Complex constraint-based adaptive time scale hydropower spot double-layer quantity price declaration method

Abstract

The invention provides a complex constraint-based adaptive time scale hydropower spot double-layer price declaration method, and the method comprises the steps: constructing a hydropower price declaration model, which comprises two stages: the first stage is to decide an optimal expected output curve by taking income maximization as a target; in the second stage, the optimal quotation curve is optimized by taking realization of the optimal expected curve as a target; through the above steps, hydroelectric spot price declaration is realized. According to the invention, the spot market electricity price prediction model is constructed, and the trading strategy and the trading model of the hydropower station participating in the spot market are researched based on the spot electricity price prediction result, so that the trading decision risk can be effectively controlled while the marketization trading income of the hydropower station is improved; according to the method, hydropower station market income-risk balance management is realized, power station comprehensive utilization constraint requirements are met, daily spot transaction declaration work of the hydropower station is effectively supported, business process carding and scientific and reasonable theoretical method guidance are provided for large hydropower station power generation project market transaction decisions, and a hydropower station marketization operation mechanism can be accelerated and perfected.

Description

Technical Field

[0001] The present invention relates to the technical field of hydropower station management, in particular to a self-adaptive time-scale hydropower spot double-layer quantity and price declaration method based on complex constraints. Background Art

[0002] As the domestic electricity market is still in its developmental stages, prioritizing the integration of thermal power and renewable energy, existing decision-making models primarily focus on renewable energy and thermal power. Existing output forecasting models for renewable energy generation typically rely on meteorological data and generation characteristics for short-term predictions, while thermal power generation optimizes scheduling and bidding based on fuel consumption and equipment status. These decision-making models are mostly based on large-scale data analysis and machine learning techniques, combined with electricity market rules, to maximize profits. However, hydropower has unique characteristics in terms of generator start-up and shutdown, reservoir capacity regulation, and output fluctuation response, as well as being constrained by various factors such as water resource scheduling and discharge restrictions. Therefore, existing models for renewable energy and thermal power cannot be directly applied to hydropower output forecasting and decision-making. Existing research on hydropower decision-making models is relatively limited, primarily focusing on theoretical research. There is a lack of mature models specifically tailored to dynamic output optimization and bidding strategies for hydropower in the spot market. Currently feasible theoretical approaches include trading decision-making models based on electricity price forecasts, game theory, and bid ranking. Among them, methods based on game theory require a large amount of data, including the bidding strategies of market entities, grid information, and the behavior of other market participants. In actual markets, complete information is difficult to obtain, which limits the practicality of game theory methods. Methods based on bid ranking usually only consider price factors, ignoring other factors that may affect electricity prices, such as market supply and demand conditions and emergencies. They lack flexibility and are difficult to adjust strategies according to changing market conditions in actual transactions. Considering the current imperfect market mechanisms and information disclosure, methods based on electricity price forecasting, by focusing on market price forecasts, can provide more flexible and forward-looking decision support without fully understanding other entities' information. This effectively reduces the impact of information asymmetry on decision-making and maximizes profits.

[0003] Chinese patent document CN111552912A describes a two-layer economic optimization method for microgrid grid connection, including: constructing a lower-layer microgrid optimization model; solving the purchase / sale potential of each microgrid in upper-layer transactions, as well as information on the purchase / sale price; constructing an upper-layer optimization model for energy transfer paths based on graph theory; and solving the lower-layer microgrid optimization model and the upper-layer optimization model using the JAYA Dijkstra algorithm to derive the optimal transmission path for each microgrid's output power. CN113705861A describes a method for optimizing hydropower station operations in a power market environment based on a genetic algorithm. The method constructs an objective function, sets constraints, and uses a genetic algorithm to solve the objective function to obtain a startup plan that minimizes total water consumption while maximizing revenue. However, none of the above-mentioned methods can be used to guide pricing decisions for large hydropower stations. Summary of the Invention

[0004] The technical problem to be solved by the present invention is to provide a dual-layer quantity and price declaration method for hydropower spot market based on complex constraints and adaptive time scale. The method can fully consider various complex constraints such as operation, vibration zone, climbing, scheduling, start and stop, storage capacity and water level discharge of large hydropower stations, set a certain time period as the minimum unit, optimize a certain time length to calculate the optimal expected output of the hydropower station, and further optimize the current quotation strategy of the hydropower unit in the multi-time scale spot market with the goal of maximizing the profit on the basis of the optimal predicted output of hydropower, and construct the optimal output and optimal strategy model of the hydropower station under complex constraints and adaptive time scale.

[0005] To solve the above technical problems, the technical solution adopted by the present invention is: a method for self-adaptive time-scale hydropower spot double-layer quantity and price declaration based on complex constraints, comprising the following steps:

[0006] S1. Construct a hydropower price declaration model, which includes two stages. The first stage is to determine the optimal expected output curve with the goal of maximizing revenue;

[0007] S2, the second stage is to optimize the optimal quotation curve with the goal of achieving the optimal expected curve;

[0008] The above steps can be used to declare the spot quantity and price of hydropower.

[0009] In the preferred solution, in step S1, the profit maximization decision model is based on the following formula:

[0010]

[0011] Where, is the winning bid power of trading unit c in the day-ahead market during period t; c is the power consumption rate of transaction unit c; Decompose the power of the contract for trading unit c in period t; Predict the day-ahead market settlement electricity price for time period t; M is the deviation between the current bid electricity and the recommended plan, Qdiff is the penalty factor; is the water discharge of reservoir r in period t, M Abq Penalty constraint for abandoned water flow; M is a sliding variable whose adjustment direction is not strictly consistent with the same library. div It is a penalty factor for inconsistent adjustment direction;

[0012] Penalty factor setting priority:

[0013]

[0014] The optimization time period can be set to a single day or multiple days.

[0015] In the preferred solution, S101, the profit maximization decision model is also constrained by the relationship between reservoirs, trading units, and dispatching units:

[0016] That is, the output of a trading unit in each period is the sum of the outputs of the dispatching units under the trading unit in the corresponding period; the output of a reservoir in each period is the sum of the outputs of the trading units under the reservoir in the corresponding period;

[0017]

[0018] Where, is the expected standard output of unit i in period t, Mark the expected force for trading unit c in period t; Denote the expected force on reservoir r during period t.

[0019] In the preferred solution, S102, the profit maximization decision model is also subject to the technical characteristics of the dispatch unit, including:

[0020] S1021, stable operating range constraints, i.e. the expected winning bid output should be within the allowable output range;

[0021]

[0022] Where, α i,t is the startup status of unit i in period t, The maximum and minimum output limits of unit i in time period t, where the minimum output limit is the upper limit of the unit vibration zone;

[0023] S1022, avoid the constraints of multiple vibration zones,

[0024] σ i,t,s =1 (6);

[0025] The highest gear vibration zone avoided by hydropower unit i in time period t is s, and the output of unit i in time period t is above the upper limit of the vibration zone s and below the lower limit of the vibration zone s+1:

[0026] ∑σ i,t,s ≤1 (7);

[0027]

[0028] Where S is the number of vibration sections, are the lower and upper limits of the output of unit i in vibration zone s, respectively;

[0029] S1023, ramp rate constraints due to shipping reasons;

[0030] S10231. The unit output change during each period should be within the allowable range of the ramp rate;

[0031]

[0032] Where, is the regulation rate of unit i, β i,t is the indicator variable for the direction of increase or decrease of unit i in period t, β i,t =1 means the period from t-1 to t is upward adjustment, β i,t =0 means the period from t-1 to t is downward adjustment;

[0033] S10232, the output variation of the trading unit, that is, the output variation of each time period should be within the allowable range of the power station ramp rate;

[0034]

[0035] Where, The adjustment rate of transaction unit c is determined according to the output amplitude of dispatch and the downstream flow of shipping, and has the function of connecting with the last period of the previous day;

[0036] S10232. Reservoir output variation, that is, the output variation in different time periods should be within the allowable range of the reservoir ramp rate;

[0037]

[0038] Where, is the regulation rate of reservoir r, and its value is determined according to the dispatching output amplitude and shipping discharge, and has the function of connecting with the last period of the previous day.

[0039] S1024, start-stop peak load regulation frequency constraint;

[0040] Statistical output 0->1, that is, the number of start-stop conversions, which must not exceed the set value;

[0041] u i,t -z i,t =α i,t -α i,t-1 (13);

[0042]

[0043] Where u i,t 、z i,t is the state variable of the startup and shutdown actions of unit i in period t, The upper limit of the number of start and stop switches of unit i in a single day.

[0044] In the preferred solution, S103, the profit maximization decision model is also subject to the hydraulic characteristics of the reservoir, including:

[0045] S1031, reservoir capacity balance constraint;

[0046]

[0047] Where V r,t is the storage capacity of reservoir r in time period t, in m 3 ; is the inflow of reservoir r in period t, is the power generation flow of reservoir r in period t, is the water discharge of reservoir r in time period t, in m 3 / s;η c,t is the water consumption rate of transaction unit c in time period t, in m 3 / kWh;

[0048] S1032, water level range constraint;

[0049]

[0050] Where, are the upper and lower limits of the storage capacity of reservoir r in time period t, respectively, in m 3 , obtained by converting the upper and lower limits of the water level according to the water level storage capacity curve;

[0051] S1033, ecological discharge flow range constraints;

[0052]

[0053] Where, The maximum discharge flow and the minimum ecological discharge flow of reservoir r are used to meet the comprehensive water demand of reservoir operation.

[0054] In the preferred solution, S104, the profit maximization decision model is also subject to power plant regulation demand constraints, including:

[0055] S1041, power station total power generation ratio constraint;

[0056]

[0057] Where, c,max 、υ c,min are the upper and lower limits of the proportion of power generation of power station c to reservoir r;

[0058] S1042, the power station daily output variation range is constrained, that is, the difference between the maximum and minimum output within the day is less than the set value:

[0059]

[0060] Where, The maximum and minimum output of the power station on that day, The maximum acceptable value of the daily maximum and minimum output fluctuations;

[0061] S1043, the reservoir daily output variation range is constrained, that is, the difference between the maximum and minimum output within a day is less than the set value:

[0062]

[0063] Where, The maximum and minimum output of the reservoir on that day, It is the maximum acceptable value of the daily maximum and minimum output fluctuations, and has the function of connecting with the last moment of the previous day;

[0064] S1043, one-way adjustment of Tongku Hydropower Plant;

[0065]

[0066] Where, β c＝1,t , β c＝2,t It is the indicator variable of the upward and downward regulation direction of the hydropower station pair in the same reservoir; It is a sliding variable whose adjustment direction is not strictly consistent with the same library;

[0067] S1044, the allowable range of deviation between the reservoir's recently awarded electricity quantity and the recommended scheduling plan;

[0068]

[0069] Where, is the hourly average output plan value recommended for reservoir r's cascade regulation, and They are the upper limits of the allowable range of positive and reverse deviations between the day's winning bid electricity and the planned value, respectively.

[0070] In the preferred solution, S105, the profit maximization decision model is also subject to power plant output constraints, including:

[0071] S1051. Control requires that the power station output be constrained throughout the day, i.e. the power station output be within the upper and lower limits of the daily output:

[0072]

[0073] Where, and are the lower and upper output limits of power station c respectively;

[0074] S1052, power station priority power generation plan output constraints;

[0075]

[0076] Where, and are the lower and upper output limits of power station c in time period t respectively;

[0077] When the reservoir comprehensively considers the output constraints of the priority power generation plan, the modeling reservoir priority power generation plan output constraints

[0078]

[0079] In the preferred solution, S106, based on LightGBM machine learning prediction, includes:

[0080] S1061, constructing a feature vector model;

[0081]

[0082] The day-ahead market settlement electricity price for period t is predicted; W t is the predicted amount of incoming water; L t is the water level control target of the reservoir, U t In order to include comprehensive utilization requirements for power generation or flood control, P j,t The electricity price forecast value of the relevant node j in time period t; H t is the holiday indicator variable, H t =1 means holidays, H t =0 means non-holiday; T t For temperature or other environmental factors, D t To include demand-side management factors such as changes in market demand;

[0083] S1062. Collect historical data, including the predicted amount of water inflow W t , the reservoir water level control target L t , comprehensive utilization requirements U t , the predicted value of electricity price of relevant nodes P j,t , Holiday Information H t , temperature Tt and demand-side management factors D t ;

[0084] S1063: Construct a feature vector based on the above data

[0085] S1064. Use the LightGBM algorithm to train historical data and optimize model parameters to minimize prediction error;

[0086] S1065, use the trained model to analyze the new feature vector Make a prediction and get the predicted electricity price.

[0087] In a preferred solution, S107, prediction based on XGBoost machine learning, includes:

[0088] S1071. Construct a feature vector model;

[0089]

[0090] The day-ahead market settlement electricity price for period t is predicted; W t is the predicted amount of incoming water; L t is the water level control target of the reservoir, U t In order to include comprehensive utilization requirements for power generation or flood control, P j,t The electricity price forecast value of the relevant node j in time period t; H t is the holiday indicator variable, H t =1 means holidays, H t =0 means non-holiday; T t For temperature or other environmental factors, D t To include demand-side management factors such as changes in market demand;

[0091] S1072. Collect historical data, including the predicted amount of water W t , the reservoir water level control target L t , comprehensive utilization requirements U t , the predicted value of electricity price of relevant nodes P j,t , Holiday Information H t , temperature T t and demand-side management factors D t ;

[0092] S1073: Construct a feature vector based on the above data

[0093] S1074. Use the XGBoost algorithm to train historical data and optimize model parameters to optimize model performance.

[0094] S1075. Evaluate the accuracy of the model through cross-validation and make adjustments based on actual conditions;

[0095] S1076. Input the data of the new time period into the trained model to obtain the predicted electricity price.

[0096] In the preferred solution, in step S2, the optimal bidding decision model is based on the following formula:

[0097] (34);

[0099] Where, is the optimal output value achieved by trading unit c through quotation, is the time period deviation between the optimal output and the optimal simulated output, M delta is the output deviation penalty factor.

[0100] In the preferred solution, S201, the optimal bid decision model is also subject to the constraints of the transaction unit declaration format, including:

[0101] S2011, Trading unit declared output constraints;

[0102] Initial declared capacity and minimum technical output P c,min The sum of the declared capacities of each segment is equal to the maximum permissible output of the trading unit;

[0103]

[0104] Where, Declare capacity for quote segment k of trading unit c, It is the rated capacity of the trading unit;

[0105]

[0106] Where, The output is converted from the ecological discharge flow of reservoir r, that is, one section is declared according to the output converted from the ecological discharge flow. If it is limited by the maximum declared flow of a single section, the number of sections is increased accordingly;

[0107] S2012, minimum declared capacity constraints;

[0108] The length of each segment shall not be less than n% of the difference between the rated active power of the trading unit and the minimum stable technical output (which can be flexibly set according to local market rules);

[0109]

[0110] S2013, maximum declared capacity constraints;

[0111] The length of each segment shall not exceed the maximum declared capacity per segment specified in the spot rules:

[0112]

[0113] Where, is the maximum declared capacity per segment specified in the spot rules.

[0114] S2014, price constraints for trading unit declarations;

[0115] The declared price of each segment shall not be lower than the declared price of the previous segment, and all segment quotes shall be within the allowable quote range;

[0116] Y min ≤Y c,k ≤Y max (39);

[0117] Y c,k -Y c,k-1 ≥Y ε (40);

[0118] Where, Y c,k is the declared price of the k-th segment of trading unit c, Y max 、Y min are the upper and lower limits of the declared price of market parameters; when the lower limit of the declared price of market parameters is less than 0, Y min takes max{lower limit of declared price of market parameters, lower limit of declared electricity price set by user}, Y ε is the minimum quote increment. The quote for the output segment corresponding to the ecological discharge flow is Y min .

[0119] In the preferred solution, through ex-post verification and correction, the starting and ending outputs corresponding to the quote segments are adjusted to integer multiples of 10,000 kW.

[0120] In the preferred solution, S202, the optimal quote decision model is also subject to the constraint of the relationship between the expected output and the segmented quotes;

[0121] If the reference price is greater than or equal to the segmented quote, the capacity of this segment wins the bid; if the reference price is lower than the segmented quote, the capacity of this segment does not win the bid.

[0122] If LMP<00000​​​​​​​​​​​​​​​​​, then the winning bid is up to the adjustable upper limit;

[0125]

[0126] Where, LMP c,t is the predicted power station node electricity price, Y c,k is the declared price of quotation segment k of transaction unit c, Indicator variable representing the relationship between the quote and the reference price in segment k. When the quotation is greater than the reference price

[0127] Contribute to the quotation strategy expectations, It is the difference between the installed capacity of the trading unit and the maximum power generation capacity.

[0128] In the preferred solution, S203, the optimal bidding decision model is also subject to the expected output and optimal output constraints;

[0129]

[0130] Where, is the deviation between the expected output and the optimal output of the quotation strategy of trading unit c.

[0131] In the preferred solution, S204, the optimal bidding decision model is also subject to (4) the constraint that the declared capacity avoids the vibration zone; σ i,t,s =1, indicating that the highest gear vibration zone avoided by hydropower unit i in time period t is s, and the output of unit i in time period t is above the upper limit of the vibration zone s and below the lower limit of the vibration zone s+1:

[0132] ∑σ i,t,s ≤1 (45);

[0133]

[0134]

[0135] Where S is the number of vibration sections, They are the lower and upper limits of the output of unit i in vibration zone s respectively.

[0136] The present invention provides a method for self-adapting time-scale hydropower spot dual-level quantity and price declaration based on complex constraints. Compared with the existing technology, the beneficial effects are as follows:

[0137] The present invention constructs a spot market electricity price prediction model, and studies the transaction strategy and transaction model of hydropower stations participating in the spot market based on the spot electricity price prediction results. While improving the market-oriented transaction income of hydropower stations, it can effectively control the transaction decision-making risks, achieve the market income-risk balance management of hydropower stations and meet the comprehensive utilization constraint requirements of power stations, effectively support the daily spot transaction declaration work of hydropower stations, provide business process sorting and scientific and reasonable theoretical method guidance for market transaction decision-making of large-scale hydropower station power generation projects, and help accelerate the improvement of the market-oriented operation mechanism of hydropower stations. DETAILED DESCRIPTION

[0138] As the construction of the electricity spot market continues to accelerate, it is imperative for hydropower to participate in spot market transactions. However, the market is still in its infancy and there are certain risks in hydropower participating in the market.

[0139] First, the transaction process for hydropower stations participating in the spot market is complicated, which may lead to trading losses due to insufficient understanding or improper operation;

[0140] Second, after participating in electricity spot trading, the uncertainty of the hydropower station's transaction volume and transaction price increases, which directly affects the power generation revenue of the station;

[0141] Third, for large hydropower stations that undertake comprehensive utilization functions, spot market transactions may not be able to meet the comprehensive utilization requirements of the power stations.

[0142] Because large hydropower stations also undertake multiple social functions such as flood control and shipping, and transmit electricity to multiple provinces and cities through dedicated cross-provincial and cross-regional supporting lines using a point-to-network approach, their trading decisions face complex constraints. This paper fully considers multiple complex constraints such as large hydropower station operation, vibration zones, ramping, scheduling, start-up and shutdown, reservoir capacity, and water level discharge. Using a time period (e.g., 15 minutes) as the minimum unit, the paper uses a custom optimized time length to calculate the optimal expected output of the hydropower station. Based on the optimal predicted hydropower output, the paper further optimizes the current bidding strategy for hydropower units in the multi-timescale spot market with the goal of maximizing returns, thus constructing a two-stage model for the optimal output and optimal bidding strategy of a hydropower station under complex constraints, with adaptive time scales.

[0143] Throughout the decision-making process, electricity price forecasts provide key market signals for decision-making. The decision-making model uses these forecast results, combined with the operating characteristics of the power generation system and market supply and demand information, to formulate the optimal application strategy and output plan. The two are coupled and jointly influence the decision-making results. For electricity price forecasting, the LightGBM and XGBoost machine learning prediction methods are used. Based on the market boundary data published by the spot market system, structural features and holiday data are processed to achieve forecasts of electricity prices for the entire network and key nodes, providing a decision-making basis for participating in the spot market. Regarding the decision-making model, based on information such as water inflow forecasts, water level control targets, comprehensive utilization requirements, and electricity price forecasts for relevant nodes, a coupling relationship between the hydropower system and the power system is constructed. As a price acceptor, the spot clearing and settlement process is simulated. With the goal of maximizing the overall benefits of the spot market, the hydropower day-ahead electricity market application quantity, price, and output plan curve are determined. In the model, hydropower plants are used as trading units (power stations) to participate in market declarations, and hydropower units are used as dispatching units for real-time dispatching. The mode of participating in spot market clearing is: participating in clearing as conventional hydropower units, with power constraints as the main constraint, while supporting the consideration of relevant constraints such as water level and downstream flow restrictions. The advantage of this decision model is that it requires less information. It only requires market supply and demand information, the spot electricity price and operating characteristics of the power generation enterprise itself, and does not require understanding of the grid structure and the bidding behavior of other market players. It has good applicability in the early stages of the power market. In addition, the method of the present invention can effectively respond to the challenge of information asymmetry, optimize the bidding strategy by focusing on market price forecasts, and thus maximize profits. The specific process steps of the entire decision model are as follows:

[0144] Example 1:

[0145] A method for self-adaptive time-scale hydropower spot double-layer quantity and price declaration based on complex constraints, comprising the following steps:

[0146] S1. Construct a hydropower price declaration model, which includes two stages. The first stage is to determine the optimal expected output curve with the goal of maximizing revenue;

[0147] S2, the second stage is to optimize the optimal quotation curve with the goal of achieving the optimal expected curve;

[0148] The above steps can be used to declare the spot quantity and price of hydropower.

[0149] The technical advantages of this invention are: 1. It has been tested and can be applied to regional spot markets. It can handle scenarios with numerous nodes (over 5,600 nodes) and complex boundary conditions and clearing mechanisms. This invention also incorporates scalability elements, allowing it to be applied to large hydropower stations participating in other spot markets.

[0150] 2. The present invention is the first to successfully test a cross-provincial, cross-regional, point-to-multi-network power transmission hydropower station that can quote quantities and participate in spot transactions. In the early stage, only thermal power and new energy quoted quantities and participated in spot transactions. The regional spot market included hydropower for the first time. The present invention is innovative in terms of hydropower participation in spot transactions. In addition to power generation, large hydropower stations also undertake multiple social functions such as flood control and shipping. They need to transmit electricity across regions and provinces, and transmit electricity to multiple provinces. Their trading strategies are more complex. For example, in the process of proportional power transmission from point to multiple networks, electricity price forecasts need to take into account multiple receiving markets, and the cross-provincial transmission fees are different from those of network-to-network power stations, which increases the complexity of market clearing and the difficulty of understanding transaction results. At the same time, the physical execution of cross-provincial priority plans adds more constraints.

[0151] 3. This invention has been tested in actual spot trading and demonstrates strong practicality. Applied to a regional spot market in 2024, the system was applied to five settlement trials totaling 53 days of spot trading. Through multiple scenarios, including dry season, dry-flood transition period, flood season, multi-unit and transmission line maintenance, and limited power station regulation capacity, it achieved 10.52 billion kWh of grid-connected electricity.

[0152] The technical advantage of this invention lies in the fact that hydropower carries out multiple social functions, such as flood control and shipping, and the unique nature of corporate power plants participating in the spot market allows for a greater range of constraints to be considered when making trading bids. Small hydropower plants abroad can participate in the spot market solely for economic reasons, generating electricity when prices are high and refraining from generating electricity or abandoning water when prices are low. This is a purely market-based behavior. However, large hydropower plants participating in the spot market must consider more than just market and economic interests; their social functions must also be considered. Based on this proposed two-stage model, the first stage primarily considers electricity price forecasts and various complex constraints on reservoirs and power plants, optimizing the optimal executable curve for the power plant. The second stage then derives a tradable price-volume curve based on market trading rules, such as bid prices and capacity requirements. This ensures that the power plant's generation curve, derived from this price-volume curve, closely matches the optimal executable curve, maximizing both power plant revenue and social benefits. Without adopting the two-stage price-volume model, directly applying the principle of profit maximization and outputting a bid-for-volume curve based on various boundary conditions can lead to conflicts within certain power plant constraints, posing risks to the operational safety of the power plant and reservoir.

[0153] In the preferred solution, in step S1, the profit maximization decision model is based on the following formula:

[0154]

[0155] Formula: Maximize medium- and long-term, as well as spot, revenues based on electricity price forecasts, taking into account plant power consumption - Deviation between actual optimized daily power generation and daily power generation based on inflow * Penalty Factor - Abandoned flow * Penalty Factor - Indicator variable for inconsistent output adjustments between power stations on the left and right banks of the same reservoir * Penalty Factor. While maximizing power generation revenue based on forecasted electricity price trends, the reservoir's social functions must also be met. These requirements include matching power generation with inflow to avoid the risk of exceeding water level and flow limits at the hydropower station; avoiding abandonment; and ensuring relative operational balance between power stations on the left and right banks of the reservoir, meeting unit maintenance requirements, and mitigating operational risks. If these factors occur, a significant penalty will be imposed, negatively impacting revenues. Therefore, the objective function will decrease, and the system will automatically seek the maximum value to avoid such situations. This maximizes economic and social benefits.

[0156] Where, is the winning bid power of trading unit c in the day-ahead market during period t; c is the power consumption rate of transaction unit c; Decompose the power of the contract for trading unit c in period t; Predict the day-ahead market settlement electricity price for time period t; M is the deviation between the current bid electricity and the recommended plan, Qdiff is the penalty factor; is the water discharge of reservoir r in period t, M Abq Penalty constraint for abandoned water flow; M is a sliding variable whose adjustment direction is not strictly consistent with the same library. div It is a penalty factor for inconsistent adjustment direction;

[0157] Penalty factor setting priority:

[0158]

[0159] The optimization time period can be set to a single day or multiple days.

[0160] In the preferred solution, S101, the profit maximization decision model is also constrained by the relationship between reservoirs, trading units, and dispatching units:

[0161] That is, the output of a trading unit in each period is the sum of the outputs of the dispatching units under the trading unit in the corresponding period; the output of a reservoir in each period is the sum of the outputs of the trading units under the reservoir in the corresponding period;

[0162]

[0163] Where, is the expected standard output of unit i in period t, Mark the expected force for trading unit c in period t; Denote the expected force on reservoir r during period t.

[0164] In the preferred solution, S102, the profit maximization decision model is also subject to the technical characteristics of the dispatch unit, including:

[0165] S1021, stable operating range constraints, i.e. the expected winning bid output should be within the allowable output range;

[0166]

[0167] Where, α i,t is the startup status of unit i in period t, The maximum and minimum output limits of unit i in time period t, where the minimum output limit is the upper limit of the unit vibration zone;

[0168] S1022, avoid the constraints of multiple vibration zones,

[0169] σ i,t,s =1 (6);

[0170] The highest gear vibration zone avoided by hydropower unit i in time period t is s, and the output of unit i in time period t is above the upper limit of the vibration zone s and below the lower limit of the vibration zone s+1:

[0171] ∑σ i,t,s ≤1 (7);

[0172]

[0173]

[0174] Where S is the number of vibration sections, are the lower and upper limits of the output of unit i in vibration zone s, respectively;

[0175] If the starting point of the first vibration zone is not 0, and the output falls before the first vibration zone, all sigmas are 0, and the output is less than the starting point of the first vibration zone. If the output falls after the first vibration zone and does not fall after the last vibration zone, the third of the three terms on the right side of the formula is 0, and the second term is Cancelling the first term, we only have Therefore, the upper limit of output is the upper limit of output in the next vibration zone. If the output falls after the last vibration zone, the second of the three terms on the right side of the formula is 0, and the third term is After canceling out the first term, the output limit is

[0176] S1023, ramp rate constraints due to shipping reasons;

[0177] S10231. The unit output change during each period should be within the allowable range of the ramp rate;

[0178]

[0179] Where, is the regulation rate of unit i, β i,t is the indicator variable for the direction of increase or decrease of unit i in period t, β i,t =1 means the period from t-1 to t is upward adjustment, β i,t =0 means the period from t-1 to t is downward adjustment;

[0180] S10232, the output variation of the trading unit, that is, the output variation of each time period should be within the allowable range of the power station ramp rate;

[0181]

[0182] Where, The adjustment rate of transaction unit c is determined according to the output amplitude of dispatch and the downstream flow of shipping, and has the function of connecting with the last period of the previous day;

[0183] S10232. Reservoir output variation, that is, the output variation in different time periods should be within the allowable range of the reservoir ramp rate;

[0184]

[0185] Where, is the regulation rate of reservoir r, and its value is determined according to the dispatching output amplitude and shipping discharge, and has the function of connecting with the last period of the previous day.

[0186] S1024, start-stop peak load regulation frequency constraint;

[0187] Statistical output 0->1, that is, the number of start-stop conversions, which must not exceed the set value;

[0188] u i,t -z i,t =α i,t -α i,t-1 (13);

[0189]

[0190] Where u i,t 、z i,t is the state variable of the startup and shutdown actions of unit i in period t, The upper limit of the number of start and stop switches of unit i in a single day.

[0191] In the preferred solution, S103, the profit maximization decision model is also subject to the hydraulic characteristics of the reservoir, including:

[0192] S1031, reservoir capacity balance constraint;

[0193]

[0194] Where V r,t is the storage capacity of reservoir r in time period t, in m 3 ; is the inflow of reservoir r in period t, is the power generation flow of reservoir r in period t, is the water discharge of reservoir r in time period t, in m 3 / s;η c,t is the water consumption rate of transaction unit c in time period t, in m 3 / kWh;

[0195] S1032, water level range constraint;

[0196]

[0197] Where, are the upper and lower limits of the storage capacity of reservoir r in time period t, respectively, in m 3 , obtained by converting the upper and lower limits of the water level according to the water level storage capacity curve;

[0198] S1033, ecological discharge flow range constraints;

[0199]

[0200] Where, The maximum discharge flow and the minimum ecological discharge flow of reservoir r meet the comprehensive water demand of reservoir operation. The leftmost side of the formula is the output, and the rightmost side is the flow. The input value is the output corresponding to the ecological discharge flow. The default value is the maximum value. The input output is converted to flow.

[0201] In the preferred solution, S104, the profit maximization decision model is also subject to power plant regulation demand constraints, including:

[0202] S1041, power station total power generation ratio constraint;

[0203]

[0204] Where, c,max 、υ c,min are the upper and lower limits of the proportion of power generation of power station c to reservoir r;

[0205] S1042, the power station daily output variation range is constrained, that is, the difference between the maximum and minimum output within the day is less than the set value:

[0206]

[0207] Where, The maximum and minimum output of the power station on that day, The maximum acceptable value of the daily maximum and minimum output fluctuations;

[0208] S1043, the reservoir daily output variation range is constrained, that is, the difference between the maximum and minimum output within a day is less than the set value:

[0209]

[0210] Where, The maximum and minimum output of the reservoir on that day, It is the maximum acceptable value of the daily maximum and minimum output fluctuations, and has the function of connecting with the last moment of the previous day;

[0211] S1043, one-way adjustment of Tongku Hydropower Plant;

[0212]

[0213] Where, β c＝1,t , β c＝2,t It is the indicator variable of the upward and downward regulation direction of the hydropower station pair in the same reservoir; It is a sliding variable whose adjustment direction is not strictly consistent with the same library;

[0214] S1044, the allowable range of deviation between the reservoir's recently awarded electricity quantity and the recommended scheduling plan;

[0215]

[0216] Where, is the hourly average output plan value recommended for reservoir r's cascade regulation, and They are the upper limits of the allowable range of positive and reverse deviations between the day's winning bid electricity and the planned value, respectively.

[0217] In the preferred solution, S105, the profit maximization decision model is also subject to power plant output constraints, including:

[0218] S1051. Control requires that the power station output be constrained throughout the day, i.e. the power station output be within the upper and lower limits of the daily output:

[0219]

[0220] Where, and They are the lower and upper limits of the power output of power station c respectively; they are mainly used to determine the power station output corresponding to the ecological discharge flow range of the reservoir;

[0221] S1052, power station priority power generation plan output constraints;

[0222]

[0223] Where, and are the lower and upper output limits of power station c in time period t, respectively; they are mainly used for the output range of power station priority power generation plan;

[0224] When the reservoir comprehensively considers the output constraints of the priority power generation plan, the modeling reservoir priority power generation plan output constraints

[0225]

[0226] In the preferred solution, S106, based on LightGBM machine learning prediction, includes:

[0227] S1061, constructing a feature vector model;

[0228]

[0229] The day-ahead market settlement electricity price for period t is predicted; W t is the predicted amount of incoming water; L t is the water level control target of the reservoir, U t In order to include comprehensive utilization requirements for power generation or flood control, P j,t The electricity price forecast value of the relevant node j in time period t; H t is the holiday indicator variable, H t =1 means holidays, H t =0 means non-holiday; T t For temperature or other environmental factors, D t To include demand-side management factors such as changes in market demand;

[0230] S1062. Collect historical data, including the predicted amount of water inflow W t , the reservoir water level control target L t , comprehensive utilization requirements U t , the predicted value of electricity price of relevant nodes P j,t , holiday information Ht, temperature Tt and demand-side management factors Dt;

[0231] S1063: Construct a feature vector based on the above data

[0232] S1064. Use the LightGBM algorithm to train historical data and optimize model parameters to minimize prediction error;

[0233] S1065, use the trained model to analyze the new feature vector Make a prediction and get the predicted electricity price.

[0234] In a preferred solution, S107, prediction based on XGBoost machine learning, includes:

[0235] S1071. Construct a feature vector model;

[0236]

[0237] The day-ahead market settlement electricity price for period t is predicted; W t is the predicted amount of incoming water; L t is the water level control target of the reservoir, U t In order to include comprehensive utilization requirements for power generation or flood control, P j,t The electricity price forecast value of the relevant node j in time period t; H t is the holiday indicator variable, H t =1 means holidays, H t =0 means non-holiday; T t For temperature or other environmental factors, D t To include demand-side management factors such as changes in market demand;

[0238] S1072. Collect historical data, including the predicted amount of water W t , the reservoir water level control target L t , comprehensive utilization requirements U t , the predicted value of electricity price of relevant nodes P j,t , holiday information Ht, temperature Tt and demand-side management factors Dt;

[0239] S1073: Construct a feature vector based on the above data

[0240] S1074. Use the XGBoost algorithm to train historical data and optimize model parameters to optimize model performance.

[0241] S1075. Evaluate the accuracy of the model through cross-validation and make adjustments based on actual conditions;

[0242] S1076. Input the data of the new time period into the trained model to obtain the predicted electricity price.

[0243] The development of the electricity spot market has introduced increased uncertainty for power generation companies, whose profits fluctuate significantly due to varying spot trading strategies. As a key factor in determining market supply and demand, predictive spot electricity prices are theoretically significantly more accurate than medium- and long-term market prices, which are dominated by financial game theory. Therefore, accurately forecasting spot electricity prices plays a crucial role in maximizing spot profits and social benefits. Spot electricity prices are highly volatile and are influenced by numerous non-deterministic factors, such as system supply and demand conditions, fuel price instability, hydropower station volatility, branch line disconnections, transmission congestion, market participant behavior, generator bidding strategies, and system and unit constraints. Currently, there are two main methods for domestic electricity price forecasting: market simulation and intelligent forecasting, including time series forecasting, artificial neural network forecasting, combined forecasting, statistical model forecasting, and machine learning forecasting. The present invention employs machine learning forecasting to predict electricity prices, primarily due to its advantages in processing complex, nonlinear data. Electricity prices are influenced by a variety of factors, including weather, electricity load, and seasonal variations. The relationships between these factors are often highly nonlinear. Traditional time series models perform poorly with nonlinear data, while machine learning methods can better capture the complex nonlinear relationships within the data, thereby improving forecast accuracy and generalization. Furthermore, compared to traditional statistical models, machine learning models are more effective at handling high-dimensional data and can automatically learn complex relationships between data without requiring specific mathematical formulas. While artificial neural networks also have the ability to handle nonlinear data, training deep neural networks requires large amounts of data and computing resources and can be more prone to overfitting. In contrast, tree models used in machine learning forecasting methods are better able to maintain high forecast accuracy while mitigating the risk of overfitting. While combined forecasting can overcome the shortcomings of single models, it often requires extensive debugging and experience in model combination and parameter selection. Machine learning methods, on the other hand, can effectively combine multiple decision trees through ensemble learning, achieving similar combined forecasting results while being more efficient in model training and tuning. Therefore, machine learning-based forecasting methods offer advantages in meeting the complexity and accuracy requirements of electricity price forecasting. In the further optimized solution, the two prediction models are combined into an adversarial network, which is iterated and optimized through weighted adjustment to further improve the prediction accuracy.

[0244] In the preferred solution, in step S2, the optimal bidding decision model is based on the following formula:

[0245]

[0246] The output curve achieved based on the electricity price forecast and taking into account plant power consumption is calculated as follows: The maximum profit is calculated by subtracting the time period deviation between the optimal output and the optimal simulated output value * the penalty factor. The optimal bidding decision model primarily considers maximizing achievable profits. The output curve cleared by the bid ultimately maximizes company profits, but it must be as close as possible to the optimal curve optimized by the profit maximization decision model. This comprehensively considers social benefits and social benefits. Any deviation is penalized, negatively impacting profits. The model thus seeks a feasible solution that is closest to the optimal curve, achieving the stated goal of the profit maximization decision model.

[0247] Where, is the optimal output value achieved by trading unit c through quotation, is the time period deviation between the optimal output and the optimal simulated output, M delta is the output deviation penalty factor.

[0248] In the preferred solution, S201, the optimal bid decision model is also subject to the constraints of the transaction unit declaration format, including:

[0249] S2011, Trading unit declared output constraints;

[0250] Initial declared capacity and minimum technical output P c,min The sum of the declared capacities of each segment is equal to the maximum permissible output of the trading unit;

[0251]

[0252] Where, Declare capacity for quote segment k of trading unit c, It is the rated capacity of the trading unit;

[0253]

[0254] Where, The output is converted from the ecological discharge flow of reservoir r, that is, one section is declared according to the output converted from the ecological discharge flow. If it is limited by the maximum declared flow of a single section, the number of sections is increased accordingly;

[0255] S2012, minimum declared capacity constraints;

[0256] The length of each segment shall not be less than n% of the difference between the rated active power of the trading unit and the minimum stable technical output (which can be flexibly set according to local market rules);

[0257]

[0258] S2013, maximum declared capacity constraints;

[0259] The length of each segment shall not exceed the maximum declared capacity per segment specified in the spot rules:

[0260]

[0261] Wherein, is the maximum declared capacity per segment specified in the spot rules.

[0262] S2014, price constraints for trading unit declarations;

[0263] The declared price of each segment shall not be lower than the declared price of the previous segment, and all segment quotes shall be within the allowable range of quotes;

[0264] Y min ≤Y c,k ≤Y max (39);

[0265] Y c,k -Y c,k-1 ≥Y ε (40);

[0266] Wherein, Y c,k is the declared price of the k-th segment of trading unit c, Y max 、Y min are the upper and lower limits of the declared price of market parameters; when the lower limit of the declared price of market parameters is less than 0, Y min takes max{lower limit of declared price of market parameters, lower limit of declared electricity price set by user}, Y ε is the minimum quote increment. The quote for the output segment corresponding to the ecological discharge flow is Y min .

[0267] In the preferred solution, through ex-post verification and correction, the starting and ending output corresponding to the quote segment is adjusted to an integer multiple of 10,000 kW.

[0268] In the preferred solution, S202, the optimal quote decision model is also subject to the constraint of the relationship between the expected output and the segmented quote;

[0269] If the reference price is greater than or equal to the segmented quote, the capacity of this segment wins the bid; if the reference price is lower than the segmented quote, the capacity of this segment does not win the bid.

[0270] If LMP c,t <Y1, the trading unit output is the minimum technical output

[0271] If Y n <LMP c,t ≤Y n+1 , then the n-th segment is full;

[0272] If LMP c,t >Y N, then the winning bid is up to the adjustable upper limit;

[0273]

[0274] Where, LMP c,t is the predicted power station node electricity price, Y c,k is the declared price of quotation segment k of transaction unit c, Indicator variable representing the relationship between the quote and the reference price in segment k. When the quotation is greater than the reference price

[0275] Contribute to the quotation strategy expectations, It is the difference between the installed capacity of the transaction unit and the maximum power generation capacity, which is the power range that cannot be achieved even if the last bid is successful.

[0276] In the preferred solution, S203, the optimal bidding decision model is also subject to the expected output and optimal output constraints;

[0277]

[0278] Where, is the deviation between the expected output and the optimal output of trading unit c's bidding strategy. Regulation requires that the lower limit of the power plant's daily output be reflected in the capacity of the first bid segment. The bid price for this segment defaults to the bid lower limit.

[0279] In the preferred solution, S204, the optimal bidding decision model is also subject to the constraint that the declared capacity avoids the vibration zone; σ i,t,s =1, indicating that the highest gear vibration zone avoided by hydropower unit i in time period t is s, and the output of unit i in time period t is above the upper limit of the vibration zone s and below the lower limit of the vibration zone s+1:

[0280] ∑σ i,t,s ≤1 (45);

[0281]

[0282] Where S is the number of vibration sections, They are the lower and upper limits of the output of unit i in vibration zone s respectively.

[0283] If the starting point of the first vibration zone is not 0, and the output falls before the first vibration zone, all sigmas are 0, and the output is less than the starting point of the first vibration zone. If the output falls after the first vibration zone and does not fall after the last vibration zone, the third of the three terms on the right side of the formula is 0, and the second term is Cancelling the first term, we only have Therefore, the upper limit of output is the upper limit of output in the next vibration zone. If the output falls after the last vibration zone, the second of the three terms on the right side of the formula is 0, and the third term is After canceling out the first term, the output limit is

[0284] The above embodiments are merely preferred technical solutions of the present invention and should not be construed as limiting the present invention. The embodiments and features in the embodiments of this application may be arbitrarily combined with each other unless they conflict. The scope of protection of the present invention shall be the technical solutions described in the claims, including equivalent alternatives to the technical features of the technical solutions described in the claims. Equivalent alternatives and improvements within this scope are also within the scope of protection of the present invention.

Claims

1. A method for double-layer quantity and price declaration of hydropower spot market based on adaptive time scale and complex constraints, characterized by: The following steps are involved: S1. Construct a hydropower price declaration model, which includes two stages. The first stage is to determine the optimal expected output curve with the goal of maximizing revenue; S2, the second stage is to optimize the optimal quotation curve with the goal of achieving the optimal expected curve; The above steps can be used to declare the spot quantity and price of hydropower.

2. The complex constraint-based adaptive time-scale hydropower spot dual-level quantity and price declaration method according to claim 1 is characterized in that in step S1, the profit maximization decision model is based on the following formula: Where, is the winning bid power of trading unit c in the day-ahead market during period t; c is the power consumption rate of transaction unit c; Decompose the power of the contract for trading unit c in period t; Predict the day-ahead market settlement electricity price for time period t; M is the deviation between the current bid electricity and the recommended plan, Qdiff is the penalty factor; is the water discharge of reservoir r in period t, M Abq Penalty constraint for abandoned water flow; M is a sliding variable whose adjustment direction is not strictly consistent with the same library. div It is a penalty factor for inconsistent adjustment direction; Penalty factor setting priority: The optimization time period can be set to a single day or multiple days.

3. The complex constraint-based adaptive time-scale hydropower spot dual-level quantity and price declaration method according to claim 2 is characterized by: S101. The profit maximization decision model is also constrained by the relationship between reservoirs, trading units, and dispatching units: That is, the output of a trading unit in each period is the sum of the outputs of the dispatching units under the trading unit in the corresponding period; the output of a reservoir in each period is the sum of the outputs of the trading units under the reservoir in the corresponding period; Where, is the expected standard output of unit i in period t, Mark the expected force for trading unit c in period t; Denote the expected force on reservoir r during period t.

4. The complex constraint-based adaptive time-scale hydropower spot dual-level quantity and price declaration method according to claim 2 is characterized by: S102. The profit maximization decision model is also subject to the technical characteristics of the dispatch unit, including: S1021, stable operating range constraints, i.e. the expected winning bid output should be within the allowable output range; Where, α i,t is the startup status of unit i in period t, The maximum and minimum output limits of unit i in time period t, where the minimum output limit is the upper limit of the unit vibration zone; S1022, avoiding multiple vibration zone constraints; s i,t,s =1 (6); The highest gear vibration zone avoided by hydropower unit i in time period t is s, and the output of unit i in time period t is above the upper limit of the vibration zone s and below the lower limit of the vibration zone s+1: ∑σ i,t,s ≤1 (7); Where S is the number of vibration sections, are the lower and upper limits of the output of unit i in vibration zone s, respectively; S1023, ramp rate constraints due to unit, power station or shipping reasons; S10231. The unit output change during each period should be within the allowable range of the ramp rate; Where, is the regulation rate of unit i, β i,t is the indicator variable for the direction of increase or decrease of unit i in period t, β i,t =1 means the period from t-1 to t is upward adjustment, β i,t =0 means the period from t-1 to t is downward adjustment; S10232, the output variation of the trading unit, that is, the output variation of each time period should be within the allowable range of the power station ramp rate; Where, The adjustment rate of transaction unit c is determined according to the output amplitude of dispatch and the downstream flow of shipping, and has the function of connecting with the last period of the previous day; S10232. Reservoir output variation, that is, the output variation in different time periods should be within the allowable range of the reservoir ramp rate; Where, is the regulation rate of reservoir r, which is determined by the output amplitude of the dispatching and the downstream flow of shipping, and has the function of connecting with the last period of the previous day; S1024, start-stop peak load regulation frequency constraint; Statistical output 0->1, that is, the number of start-stop conversions, which must not exceed the set value; you i,t -z i,t =a i,t -a i,t-1 (13); Where u i,t 、z i,t is the state variable of the startup and shutdown actions of unit i in period t, The upper limit of the number of start and stop switches of unit i in a single day.

5. The complex constraint-based adaptive time-scale hydropower spot dual-level quantity and price declaration method according to claim 2 is characterized by: S103. The profit maximization decision model is also subject to the hydraulic characteristics of the reservoir, including: S1031, reservoir capacity balance constraint; Where V r,t is the storage capacity of reservoir r in time period t, in m 3 ; is the inflow of reservoir r in period t, is the power generation flow of reservoir r in period t, is the water discharge of reservoir r in time period t, in m 3 / s;η c,t is the water consumption rate of transaction unit c in time period t, in m 3 / kWh; S1032, water level range constraint; Where, are the upper and lower limits of the storage capacity of reservoir r in time period t, respectively, in m 3 , obtained by converting the upper and lower limits of the water level according to the water level storage capacity curve; S1033, ecological discharge flow range constraints; Where, The maximum discharge flow and the minimum ecological discharge flow of reservoir r are used to meet the comprehensive water demand of reservoir operation.

6. The complex constraint-based adaptive time-scale hydropower spot dual-level quantity and price declaration method according to claim 2 is characterized by: S104. The profit maximization decision model is also subject to power plant regulation requirements, including: S1041, power station total power generation ratio constraint; Where, c,max 、υ c,min are the upper and lower limits of the proportion of power generation of power station c to reservoir r; S1042, the power station daily output variation range is constrained, that is, the difference between the maximum and minimum output within the day is less than the set value: Where, The maximum and minimum output of the power station on that day, The maximum acceptable value of the daily maximum and minimum output fluctuations; S1043, the reservoir daily output variation range is constrained, that is, the difference between the maximum and minimum output within a day is less than the set value: Where, The maximum and minimum output of the reservoir on that day, It is the maximum acceptable value of the daily maximum and minimum output fluctuations, and has the function of connecting with the last moment of the previous day; S1043, one-way adjustment of Tongku Hydropower Plant; Where, β c＝1,t , β c＝2,t It is the indicator variable of the upward and downward regulation direction of the hydropower station pair in the same reservoir; It is a sliding variable whose adjustment direction is not strictly consistent with the same library; S1044, the allowable range of deviation between the reservoir's recently awarded electricity quantity and the recommended scheduling plan; Where, is the hourly average output plan value recommended for reservoir r's cascade regulation, and They are the upper limits of the allowable range of positive and reverse deviations between the day's winning bid electricity and the planned value, respectively.

7. The complex constraint-based adaptive time-scale hydropower spot dual-level quantity and price declaration method according to claim 2 is characterized by: S105. The profit maximization decision model is also subject to power plant output constraints, including: S1051. Control requires that the power station output be constrained throughout the day, i.e. the power station output be within the upper and lower limits of the daily output: Where, and are the lower and upper output limits of power station c respectively; S1052, power station priority power generation plan output constraints; Where, and are the lower and upper output limits of power station c in time period t respectively; When the reservoir comprehensively considers the output constraints of the priority power generation plan, the output constraints of the reservoir priority power generation plan are modeled:

8. The method for self-adaptive time-scale hydropower spot double-layer quantity and price declaration based on complex constraints according to claim 2 is characterized by: S106, based on LightGBM machine learning prediction, including: S1061, constructing a feature vector model; The day-ahead market settlement electricity price for period t is predicted; W t is the predicted amount of incoming water; L t is the water level control target of the reservoir, U t In order to include comprehensive utilization requirements for power generation or flood control, P j,t The electricity price forecast value of the relevant node j in time period t; H t is the holiday indicator variable, H t =1 means holidays, H t =0 means non-holiday; T t For temperature or other environmental factors, D t To include demand-side management factors such as changes in market demand; S1062. Collect historical data, including the predicted amount of water inflow W t , the reservoir water level control target L t , comprehensive utilization requirements U t , the predicted value of electricity price of relevant nodes P j,t , Holiday Information H t , temperature T t and demand-side management factors D t ; S1063: Construct a feature vector based on the above data S1064. Use the LightGBM algorithm to train historical data and optimize model parameters to minimize prediction error; S1065, use the trained model to analyze the new feature vector Make a prediction and get the predicted electricity price.

9. The method for self-adaptive time-scale hydropower spot double-layer quantity and price declaration based on complex constraints according to claim 2 is characterized by: S107, based on XGBoost machine learning prediction, including: S1071. Construct a feature vector model; The day-ahead market settlement electricity price for period t is predicted; W t is the predicted amount of incoming water; L t is the water level control target of the reservoir, U t In order to include comprehensive utilization requirements for power generation or flood control, P j,t The electricity price forecast value of the relevant node j in time period t; H t is the holiday indicator variable, H t =1 means holidays, H t =0 means non-holiday; T t For temperature or other environmental factors, D t To include demand-side management factors such as changes in market demand; S1072. Collect historical data, including the predicted amount of water W t , the reservoir water level control target L t , comprehensive utilization requirements U t , the predicted value of electricity price of relevant nodes P j,t , Holiday Information H t , temperature T t and demand-side management factors D t ; S1073: Construct a feature vector based on the above data S1074. Use the XGBoost algorithm to train historical data and optimize model parameters to optimize model performance. S1075. Evaluate the accuracy of the model through cross-validation and make adjustments based on actual conditions; S1076. Input the data of the new time period into the trained model to obtain the predicted electricity price.

10. The method for self-adaptive time-scale hydropower spot dual-level quantity and price declaration based on complex constraints according to claim 1, wherein in step S2, the optimal bidding decision model is based on the following formula: Where, is the optimal output value achieved by trading unit c through quotation, is the time period deviation between the optimal output and the optimal simulated output, M delta is the output deviation penalty factor.

11. The complex constraint-based adaptive time-scale hydropower spot dual-level quantity and price declaration method according to claim 10 is characterized by: S201. The optimal bid decision model is also subject to the trading unit declaration format constraints, including: S2011, Trading unit declared output constraints; Initial declared capacity and minimum technical output P c,min The sum of the declared capacities of each segment is equal to the maximum permissible output of the trading unit; Where, Declare capacity for quote segment k of trading unit c, is the rated capacity of the trading unit; Where, The output is converted from the ecological discharge flow of reservoir r, that is, one section is declared according to the output converted from the ecological discharge flow. If it is limited by the maximum declared flow of a single section, the number of sections is increased accordingly; S2012, minimum declared capacity constraints; The length of each segment shall not be less than n% of the difference between the rated active power of the trading unit and the minimum stable technical output (which can be flexibly set according to local market rules); S2013, maximum declared capacity constraints; The length of each segment shall not exceed the maximum declared capacity of a single segment specified in the spot trading rules: Where, The maximum declared capacity of a single segment as specified in the spot trading rules; S2014, trading unit declared price constraints; The price declared for each segment is not lower than the price declared for the previous segment, and all segment quotations are within the quotation allowable range; AND min ≤Y c,k ≤Y max (39); AND c,k -AND c,k-1 ≥And ε (40); Where Y c,k is the declared price of transaction unit c quotation segment k, Y max 、Y min The upper and lower limits of the market parameter price declaration; when the lower limit of the market parameter price declaration is less than 0, Y min Take max{market parameter declared price lower limit, user set declared electricity price lower limit}, Y ε is the minimum quotation increment; the quotation for the output section corresponding to the ecological discharge flow is Y min .

12. The complex constraint-based adaptive time-scale hydropower spot dual-level quantity and price declaration method according to claim 11 is characterized by: Through subsequent verification and correction, the starting and ending outputs corresponding to the quotation segment are adjusted to integer multiples of 10,000 kW.

13. The complex constraint-based adaptive time-scale hydropower spot dual-level quantity and price declaration method according to claim 10 is characterized by: S202, the optimal bidding decision model is also constrained by the relationship between expected output and segmented bidding; If the reference electricity price is greater than or equal to the bid price for the segment, the capacity for that segment will be awarded; if the reference electricity price is lower than the bid price for the segment, the capacity for that segment will not be awarded. If LMP c,t <Y1, the output of the trading unit is the minimum technical output If Y n <LMP c,t ≤Y n+1 , then the n segment is full; If LMP c,t >Y N , then the winning bid is up to the adjustable upper limit; Where, LMP c,t is the predicted power station node electricity price, Y c,k is the declared price of quotation segment k of transaction unit c, Indicator variable representing the relationship between the quote and the reference price in segment k. When the quotation is greater than the reference price Contribute to the quotation strategy expectations, It is the difference between the installed capacity of the trading unit and the maximum power generation capacity.

14. The method for self-adaptive time-scale hydropower spot double-layer quantity and price declaration based on complex constraints according to claim 10 is characterized by: S203, the optimal bidding decision model is also subject to the expected output and optimal output constraints; Where, is the deviation between the expected output and the optimal output of the quotation strategy of trading unit c.

15. The method for self-adaptive time-scale hydropower spot double-layer quantity and price declaration based on complex constraints according to claim 10 is characterized by: S204, the optimal bidding decision model is also subject to the constraint that the declared capacity avoids the vibration zone; σ i,t,s =1, indicating that the highest gear vibration zone avoided by hydropower unit i in time period t is s, and the output of unit i in time period t is above the upper limit of the vibration zone s and below the lower limit of the vibration zone s+1: ∑σ i,t,s ≤1(45); Where S is the number of vibration sections, They are the lower and upper limits of the output of unit i in vibration zone s respectively.

Images