A dynamic risk assessment method and system for power spot market risk control

Abstract

This invention provides a dynamic risk assessment method and system for risk control in the electricity spot market, relating to the field of hydropower station management technology. It constructs a unified database by collecting and integrating heterogeneous data from multiple sources, including meteorology, power grid, and the market. A multi-timescale dynamic risk assessment model is used to predict key uncertainty factors, and a multi-objective optimization scheduling model is constructed using mixed-integer linear programming to achieve cascade linkage and risk-return balance. Simultaneously, it generates position optimization and robust pricing strategies, and introduces a CVaR (Continuous Value Assurance) pricing model to dynamically adjust risk hedging strategies. The system comprises a data layer, a model layer, an application layer, and a presentation layer, enabling data management, model calculation, interactive decision-making, and visualization, effectively improving the risk resilience and economic benefits of large hydropower stations in the spot market.

Description

Technical Field

[0001] This invention relates to the field of hydropower station management technology, specifically to a method and system for risk control in the electricity spot market based on dynamic risk assessment. Background Technology

[0002] With the continued deepening of power market reforms, large hydropower stations that have long operated under government-set prices or priority power generation plans and enjoyed relatively stable revenues are being included in the spot market, leading to a fundamental change in their revenue model. In the regional spot market environment, the core risk of participating in "volume and price" transactions lies in revenue risk, specifically manifested as volume deviation risk and pricing strategy risk. The former stems from the mismatch between actual power generation capacity and declared power volume, such as errors in water inflow forecasts or limitations imposed by other factors like flood control and irrigation requirements, resulting in penalties or losses from low-price substitution transactions. The latter manifests as the deviation between the declared price and the market clearing price; excessively high bids may result in the unit failing to win the bid, while excessively low bids sacrifice marginal revenue, both directly impacting revenue. Compared to traditional medium- and long-term markets, the time-sharing trading and high-frequency clearing characteristics of the spot market significantly increase decision-making complexity. Large hydropower stations need to comprehensively consider multiple factors such as water inflow uncertainty, grid dispatch constraints, and market supply and demand changes within a short period, dynamically adjusting their volume and price strategies. Adhering to the conventional thinking of maintaining volume and price in the medium- and long-term market could exacerbate revenue volatility due to decision-making delays or rigid strategies.

[0003] Existing risk control methods suffer from the following significant shortcomings: passive risk assessment, which can only simulate risk but cannot proactively adjust strategies to mitigate it; crude models with simplistic assumptions about the correlation between water inflow and electricity prices, ignoring the tail risk of low prices during the flood season; lack of cascade coupling, failing to consider the cascading effects of upstream power plant decisions on downstream areas; and poor dynamism, lacking real-time monitoring and closed-loop feedback mechanisms, making it difficult to cope with high-frequency market fluctuations. Therefore, establishing a risk quantification model adapted to the spot market, an agile decision-making mechanism, and implementing the aforementioned risk control mechanisms through information systems have become key breakthroughs in balancing returns and risks.

[0004] To address key pain points in spot market participation, such as price volatility, power consumption, and dispatch constraints, and to precisely meet the risk management needs of large hydropower stations participating in the Southern China spot market using a "volume-based bidding" model, this invention focuses on multi-dimensional collaborative optimization dispatch, dynamic risk assessment, and real-time strategy adjustment. By quantifying the revenue and risk of large hydropower stations in both the day-ahead and real-time markets, it covers multiple dimensions including water inflow uncertainty, grid dispatch constraints, and market supply and demand changes. Through the construction of a multi-timescale dynamic risk assessment model, it enables risk monitoring, early warning, and optimized decision support for large hydropower stations in spot market transactions, effectively meeting the refined and intelligent risk management needs of large hydropower stations under the "volume-based bidding" trading mechanism in the Southern China spot market. Summary of the Invention

[0005] The technical problem to be solved by this invention is to provide a method and system for risk control in the electricity spot market based on dynamic risk assessment. This method and system solves the core problems of price fluctuations, power consumption and dispatch constraints in the process of participating in the spot market, thereby improving the risk resistance and economic benefits of large hydropower in the spot market.

[0006] The core idea of ​​this invention is to achieve comprehensive management and dynamic control of the risks of large hydropower stations participating in the spot market through the following different levels of collaboration and intelligent decision-making.

[0007] This invention provides a method for risk control in the electricity spot market based on dynamic risk assessment, comprising the following steps: S1. During the annual trading period, collect historical data, establish a multi-scenario probability prediction model, predict the Value at Risk (VaR) and Conditional Value at Risk (CvaR) under multiple scenarios, and optimize the monthly and hourly holding targets for the annual trading. S2. During the monthly trading period, update historical data, establish a multi-scenario probability prediction model to predict risk values, and optimize the weekly intraday position targets for monthly trading based on the comparison between the expected transaction price in the mid-to-long term and the predicted spot price. S3. During weekly and multi-day trading periods, update historical data, establish multi-scenario probability prediction models to predict risk values, and optimize daily and hourly position targets for weekly trading. S4. During the day-ahead declaration and real-time operation period, based on the supply and demand conditions of the spot market and the operating constraints of the hydropower station, a control objective function is constructed to generate a segmented pricing strategy; S5. During the post-operation and settlement review period, the actual profit and loss are calculated based on the daily clearing and monthly settlement statements, and the price prediction model and risk value probability distribution model are updated based on the prediction error. By constructing a multi-timescale dynamic risk assessment model, risk monitoring, early warning, and optimization decision support for large-scale hydropower stations in spot market transactions can be achieved.

[0008] Preferably, in step S1, the multi-scene prediction model includes: The annual monthly power generation forecast for this hydropower station is generated under multiple scenarios, which stem from the uncertainty in the prediction of water inflow. The forecast values ​​for monthly spot clearing prices and settlement prices in the Southern Power Grid region's spot market throughout the year are generated under multiple scenarios. These multiple scenarios stem from the uncertainties in load demand, supply capacity, and the bidding behavior of market participants.

[0009] Preferably, in step S2, the objective of optimizing the weekly time-sharing position of monthly transactions is to maximize the expected return and the risk-adjusted return, as shown in the following formula (1): (1) in Represents the scene Expected returns To sell electricity at a higher price, In order to buy back the electricity price, For medium- to long-term electricity supply, For spot electricity volume, This is the risk aversion coefficient. This is the conditional risk value.

[0010] Preferably, step S2 also includes the following constraints: the total amount of shares sold each week does not exceed the total expected electricity generation to the grid for the whole month, and the average hourly electricity generation to the grid does not exceed the installed capacity.

[0011] Preferably, in step S3, the multi-scenario probability prediction model is based on medium-term weather forecast data, which is a 3-10 day forecast, and the optimization process takes into account the risk of power transmission channel quota and the constraints of hydropower station comprehensive utilization tasks.

[0012] Preferably, the control objective function in step S4 is shown in equation (2) below: (2) in, This refers to the electricity volume won in the recent bid. For time period The recent market clearing price, For the scene Expected returns and They are the scenes Real-time adjustment of power consumption up and down. In the scene Next period The real-time market clearing price, It is the risk aversion coefficient. This is the conditional risk value.

[0013] Preferably, in step S5, the review period is after the daily clearing statement is disclosed, the annual total risk value VaR and conditional risk value CvaR are updated based on actual data, and the probability distribution of the multi-scenario generation module is corrected.

[0014] This invention provides a dynamic risk assessment system for the electricity spot market that operates the aforementioned method, comprising a data layer, a model layer, an application layer, and a presentation layer: The data layer is used to collect and manage multi-source data such as historical hydrological and meteorological data, power grid operation data, market transaction data, and power plant equipment parameters. It adopts a hybrid storage scheme of time-series database and relational database, and performs outlier removal and rationality verification on the data. The model layer includes an uncertainty modeling and multi-scenario generation module, a multi-objective hydropower station collaborative operation optimization model, a medium- and long-term contract position optimization and spot market remaining capacity space generation optimization model, and a spot market declaration optimization model based on risk preference. The application layer is used to realize the interaction between users and the system, and supports model calling, parameter setting, transaction scheme generation and result feedback; The presentation layer uses visualization techniques to intuitively display prediction results, strategy recommendations, risk indicators, and performance evaluation results.

[0015] Preferably, the uncertainty modeling and multi-scenario generation module of the model layer uses a fusion method of LSTM-GAN and Monte Carlo simulation to generate multiple scenarios. The multi-objective hydropower station collaborative operation optimization model takes maximizing power generation revenue, minimizing CvaR risk, and maximizing the satisfaction of comprehensive utilization tasks as its objective functions.

[0016] Preferably, the combination of time series charts for weather and power output forecasts, heat maps of electricity prices for multiple scenarios, mulberry charts for medium- and long-term contract strategies, dashboards for spot market declaration strategies, radar charts of risk indicators, and bar and line charts for performance evaluation is used.

[0017] Compared with the prior art, the beneficial effects of the present invention are as follows: 1. Multi-source data fusion and prediction accuracy improvement: Solving the problem of coarseness of existing models: The LSTM-GAN hybrid model is used to enhance small sample extreme water inflow scenario data, obtain water inflow time series data features, use the Transformer model to capture the long-distance dependence of electricity price changes, use graph neural network to construct regional power grid topology map, model the spatial correlation of load, realize the deep integration of meteorological, power grid and market data, and provide comprehensive data support for risk assessment. 2. Cascade Coordination and Risk-Return Balance Optimization: A multi-timescale dynamic risk assessment model is constructed, combined with a multi-objective optimization scheduling model based on mixed-integer linear programming. For the first time, a quantitative modeling of the hydraulic-electric coupling relationship of cascade hydropower stations is achieved. This model comprehensively considers power generation revenue, risk avoidance, and integrated utilization tasks such as flood control and irrigation. It solves the problems of existing methods ignoring the cascade chain effect and revenue fluctuations or task conflicts caused by single-objective optimization. It achieves a dynamic balance between risk and return and is suitable for the time-sharing and high-frequency trading characteristics of the spot market.

[0018] 3. Enhanced Risk Quantification and Strategy Robustness: The CVaR risk measurement indicator is introduced, and the risk threshold is calculated using the Monte Carlo method. Combined with the time-of-use electricity price volatility matrix, the position ratio of medium- and long-term contracts and the spot market is optimized to generate a robust pricing strategy that adapts to risk preferences. This solves the shortcomings of existing risk control methods, such as the lack of quantitative indicators and rigid position and pricing strategies. It effectively hedges the profit risks caused by the deviation between supply and demand and the price deviation, and significantly improves the risk resistance capability under extreme scenarios.

[0019] 4. Adaptation to Comprehensive Utilization Tasks: The multi-objective optimization scheduling model incorporates the cascade linkage constraints and the satisfaction of comprehensive utilization tasks into the objective function. This solves the shortcomings of existing methods that do not consider the chain reaction of upstream power station decisions on downstream areas and are prone to conflicts with tasks such as flood control and ecological flow. While ensuring the revenue from spot market transactions, it ensures the compliant execution of comprehensive utilization tasks and meets the actual operational needs of multi-task collaboration in large hydropower stations. Attached Figure Description

[0020] The present invention will be further described below with reference to the accompanying drawings and embodiments: Figure 1 This is a diagram of the architecture of a dynamic risk assessment system for the electricity spot market according to the present invention. Figure 2 This invention relates to a method and system flowchart for risk control in the spot market for electricity from large hydropower stations. Detailed Implementation

[0021] Example 1 like Figure 1 As shown, a dynamic risk assessment method for risk control in the electricity spot market is presented.

[0022] S1. Quantification and control of risk during the annual trading period: If year Y is the year of the trading target, the annual trading usually takes place at the end of year Y-1. Assuming that the annual trading for year Y is organized in December of year Y-1, with day D as the trading execution date, the organization period for the annual trading is from day D-30 to day D-400.

[0023] Historical meteorological data such as temperature, precipitation, and wind and solar resources, historical fuel price data, and market and power plant clearing prices are collected to construct historical series of electricity market prices, hydropower station spot clearing prices, and hydropower station settlement prices in the southern region. Based on the multi-scenario prediction results of historical data and future boundary conditions, a multi-scenario probabilistic prediction model is established to predict risk values ​​under multiple scenarios, including Value at Risk (VaR) and Conditional Value at Risk (CvaR). According to the risk of different scenarios, the monthly and time-sharing position targets of annual transactions in each scenario are optimized. After the annual transaction is completed, the two risk values ​​are updated based on the monthly volume and price data of the completed transactions.

[0024] There are two types of intermediate variables with multiple scenarios: the predicted monthly on-grid power generation capacity of this hydropower station throughout the year, and the predicted monthly clearing and settlement prices of this hydropower station in the Southern Power Grid spot market throughout the year. The reason for the multiple scenarios in the spot market for this hydropower station is that the accuracy of predicting the monthly water inflow for each month of year Y before year Y is relatively poor, resulting in various scenarios. The reason for the multiple scenarios in the Southern Power Grid spot market is that demand in the Southern Power Grid spot market is affected by three factors: load demand, supply capacity, and the bidding behavior of market participants. Load demand is affected by factors such as weather and holidays; supply capacity is affected by wind, solar, and hydropower resources throughout the Southern Power Grid; and bidding behavior is affected by international coal and gas prices. Therefore, before the annual transaction, the uncertainty of the above three factors is significant, resulting in multiple scenarios.

[0025] For the annual timescale, the main risks to focus on are: extreme changes in market supply and demand conditions during the operation period, leading to significant deviations in the monthly and hourly price characteristics of the spot market compared to the predicted values; and significant deviations in the power plant's generating capacity compared to normal seasons due to extreme weather events, such as a reversal of the wet season to a dry season, or vice versa. Operational constraints are not currently considered, while risks arising from trading and settlement mechanisms are long-term and impact the market through electricity price and power generation risks. The annual cross-provincial priority plan holdings, price, and their risk value and conditional risk value will be output.

[0026] S2. Monthly Trading Period: If month M is the target month for trading, monthly trading usually takes place around January 20th (M-1). With day D as the execution date, the organization period for monthly trading is from day D-15 to day D-45.

[0027] Collect meteorological forecast data such as temperature, precipitation, and wind and solar resources for the target month and from the target month to the end of the year; update fuel price data; update market clearing prices from the beginning of the year to the most recent date and clearing prices at the nodes where power plants are located; update historical series of electricity market prices in the southern region, historical series of spot clearing prices for hydropower stations, and historical series of settlement prices for hydropower stations; and establish a multi-scenario probability prediction model based on the multi-scenario prediction results of historical data and future boundary conditions to predict risk values ​​under multiple scenarios, including risk value and conditional risk value. Based on the risk of different scenarios, optimize the weekly intraday holding targets for the next month's transactions in each scenario.

[0028] Risk control methods when formulating trading strategies before monthly trading: The known conditions before monthly trading are: the existing medium- and long-term holding volume and price of hydropower stations for the next month, and the holding volume and price are all time-sharing monthly averages; the monthly predicted grid-connected power generation with probability distribution for multiple scenarios, which can be optimized by shifting between daily and time-sharing; the monthly spot clearing price with probability distribution for multiple scenarios, the weekly average price, and the monthly time-sharing average price.

[0029] The optimization objective for monthly trading is as follows: If the current mid-to-long-term projected transaction price for a certain week of the next month is higher than the predicted spot price for that week, then increase the mid-to-long-term selling volume in the electricity market to increase mid-to-long-term open interest; if the current mid-to-long-term projected transaction price for a certain week of the next month is lower than the predicted spot price for that week, then sell mid-to-long-term contracts to reduce mid-to-long-term open interest.

[0030] Constraints: The total weekly sales volume shall not exceed the expected monthly power generation and grid connection volume; the hourly average power generation and grid connection volume after adding the monthly sales volume to the annual monthly sales volume shall not exceed the installed capacity; and the monthly buyback volume shall not exceed the annual monthly sales volume. Based on a certain risk aversion coefficient, decisions are made regarding the amount of electricity sold or bought back each week, along with the corresponding price. The objective function is... To maximize the expected return and the risk-adjusted return, as shown in equation (3): (3) in Represents the scene Expected returns To sell electricity at a higher price, In order to buy back the electricity price, For medium- to long-term electricity supply, For spot electricity volume, This is the risk aversion coefficient. Conditional value.

[0031] After monthly transactions are completed, the two risk values ​​for the entire year are updated based on historical completed settlement volume and price data, cumulative transaction volume and price data for the next month, and updated weather forecast data, coal price and gas price forecasts for the remaining time of the year. For the monthly timescale, the main risk focuses include: extreme changes in market supply and demand conditions during the operation period, leading to significant deviations from the predicted values ​​for monthly, weekly, and hourly price characteristics in the spot market; significant deviations in the power plant's generating capacity compared to normal seasons due to extreme weather events, such as a reversal of high water levels during the wet season or vice versa; and special operational requirements in the water allocation plan, such as constraints on the power plant's generating capacity due to multiple comprehensive utilization tasks including flood control, water supply, ecological flow guarantee, and navigation. Although such constraints are exempt from assessment under current rules in the Southern Power Grid region, they still affect the actual settled electricity volume. For cascade power plants, upstream and downstream constraints also need to be considered. The updated values ​​for medium- and long-term holdings volume and price, the current year's risk value, and the conditional risk value are output for the next month.

[0032] S3, medium- to long-term trading on a weekly or multi-day basis, uses D day as the target date, and the trading window is usually from D-5 to D-14.

[0033] Collect meteorological forecast data such as temperature, precipitation, and wind and solar resources for the target week and from the current month to the end of the year; update the latest fuel price data; update the market clearing price from the beginning of the year to the most recent date and the clearing price at the node where the power plant is located; update the historical series of electricity market prices in the southern region, the historical series of spot clearing prices for hydropower stations, and the historical series of settlement prices for hydropower stations; and establish a multi-scenario probability prediction model based on the multi-scenario prediction results of historical data and future boundary conditions to predict the risk value under multiple scenarios, including risk value and conditional risk value. Based on the risk of different scenarios, optimize the daily and time-sharing position targets for weekly trading in each scenario.

[0034] The risk control methods used when formulating trading strategies before weekly trading are similar to those used for monthly trading, except that the decision-making content changes from buying and selling on a weekly basis to buying and selling on a daily basis.

[0035] After the weekly transaction is completed, the two risk values ​​for the whole year are updated based on the historical completed settlement volume and price, the cumulative transaction volume and price data for the next week, and the updated weather forecast data, coal price and gas price forecast data for the remainder of the year.

[0036] For a weekly timescale, compared to the 45-day long-term weather forecasts used in monthly trading, medium-term (3-10 day) weather forecasts have achieved a certain level of accuracy; the fluctuation range of fuel market price data forecasts has also been significantly reduced; large hydropower stations have large reservoir capacities, and by the time of weekly trading, the weekly on-grid power generation capacity of the target asset can be transformed from a predictive capability into a target for optimizing trading strategies. Therefore, the accuracy of forecasts for multiple variables, including spot market clearing prices and settlement prices, and the on-grid power generation capacity of this hydropower station, has been significantly improved, and risk has gradually been transformed into certain profit and loss.

[0037] The main risks before and after the week's trading session are: changes in the transmission capacity limits of large hydropower stations. Generally, the possibility of large hydropower stations curtailing power due to transmission capacity limits is small, but such events cannot be ruled out in extreme cases; medium-term (3-10 day) weather forecasts, while possessing a certain degree of accuracy, still have significant forecast biases, which are a major source of market price volatility risk; and special operational requirements arising from water diversion plans, as the power plant's generating capacity may be constrained due to its multiple comprehensive utilization tasks, including flood control, water supply, ecological flow guarantee, and navigation. For cascade hydropower stations, upstream and downstream constraints also need to be considered, ultimately resulting in updated mid-to-long-term holdings, price, overall risk value for the year, and conditional risk value.

[0038] S4. Day-ahead declaration and real-time operation: Day D is the target date, Day D-1 is the day-ahead market declaration, and the hydropower station on Day D is actually operated according to the real-time bidding results.

[0039] With the supply and demand conditions in the spot market almost fixed, thus determining the day-ahead and real-time price levels, large hydropower stations, due to their high reservoir capacity, can almost meet various demand curves. Therefore, the main risks during the day-ahead bidding and real-time operation periods stem from factors affecting the actual output of large hydropower stations, primarily operational constraints. During this phase, day-ahead bidding needs to meet the risk control targets imposed by the trading mechanism, including the segmented pricing monotonicity and operational constraints.

[0040] The control objective function at the time of the recent declaration is shown in equation (4) below: (4) in, This refers to the electricity volume won in the recent bid. For time period The recent market clearing price, For the scene Expected returns and They are the scenes Real-time adjustment of power consumption up and down. In the scene Next period The real-time market clearing price, It is the risk aversion coefficient. This is the conditional risk value.

[0041] In addition to the rule-required conditions such as the declared price needing to be non-decreasing and the annual inter-provincial plan needing to be cleared in the market before the due date, the constraints also include hydropower-related operational constraints, such as water balance constraints and unit vibration zone constraints. The updated values ​​of the total value at risk and conditional value at risk for the year are output.

[0042] S5. Review after operation and settlement (>=5 days), with D day as the target date. The daily settlement statement disclosure time is usually from D+6 to D+8, and the monthly settlement statement disclosure date is around the 13th of the following month.

[0043] At this stage, all transactions and operations have been completed. Actual profits and losses need to be calculated based on daily and monthly settlement statements. In particular, it is necessary to verify the daily and monthly settlement statements and calculate the profits and losses from medium- and long-term transactions, spot transactions, and adjustments to settlement rules. Based on prediction errors, various price and power prediction models are updated, and risk value statistics or probability distribution models are updated or corrected. The output of this node is updated again with the annual total value at risk and conditional value at risk based on actual data.

[0044] The above steps can be achieved by adding risk functions to the power market auxiliary decision-making system of the hydropower station or by establishing a separate risk control system.

[0045] Example 2 like Figure 2 As shown, a dynamic risk assessment and control system for the electricity spot market is presented.

[0046] The data layer is responsible for collecting and managing various types of data, including historical hydrological and meteorological data, power grid operation data, market transaction data, and power plant equipment parameters. Historical hydrological and meteorological data covers temperature, precipitation, and wind and solar resources in the southern region for over 10 years, as well as special meteorological data such as typhoon paths, landfall frequency, and intensity, illustrating their impact on hydropower transactions under different meteorological scenarios. Real-time collection of regional power grid load data, transmission line power flow data, and unit grid-connected operation status reflects the limitations imposed on transactions by power grid physical constraints. Simultaneously, it analyzes historical spot prices, medium- and long-term contract transaction data, and ancillary service market data in the southern regional power grid to construct a complete market price system and its mapping relationship with transaction rules. Power plant equipment parameters store static parameters such as rated output, minimum output, vibration zone range, reservoir capacity curve, and turbine efficiency curve of hydropower station units, as well as dynamic parameters such as equipment maintenance plans and fault history records, supporting physical constraint modeling.

[0047] The data management mechanism of the data layer adopts a hybrid storage scheme of time-series and relational databases. The time-series database stores dynamic real-time data, enabling efficient processing of time-series data writing and querying operations, while the relational database stores structured static data, ensuring the accuracy of data association queries. Outliers are removed using the 3σ principle for meteorological and price data. For equipment parameters and power grid constraints, rationality verification is performed based on the experience of hydropower industry experts to ensure the data quality input to the model. Meteorological data, power grid operation data, and market transaction data are updated in real time, power plant equipment parameters are updated based on event triggers, and historical data is archived daily or monthly, providing comprehensive data support for the model layer.

[0048] The model layer comprises five core algorithms and model libraries. The uncertainty modeling and multi-scenario generation module, based on meteorological and market data from the data layer, utilizes a fusion method of LSTM-GAN and Monte Carlo simulation to generate multiple scenarios covering uncertainties in the electricity market and water temperature and meteorological conditions. The multi-objective hydropower station collaborative operation optimization model, considering uncertainty, coordinates the physical constraints and multi-objective requirements of cascade hydropower stations. It constructs a multi-objective optimization model with three objective functions: maximizing power generation revenue, minimizing CvaR risk, and maximizing the satisfaction of comprehensive utilization tasks, outputting a collaborative operation strategy. The medium- and long-term contract position optimization and spot market remaining capacity space generation optimization model, considering uncertainty, mainly optimizes the medium- and long-term contract position structure for the year or month, matching spot market to lock in profits and control risks. Using the electricity price and output scenarios from the multi-scenario generation module as input, it calculates the returns and risks of different contract time-sharing positions.

[0049] It also includes a spot market bidding optimization model based on risk appetite. By integrating risk appetite with real-time market information, a robust spot market is generated. By introducing a risk aversion parameter λ and a risk trade-off parameter γ, a return and risk trade-off model is constructed. On the return side, it combines multi-scenario electricity price forecasts to calculate the expected winning bid returns for different bids. On the risk side, it uses a GARCH model to calculate the real-time electricity price volatility σ and combines it with CVaR to assess extreme losses. The rolling closed-loop dynamic adjustment mechanism dynamically corrects model parameters and strategies based on actual market operation data, forming a closed loop of prediction, decision-making, execution, and feedback. It updates daily based on spot market clearing results and weather forecasts, calls up various modules in the model layer to recalculate, adjusts daily bidding strategies and unit output plans, and then backtests based on monthly or quarterly settlement data to count the number of times VaR and CVaR exceedances, corrects the probability distribution of the multi-scenario generation module to optimize the objective function weights of the model, and continuously improves the model's adaptability.

[0050] The application layer primarily serves as the interaction between users and the system, enabling the entire process of model invocation, strategy generation, and result feedback. Its main functions include: users can select and invoke corresponding model modules based on the trading cycle; for example, annual trading will invoke the uncertainty modeling and multi-scenario generation model, as well as the medium-to-long-term contract position optimization model. Model parameters, such as the risk aversion parameter λ, confidence level β, and comprehensive task priority, can be set through a visual interface to flexibly adapt to different decision-making needs. Based on user configuration, the system automatically invokes model layer algorithms to generate multi-dimensional trading plans, including medium-to-long-term trading plans and spot reporting plans. The medium-to-long-term trading plan includes initial cross-provincial priority plan positions, monthly contract adjustment volumes, and intraday position structures. The spot reporting plan includes day-ahead reporting curves and real-time risk hedging strategies. It also supports plan version management and comparative analysis to assist user decision-making, and feeds back the model calculation results to the user.

[0051] The presentation layer uses visualization techniques to transform complex model outputs into intuitive and easy-to-understand decision-making support. Forecast result visualization includes weather and power output forecasts and electricity price forecasts. Weather and power output forecasts use event sequence charts to display water inflow and wind / solar power output forecasts for the next month, overlaid with historical quantile intervals to visually represent uncertainty. Electricity price forecasts use multi-scenario electricity price heatmaps to show peak-valley electricity price distribution and the probability and impact of extreme scenarios. Strategy suggestion visualization includes medium- and long-term contract strategies and spot bidding strategies. Medium- and long-term contract strategies use a mulberry chart to show contract power flow, marking risk thresholds and channel limits at each stage, clearly presenting the position optimization logic. Spot bidding strategies generate a bidding curve dashboard that displays segmented bids and the probability distribution of winning bids in real time, allowing users to drag and adjust bid ranges and calculate risk changes in real time. Risk indicators use radar charts to display the core risk indicators of the current strategy, comparing them with historical strategies to help determine the degree of risk exposure. Performance evaluation uses a combination of bar charts and line charts to display actual returns and prediction error rates for different trading periods, quantifying model accuracy and strategy effectiveness, and providing a basis for subsequent parameter optimization.

[0052] The specific embodiments described above do not constitute a limitation on the scope of protection of this disclosure. Those skilled in the art should understand that various modifications, combinations, sub-combinations, and substitutions can be made according to design requirements and other factors. Although embodiments of this application have been shown and described above, it is understood that the above embodiments are exemplary and should not be construed as limiting this application. Those skilled in the art can make changes, modifications, substitutions, and variations to the above embodiments within the scope of this application.

Claims

1. A method for dynamic risk assessment of power spot market risk control, characterized in that, The method comprises the following steps: S1, during the annual transaction, historical data is collected, a multi-scenario probability prediction model is established, risk values VaR and CvaR are predicted, and the monthly and hourly holding targets of the annual transaction are optimized; S2, during the monthly transaction, historical data is updated, a multi-scenario probability prediction model is established to predict risk values, the weekly and hourly holding targets of the monthly transaction are optimized according to the comparison relationship between the long-term predicted transaction price and the predicted spot price; S3, during the weekly and multi-day transaction, historical data is updated, a multi-scenario probability prediction model is established to predict risk values, and the daily and hourly holding targets of the weekly transaction are optimized; S4, during the day-ahead declaration and real-time operation, a control target function is constructed based on the supply and demand conditions of the spot market and the operation constraints of the hydropower station, and a segmented bidding strategy is generated; S5, after the operation and settlement, the actual profit and loss are counted according to the daily clearing order and the monthly settlement order, the price prediction model and the risk value probability distribution model are updated based on the prediction error; Through the construction of a multi-time scale dynamic risk assessment model, risk monitoring, early warning and optimized decision support in the spot market transaction of a large hydropower station are realized.

2. The method of claim 1 wherein, In step S1, the multi-scenario prediction model comprises: The multi-scenario generation of the annual monthly predicted value of the power generation and grid access of the hydropower station, the multi-scenario is derived from the uncertainty of the prediction of the water inflow; The multi-scenario generation of the annual monthly predicted value of the spot market clearing price and settlement price of the regional spot market of the South Power Grid, the multi-scenario is derived from the uncertainty of the load demand, supply capacity and bidding behavior of market participants.

3. The method of claim 1 wherein, In step S2, the weekly and hourly holding targets of the monthly transaction are optimized to maximize the expected return and risk-adjusted return, as shown in the following formula (1): (1)； wherein represents the expected benefit from the scenario, is the selling price, is the buy-back price, is the medium- and long-term electricity quantity, is the spot electricity quantity, is the risk aversion coefficient, is the conditional risk value.​ 4. The method of claim 1 wherein, In step S2, the constraint condition is that the total holding amount of each week is not more than the predicted monthly power generation and grid access, and the hourly average power generation and grid access power does not exceed the installed capacity.

5. The method of claim 1 wherein, In step S3, the multi-scenario probability prediction model is based on medium-term meteorological forecast data, and the optimization process considers the transmission channel limit risk and the comprehensive utilization task constraints of the hydropower station.

6. The method of claim 1 wherein, In step S4, the control target function is as shown in the following formula (2): (2)； in, This refers to the electricity volume won in the recent bid. For time period The recent market clearing price, For the scene Expected returns and They are the scenes Real-time adjustment of power consumption up and down. In the scene Next period The real-time market clearing price, It is the risk aversion coefficient. This is the conditional risk value.

7. The method of claim 1 wherein, In step S5, after the daily clearing order is disclosed, the overall risk value VaR and the conditional risk value CvaR are updated based on the actual data, and the probability distribution of the multi-scenario generation module is corrected.

8. A dynamic risk assessment power spot market risk control system, characterized by, It comprises a data layer, a model layer, an application layer and a display layer: The data layer is used for collecting and managing historical hydrological and meteorological data, power grid operation data, market transaction data and multi-source data of power station equipment parameters, adopts a hybrid storage scheme of time series database and relational database, and performs outlier elimination and reasonableness verification on the data; The model layer comprises an uncertain modeling and multi-scenario generation module, a multi-objective hydropower station collaborative operation optimization model, a long-term contract holding optimization and residual capacity space generation optimization model of the spot market, and a spot market declaration optimization model based on risk preference; The application layer is used for realizing the interaction between the user and the system, supporting model calling, parameter setting, transaction scheme generation and result feedback; The display layer visually displays the prediction results, strategy suggestions, risk indicators and performance evaluation results through visualization means.

9. The system of claim 8, wherein, The uncertainty modeling and multi-scenario generation module of the model layer generates multi-scenarios by using a fusion method of LSTM-GAN and Monte Carlo simulation; The multi-objective hydropower station collaborative operation optimization model takes the maximum power generation income, the minimum CvaR risk and the highest comprehensive utilization task satisfaction as objective functions.

10. The system of claim 8, wherein, The time series chart of weather and output prediction, the multi-scenario electricity price heat map, the mulberry chart of medium and long-term contract strategy, the bidding curve instrument panel of spot bidding strategy, the radar chart of risk index and the combination of bar chart and line chart of performance evaluation.

Images