, thermal power generation forecast, and new energy generation forecast using the energy storage sensitivity coefficient and the annual growth rate of new energy, so as to obtain the weight coefficients corresponding to each of the three. Based on the respective weight coefficients of the three, the load forecast, the thermal power generation forecast, and the new energy power generation forecast are weighted and calculated to obtain the weighted similarity between the current trading day and historical days; The number of screening days is determined based on the historical price dispersion, and a corresponding number of historical days are selected from high to low based on the weighted similarity to generate the preliminary set of similar days.

3. The charging and discharging control method for independent energy storage stations based on similar spot trading days as described in claim 2, characterized in that, The step of performing causal intervention processing on the historical spot price series corresponding to the preliminary similar day set to generate a causal intervention vector for isolating the impact of changes in energy storage scale on spot prices includes: Based on the energy storage change dataset, determine the energy storage scale status for the current trading day and the energy storage scale status for historical trading days; Calculate the energy storage capacity difference between the current trading day's energy storage capacity status and the historical trading day's energy storage capacity status; The causal intervention process is applied to the historical spot price series, and the impact of changes in energy storage scale on prices is assessed based on the differences in energy storage scale to obtain the price adjustment component. Based on the price adjustment component, a causal intervention vector is generated to isolate the impact of changes in energy storage scale on spot prices.

4. The charging and discharging control method for independent energy storage stations based on similar spot trading days as described in claim 3, characterized in that, The online optimization process based on time-varying exploratory noise is performed on the historical spot price series corresponding to the causal intervention vector, the meteorological deviation factor, and the preliminary similar day set to obtain the dynamic correction coefficient for each historical day; including: The causal intervention vector, the meteorological deviation factor, and the historical spot price series are input into the online optimization process based on time-varying exploration noise; In the online optimization process based on time-varying exploration noise, an optimization objective including a correction bias penalty term is constructed; The time-varying exploration noise is generated, and the noise amplitude of the time-varying exploration noise is dynamically adjusted according to the energy storage sensitivity coefficient; The time-varying exploration noise is used to perturb the current correction coefficients, and the corrected correction coefficients are iteratively updated based on the optimization objective; When the optimization objective satisfies the convergence condition, the iteratively updated correction coefficient is output as the dynamic correction coefficient for each historical day.

5. The charging and discharging control method for independent energy storage stations based on similar spot trading days as described in claim 4, characterized in that, The process of correcting the historical spot price series using the dynamic correction coefficient to obtain a corrected price reference series, and calculating a user-side charging price prediction series to characterize charging costs and a generator-side discharging price prediction series to characterize discharging revenue based on the corrected price reference series, includes: The historical spot price series is corrected using the dynamic correction coefficient to obtain the corrected price reference series; Adaptive time-series fusion processing is performed on the corrected price reference sequence to extract fused time-series features containing information from multiple time scales; Based on the charging and discharging roles of independent energy storage power stations, the fused timing features are classified and mapped into user-side charging timing features and generator-side discharging timing features. The energy storage conversion efficiency of the independent energy storage power station is obtained, and the energy storage conversion efficiency is used to numerically calibrate the user-side charging timing characteristics and the generator-side discharging timing characteristics, respectively. Based on the numerically calibrated features, the user-side charging price prediction sequence and the generator-side discharging price prediction sequence are generated.

6. The charging and discharging control method for independent energy storage stations based on similar spot trading days as described in claim 5, characterized in that, The step of generating a charging and discharging pricing strategy for independent energy storage power stations in the spot market based on the user-side charging price prediction sequence and the generator-side discharging price prediction sequence includes: The charging cost is calculated based on the user-side charging price prediction sequence, and the discharge revenue is calculated based on the generator-side discharge price prediction sequence. Combining the charging cost and the discharging benefit, an optimization objective is constructed with the goal of maximizing transaction revenue; The operational constraints of the independent energy storage power station are incorporated into the optimization objective, and the strategy optimization process is executed to obtain the optimal charging and discharging power plan that satisfies the spot trading rules. Based on the optimal charging and discharging power plan, a charging and discharging pricing strategy for the independent energy storage power station in the spot market is generated.

7. A charging and discharging control system for an independent energy storage station based on similar spot trading days, characterized in that, include: The acquisition module is used to acquire the power boundary dataset, meteorological dataset, and energy storage change dataset required for spot trading of independent energy storage power stations; The calculation module is used to calculate the energy storage sensitivity coefficient based on the energy storage change dataset, and to perform time-varying feature weight allocation processing on the power boundary dataset, calculate the similarity between the current trading day and historical days, and filter to obtain a preliminary set of similar days; the weight coefficient of the time-varying feature weight allocation processing is determined by the energy storage sensitivity coefficient and the annual growth rate of new energy. The vector generation module is used to perform causal intervention processing on the historical spot price series corresponding to the preliminary similar day set, and generate a causal intervention vector to isolate the impact of changes in energy storage scale on spot prices; The optimization module is used to perform online optimization processing based on time-varying exploratory noise on the historical spot price series corresponding to the causal intervention vector, the meteorological deviation factor, and the preliminary similar day set, so as to obtain the dynamic correction coefficient of each historical day. The correction module is used to correct the historical spot price series using the dynamic correction coefficient to obtain a corrected price reference series, and based on the corrected price reference series, to calculate a user-side charging price prediction series that characterizes charging costs and a generator-side discharging price prediction series that characterizes discharging revenue. The strategy generation module is used to generate a charging and discharging pricing strategy for independent energy storage power stations in the spot market based on the user-side charging price prediction sequence and the generator-side discharging price prediction sequence. The control module is used to control the independent energy storage power station to perform charging and discharging operations based on the charging and discharging pricing strategy.

8. The charging and discharging control system for an independent energy storage station based on similar spot trading days as described in claim 7, characterized in that, The calculation module is specifically used for: Extract the energy storage capacity growth rate and the new energy installed capacity growth rate from the energy storage change dataset; The ratio of the energy storage capacity growth rate to the new energy installed capacity growth rate is calculated to obtain the energy storage sensitivity coefficient; The load forecast, thermal power generation forecast, and new energy generation forecast contained in the power boundary dataset are obtained. The time-varying feature weight allocation processing is performed on the load forecast, thermal power generation forecast, and new energy generation forecast using the energy storage sensitivity coefficient and the annual growth rate of new energy, so as to obtain the weight coefficients corresponding to each of the three. Based on the respective weight coefficients of the three, the load forecast, the thermal power generation forecast, and the new energy power generation forecast are weighted and calculated to obtain the weighted similarity between the current trading day and historical days; The number of screening days is determined based on the historical price dispersion, and a corresponding number of historical days are selected from high to low based on the weighted similarity to generate the preliminary set of similar days.

9. The charging and discharging control system for an independent energy storage station based on similar spot trading days as described in claim 7, characterized in that, The vector generation module is specifically used for: Based on the energy storage change dataset, determine the energy storage scale status for the current trading day and the energy storage scale status for historical trading days; Calculate the energy storage capacity difference between the current trading day's energy storage capacity status and the historical trading day's energy storage capacity status; The causal intervention process is applied to the historical spot price series, and the impact of changes in energy storage scale on prices is assessed based on the differences in energy storage scale to obtain the price adjustment component. Based on the price adjustment component, a causal intervention vector is generated to isolate the impact of changes in energy storage scale on spot prices.

10. The charging and discharging control system for an independent energy storage station based on similar spot trading days as described in claim 7, characterized in that, The optimization module is specifically used for: The causal intervention vector, the meteorological deviation factor, and the historical spot price series are input into the online optimization process based on time-varying exploration noise; In the online optimization process based on time-varying exploration noise, an optimization objective including a correction bias penalty term is constructed; The time-varying exploration noise is generated, and the noise amplitude of the time-varying exploration noise is dynamically adjusted according to the energy storage sensitivity coefficient; The time-varying exploration noise is used to perturb the current correction coefficients, and the corrected correction coefficients are iteratively updated based on the optimization objective; When the optimization objective satisfies the convergence condition, the iteratively updated correction coefficient is output as the dynamic correction coefficient for each historical day.

Images