A method and system for evaluating power and energy balance of a cascade hydropower station

Abstract

The present application relates to the technical field of power and electricity balance evaluation of cascade hydropower stations, and particularly relates to a power and electricity balance evaluation method and system for cascade hydropower stations; by constructing multiple groups of scenarios containing the declared quantity-price curve sequence of the upstream power station in multiple time periods and the interval inflow sequence of the downstream power station, and establishing a complete calculation method from upstream water outflow, hydraulic evolution to downstream economic loss, the risk of hydraulic and power mismatch is quantitatively evaluated in the scenario of multi-owner cascade hydropower stations, which provides a solution to the problem that the downstream power station cannot predict the winning power of the upstream power station but needs to declare the contract electricity quantity, introduces the conditional value at risk in the financial field into the risk evaluation of the hydropower market, constructs a contract electricity quantity optimization model by minimizing the tail loss expectation, and gives the linear programming equivalent form of CVaR to realize efficient solution and generation of hedging strategies, and provides a complete technical solution from risk identification, risk quantification to risk hedging for the downstream power station.

Description

Technical Field

[0001] This invention relates to the field of power balance assessment technology for cascade hydropower stations, and in particular to a method and system for power balance assessment of cascade hydropower stations. Background Technology

[0002] In my country, upstream and downstream power plants in the same river basin are operated by different entities. Under the current electricity market rules, each independent cascade power plant needs to submit its bidding volume and price curves to the power trading center to participate in the day-ahead market clearing. However, due to the obstruction of information sharing between upstream and downstream power plants, and the fact that each cascade power plant makes its bidding decisions with the goal of maximizing its own interests, the market clearing results show a mismatch between hydropower and electricity.

[0003] The problem stems from the fact that the upstream power plant's winning bid is entirely determined by its own bidding strategy, while the amount of water used for power generation by the downstream power plant depends on the actual discharge flow from the upstream power plant. Therefore, when the upstream power plant's winning bid is low, resulting in insufficient discharge flow, the downstream power plant faces a situation where it has no water to generate electricity. This leads to the downstream power plant being unable to fulfill its bid, and it is forced to purchase electricity at a high price in the spot market to make up for the shortfall, or accept system imbalance penalties. Conversely, when the upstream power plant's winning bid is high and it releases water in a concentrated manner, the downstream power plant may be forced to abandon water due to limited reservoir capacity, resulting in a waste of water energy. This problem is particularly prominent in run-of-river downstream power plants with poor regulation performance, where their power generation is almost entirely dependent on the bidding results of the upstream power plant, and it is difficult to solve this problem.

[0004] Existing methods for scheduling cascade hydropower stations and power balance assessment technologies generally assume that upstream and downstream power stations belong to the same operating entity and can coordinate and formulate power generation plans. Therefore, existing methods are unable to solve the risk assessment problem caused by the lack of information that downstream power stations cannot predict the output of upstream stations when multiple owners bid independently. Summary of the Invention

[0005] To address the shortcomings of existing technologies, this invention provides a method and system for assessing the power balance of cascade hydropower stations. It solves the technical problem that existing methods struggle to address in scenarios where multiple owners bid independently, leading to a lack of information regarding the downstream power station's inability to predict the upstream power output.

[0006] To address the aforementioned technical problems, this invention provides the following technical solution: a method for assessing the power balance of a cascade hydropower station, applicable to scenarios where multi-owner cascade hydropower stations participate in the electricity spot market. This method specifically includes the following steps: S1. Collect historical auction clearing data of upstream power plants, inflow forecast data of downstream power plants, and historical clearing price data of the spot market respectively; S2. Based on the historical bidding clearing data and the interval water inflow forecast data, generate multiple sets of scenarios. Each set of scenarios includes the bidding volume and price curve sequence of the upstream power station in multiple time periods and the interval water inflow sequence of the downstream power station. S3. For each scenario, the spot market clearing is simulated based on the declared quantity and price curve sequence of the upstream power station to obtain the winning bid output process of the upstream power station, and the downstream discharge flow process of the upstream power station is calculated. The inflow flow process of the downstream power station is calculated by combining the interval water inflow sequence. Then, the actual power generation of the downstream power station is calculated by combining the unit power characteristic curve of the downstream power station. The actual power generation is compared with the market winning bid contract power of the downstream power station in each time period to calculate the deviation power generation sequence of the downstream power station in this scenario. The deviation power generation sequence is mapped to the economic loss sequence in this scenario. The economic loss includes at least the power purchase penalty cost caused by the power shortage and the resource waste loss caused by water abandonment. S4. Analyze multiple sets of economic loss sequences to obtain the probability distribution function of economic loss, and introduce the Conditional Value at Risk (CVaR) model to calculate the average loss that downstream power plants may suffer under the preset confidence level. S5. With the goal of minimizing conditional risk value, the algorithm optimizes the solution to output the recommended contract power allocation scheme for downstream power plants and the risk hedging strategy with upstream power plants.

[0007] Preferably, the specific steps for generating multiple sets of scenarios consisting of upstream bidding strategies and interval water inflow data are as follows: S21. Extract features from historical auction clearing data and construct an auction behavior feature vector for each trading day and time period, including base price, peak premium, price elasticity coefficient, and risk preference factor. S22. Based on the bidding behavior feature vectors of multiple historical trading days in various time periods, calculate the mean vector and covariance matrix for each time period; S23. Construct a four-variable normal distribution based on the mean vector and covariance matrix, and sample from it multiple times to obtain the bidding behavior feature vectors for multiple consecutive time periods. S24. Reconstruct the bidding volume and price curve of the upstream power station in that period according to the bidding behavior feature vector obtained in each sampling, and arrange them in order to obtain the bidding volume and price curve sequence. S25. Process the historical water inflow forecast data of the downstream power station, extract multiple uncertainty statistical parameters to characterize the uncertainty of water inflow, and perform time series modeling to obtain an interval water inflow prediction model to characterize the water inflow value of the downstream power station. S26. Calculate the interval water inflow values ​​that are equal to the number of declared quantity and price curves according to the interval water inflow prediction model, and arrange them in order to obtain the interval water inflow sequence. S27. Construct a scenario consisting of a sequence of declared volume and price curves of upstream power stations in multiple consecutive time periods and a sequence of water inflow of downstream power stations in intervals. S28. Repeat steps S21-S27 to obtain multiple scenes.

[0008] As a preferred method, the reconstructed volume-price curve is generated based on an exponentially decaying volume-price curve reconstruction function, the expression of which is: In the above formula, This represents the bid volume and price curve for the nth scenario in time period t, where p is the bid price variable. To maximize the technical output of the upstream power station during time period t, This is the base pricing parameter; when the price is lower than this value, the power plant will declare all available capacity. This is the price elasticity coefficient, which has a value greater than 0. It is used to control the rate at which the bid power decreases as the price increases.

[0009] Preferably, the interval water inflow prediction model is an AR(1) model, which includes at least the seasonal component sequence, the autoregressive coefficient sequence, and the white noise standard deviation sequence as its uncertainty statistical parameters. The interval water inflow sequence is generated recursively based on the AR(1) model. The specific steps are as follows: S261. Initialize the value of the random disturbance term for the initial time period; S262. Starting from the initial time period, perform the following operations sequentially for each time period to generate random disturbance terms: independently sample white noise from the standard normal distribution, and then calculate the random disturbance terms for the current time period according to the AR(1) model; S263. Add the random disturbance term of each time period to the seasonal component to obtain the interval inflow rate for that time period; S264. Physical constraint verification of the inflow rate of the interval is performed to ensure compliance with the basic laws of hydrology.

[0010] Preferably, step S4 specifically includes the following steps: S41. Construct an empirical distribution function of economic loss based on economic loss sequences from multiple scenarios. Its expression is: In the above formula, This represents the cumulative probability that the economic loss does not exceed a threshold x. Let n be the total economic loss in the nth scenario. This indicates an indicator function that takes the value 1 when the condition inside the curly braces is true, and 0 otherwise. R represents the set of real numbers. S42. Arrange the total economic loss values ​​of each scenario in ascending order to obtain the ordinal statistics; S43, at the preset confidence level Below, calculate the ordinal statistics. Quantiles are used to obtain value at risk. ; S44, at the same confidence level Below, the total economic loss is calculated to exceed the value at risk. The conditional expected value at time, in order to obtain the conditional value of risk. ; S45. Constructing Conditional Value at Risk The equivalent form of linear programming; S46. The calculated value of risk Conditional Value at Risk Perform a statistical stability test. If the test passes, proceed to step S5; otherwise, return to step S2 and increase the number of scenarios.

[0011] The present invention also provides an electricity balance assessment system, including a processor and a memory, wherein the memory is used to store a computer program, and the computer program, when executed by the processor, implements the electricity balance assessment method.

[0012] By employing the above technical solution, the present invention provides a method and system for assessing the power balance of a cascade hydropower station, which has at least the following beneficial effects: 1. This invention constructs multiple scenarios containing the declared quantity and price curve sequences of upstream power plants at multiple time periods and the interval water inflow sequence of downstream power plants, and establishes a complete calculation method from upstream clearing, hydraulic evolution to downstream economic losses. In this way, it can quantitatively assess the risk of mismatch between hydraulic and power in the scenario of multi-owner cascade power plants, and provides a solution to the problem that downstream power plants cannot predict the upstream winning bid output but still need to declare the contracted electricity volume.

[0013] 2. This invention simplifies the complex upstream bidding behavior of upstream power plants into characteristic parameters such as basic price and price elasticity coefficient. By constructing a quaternary normal distribution and reconstructing the price-volume curve of exponential decay, it can characterize the uncertainty of market behavior. At the same time, it introduces the AR(1) time series model to retain the correlation of the runoff process, so that the generated upstream bidding strategy and the water inflow scenario in the interval conform to the actual statistical law, which greatly improves the accuracy of risk assessment.

[0014] 3. This invention introduces conditional value of risk from the financial field into the risk assessment of the hydropower market, constructs a contract power optimization model by minimizing the expected tail loss, and provides an equivalent linear programming form of CVaR to achieve efficient solution. At the same time, hedging strategies can be generated based on the optimization results, providing downstream power plants with a complete technical solution from risk identification, risk quantification to risk hedging. Attached Figure Description

[0015] The accompanying drawings, which are included to provide a further understanding of this application and form part of this application, illustrate exemplary embodiments and are used to explain this application, but do not constitute an undue limitation of this application. In the drawings: Figure 1 This is a flowchart of the power balance assessment method for cascade hydropower stations according to the present invention. Detailed Implementation

[0016] To make the above-mentioned objects, features, and advantages of the present invention more apparent and understandable, the present invention will be further described in detail below with reference to the accompanying drawings and specific embodiments. This will allow for a full understanding of how the present application uses technical means to solve technical problems and achieve technical effects, and to facilitate its implementation.

[0017] To address the technical challenge of assessing the risk arising from the lack of information—specifically, the inability of downstream power plants to predict the output of upstream power plants in scenarios involving multiple owners bidding independently—this invention provides a power balance assessment method for cascade hydropower stations. Figure 1 As shown, this method, applied to a scenario where multi-owner cascade hydropower stations participate in the electricity spot market, specifically includes the following steps: S1. In a scenario where multiple hydropower stations in the same river basin participate in the electricity spot market, the core challenge faced by downstream power stations is their inability to predict the bidding behavior of upstream power stations or control the natural fluctuations in water inflow. They are required to declare their contracted electricity volume in advance. This information asymmetry leads to two situations for downstream power stations: either they are penalized for not generating electricity, or they are forced to abandon water due to excessive inflow. Therefore, this embodiment analyzes the uncertainties reflected in three dimensions: market behavior, natural conditions, and economic consequences. These sources are: historical bidding clearing data reflecting the market behavior of upstream power stations; water inflow forecast data reflecting the uncertainty of natural conditions for downstream power stations; and historical clearing price data of the spot market reflecting the severity of market penalties. The historical bidding clearing data of upstream power stations generally includes the bidding volume and price curves and corresponding clearing results for each trading day and time period (usually one hour) over a past period (e.g., at least one year). This data records the bidding strategies and winning bids of upstream power stations under different market conditions and forms the basis for analyzing the patterns of upstream bidding behavior. Downstream power plant inflow forecast data typically includes historical inflow data for the same period over a past period (e.g., ten years) and inflow forecast data for several future trading days. This data is used to analyze the seasonal patterns and random fluctuations of inflow. Historical clearing price data in the spot market typically includes the uniform clearing price for each trading day and time period over a past period (e.g., one year), as well as the penalty price for deviations in electricity volume as stipulated in market rules. This data is used to quantify subsequent economic losses.

[0018] S2. Traditional deterministic assessment methods can only provide a single prediction result and cannot quantify the risks brought about by uncertainty. Therefore, this embodiment transforms uncontrollable upstream bidding behavior and inter-regional water inflow into multiple quantifiable and calculable scenarios. By generating a large number of possible future operating conditions, it provides input for subsequent probabilistic risk assessment. That is, multiple scenarios are generated based on historical bidding clearing data and inter-regional water inflow forecast data. Each scenario includes the upstream power station's bid volume and price curve sequence for multiple time periods and the downstream power station's inter-regional water inflow sequence. Subsequently, based on the different bidding strategies that the upstream may adopt and the different fluctuation patterns that the inter-regional water inflow may exhibit, the downstream power station needs to evaluate its performance in all these scenarios. The specific steps for generating multiple scenarios composed of upstream bidding strategies and inter-regional water inflow data are as follows: S21. Extract features from historical auction clearing data to construct a bidding behavior feature vector for each trading day and each time period, including the base price, peak premium, price elasticity coefficient, and risk preference factor. For example, for M historical trading days, if each trading day d (d=1,2,...,M) contains 24 time periods t (t=1,2,...,24), then for each time period, the bid volume and price curve of the upstream power station in that time period can be extracted and quantified to include at least the base price. Peak hour premium Price elasticity coefficient and risk preference factors The bidding behavior feature vector of these four feature parameters The base price reflects the marginal cost of the power plant, typically taken as the lower quantile (e.g., 10th percentile) of the bid price for that period. The peak-hour premium reflects the increase in price by the power plant during peak hours (e.g., evening peak) relative to the base price. The bid elasticity coefficient reflects the rate at which the bid power decreases as the bid price increases; it can be obtained by fitting the slope of the bid volume-price curve. The risk preference factor reflects whether the power plant tends to bid high for higher returns or low to secure a bid; it can be estimated by the relationship between historical bid success rates and bid levels. Therefore, the bidding behavior characteristic vector can be represented as:

[0019] S22. Based on the bidding behavior feature vectors of multiple historical trading days in various time periods, calculate the mean vector and covariance matrix for each time period. The calculation formula is as follows: , In the above formula, the mean vector The covariance matrix describes the average level of bidding behavior during that period. It is used to characterize the correlation and fluctuation range between various feature parameters.

[0020] S23. In this implementation, it is assumed that the characteristic vector of bidding behavior follows a multivariate normal distribution. Therefore, a four-variate normal distribution is constructed based on the mean vector and the covariance matrix. And sampling is performed repeatedly to obtain bidding behavior feature vectors for multiple consecutive time periods. That is, for the nth scenario, sampling is performed sequentially for time periods t=1,2,...,24. The bidding behavior feature vector extracted for the tth time period in the nth scenario can be expressed as: ,in, This is the base price for the t-th time period in the n-th scenario. This refers to the peak-hour premium for the t-th time period in the n-th scenario. Let $\frac{ ... Let be the risk preference factor for the t-th time period in the n-th scenario. When sampling, it's best to also consider the correlation between time periods. Since bidding strategies in adjacent time periods are usually continuous, a structure that preserves the covariance between time periods can be used (e.g., by processing the joint distribution across all time periods through Cholesky decomposition), or it can be simply treated as independent sampling for each time period to reduce computational complexity.

[0021] S24. The bidding volume and price curves of the upstream power station for that time period are reconstructed sequentially based on the bidding behavior feature vectors obtained from each sampling, and then arranged sequentially to obtain a sequence of bidding volume and price curves. In this embodiment, the bidding volume and price curve reconstruction is generated based on an exponentially decaying volume and price curve reconstruction function. For time period t of the nth scenario, its expression can be: In the above formula, This represents the bid volume and price curve for the nth scenario in time period t, where p is the bid price variable. The maximum technical output of the upstream power station during time period t is determined by the power station's equipment parameters. This is the base pricing parameter; when the price is lower than this value, the power plant will declare all available capacity. The bid elasticity coefficient, which is greater than 0, controls the rate at which the bid power decreases as the bid price increases. A smaller value results in a steeper curve, indicating a more aggressive bidding strategy. This function satisfies smoothness, boundedness, and asymptoticity, and can describe bid volume-price curves of arbitrary shapes using only two parameters, significantly reducing modeling complexity. Arranging the curves of each time period in chronological order yields a bid volume-price curve sequence. For example, for a scenario with 24 time periods, its bid volume-price curve sequence can be represented as follows: .

[0022] S25. The collected historical inflow forecast data of the downstream power station is processed to extract multiple uncertainty statistical parameters to characterize the uncertainty of the inflow, and time series modeling is performed to obtain an interval inflow prediction model to characterize the inflow value of the downstream power station. In this embodiment, firstly, the historical inflow data is seasonally decomposed, that is, based on the interval inflow data of M historical trading days. , That is, the water inflow during time period t on day d, and the seasonal component of any time period t is calculated. This is used to characterize the overall level of water inflow within a given time period t, reflecting its general pattern. Its calculation can be expressed by the following formula for the average value:

[0023] The seasonal components for each date at each time period are calculated and arranged in chronological order to obtain the seasonal component sequence. Then, the random disturbance term is extracted. That is, the random disturbance term corresponding to time period t on day d, which is used to characterize the random variable of water inflow in the interval at any time period t based on the seasonal component. Its calculation can be expressed by the formula:

[0024] Then time series modeling can be performed. In this embodiment, an interval water inflow prediction model is constructed based on a first-order autoregressive model AR(1) and a random disturbance term. Its expression can be expressed as:

[0025] Finally, the least squares method was used to estimate the autoregressive coefficients for time period t. and white noise standard deviation Its calculation expression is: , In the above formula, Let represent the theoretical white noise term for time period t on day d, which follows a pattern with a mean of 0 and a variance of . The normal distribution, i.e. , This represents the residual estimate for time period t on day d. It is the residual actually calculated after model fitting, and it is the realized value of white noise estimated using sample data.

[0026] By calculating the autoregressive coefficients and white noise standard deviations for each time period and arranging them in chronological order, the autoregressive coefficient sequence and white noise standard deviation sequence can be obtained. Therefore, the following statistical parameters for characterizing the uncertainty of inflow water are ultimately obtained: the seasonal component sequence, the autoregressive coefficient sequence, and the white noise standard deviation sequence.

[0027] S26. Calculate the interval water inflow values ​​that are equal in number to the declared quantity and price curves according to the interval water inflow prediction model, and arrange them sequentially to obtain the interval water inflow sequence. The interval water inflow sequence is generated recursively through the AR(1) model. The specific steps are as follows: S261. In this embodiment, for the nth scenario, the value of the random disturbance term in the initial time period is initialized, and its value can be 0, i.e. Alternatively, the statistical value from the same historical period can be used.

[0028] S262. Subsequently, starting from the initial time period, the following operations are performed sequentially for each time period to generate random disturbance terms: White noise is independently sampled from the standard normal distribution, that is, the white noise sampled from time period t of the nth scene can be expressed as: Then, the random disturbance term for the current time period is calculated based on the AR(1) model. The calculation formula can be expressed as:

[0029] S263. Add the random disturbance term and seasonal component to each time period to obtain the interval inflow rate for that time period. For the interval inflow rate of time period t in the nth scenario, It is calculated as the random disturbance term for that period. With seasonal ingredients The sum can be expressed as:

[0030] S264. Perform physical constraint verification on the interval inflow rate. For example, if the interval inflow rate calculated in step S263 is less than 0, set the value to 0. If it exceeds the reasonable historical range, truncate it. If the rate of change is too large, resample or smooth it. In specific cases, constraints can be added according to the actual situation to ensure that the generated interval inflow rate conforms to hydrological laws.

[0031] S27. Construct a scenario consisting of a sequence of declared volume and price curves from upstream power plants over multiple consecutive time periods and a sequence of water inflow from downstream power plants within a given interval. For example, for the nth scenario, if it contains 24 time periods, its declared volume and price curve sequence and the sequence of water inflow from downstream power plants within a given interval can be represented as follows: and .

[0032] S28. Repeat steps S21-S27 to obtain multiple scenes. The total number of scenes can be set, generally between 5000 and 10000. Repeat steps S21-S27 the corresponding number of times to obtain that number of scenes.

[0033] S3. Next, it is necessary to assess the potential economic consequences for downstream power plants under each scenario. For each set of scenarios, such as the nth set, the spot market clearing is first simulated based on the bidding volume and price curve sequence of the upstream power plants to obtain the bidding output process of the upstream power plants. According to market rules, the bidding output of the upstream power plants in time period t is... The declared volume and price curves for the clearing price The corresponding power value is expressed as follows:

[0034] If a segmented pricing rule is adopted, the declared quantity-price curve is a discrete set of points, which can be represented as: ,in, This represents the k-th price point in the discrete price-volume curve. This indicates the price point of the upstream power station in the nth scenario at time t. The power of the application This represents the total number of discretized bid points. The winning bid force can then be calculated using linear interpolation, and the calculation formula can be expressed as: In the above formula, This represents the winning bid output of the upstream power station in the nth scenario during time period t. This represents the spot market clearing price for time period t.

[0035] The above formula can be used to calculate the winning bid output for the nth scenario in each time period. In this embodiment, taking a scenario with 24 time periods as an example, the winning bid output process of the upstream power station can be expressed as follows: Then, by combining the power characteristic curves of the upstream power plants, the required power generation flow process (i.e., the downstream flow process) for each time period can be deduced. The power characteristic curve of the units is usually expressed as the output. With power generation flow Functional relationship with head H: Given the winning bid output of the power station. and corresponding water head In this case, the water head can be calculated based on the changes in the reservoir capacity of the upstream reservoir, and then the winning bid output can be determined. and corresponding water head Substituting the above functional relationship, the power generation flow rate can be obtained by inverse solution. The change in upstream reservoir capacity can be recursively calculated using the water balance equation. Taking the nth scenario as an example, the expression of this equation can be expressed as: In the above formula, and This represents the reservoir capacity of the upstream reservoir in the nth scenario, at time interval t+1 and time interval t. This represents the inflow rate (including inter-regional water flow) of the upstream reservoir in the nth scenario during time period t. This represents the water discharge rate of the nth scenario in time period t. This indicates the length of a time period, typically 1 hour, or 3600 seconds. Let represent the power generation flow rate of the nth scenario in time period t. By calculating the reservoir capacity for each time period sequentially, the discharge flow rate for each time period can be calculated sequentially, thus obtaining the discharge flow rate process for that scenario. In actual calculations, if the upstream reservoir capacity is large, its regulating effect may smooth the discharge flow rate. However, to simplify the calculation, it can be approximated that the upstream power generation flow rate is constant according to the bid output, without considering the lag effect of reservoir capacity regulation. Of course, the specific calculation method can be set according to the actual situation, and this embodiment will not list them all.

[0036] After obtaining the downstream discharge flow from the upstream power station, the inflow flow from the downstream power station can be calculated by combining the inflow sequence of the interval. In this embodiment, the upstream discharge flow is superimposed with the inflow of the interval, and the hydraulic time lag of the river channel is considered to calculate the inflow flow from the downstream power station. Specifically, the propagation time required for water to flow from the upstream power station to the downstream power station is considered. The time it takes is typically several hours and can be estimated based on factors such as river length and flow velocity. Losses along the way may also occur due to evaporation and seepage; therefore, a flow attenuation coefficient is introduced. , Therefore, in time period t of scenario n, the inflow of the downstream power station is... The calculation formula can be expressed as: In the above formula, Indicates the upstream power station during the time period The power generation flow rate, if Then, the data for the corresponding time period of the previous day will be used. The interval inflow rate of the nth scenario is calculated in step S26. The inflow rate process of the downstream power station can be obtained by calculating the inflow rate of the downstream power station in each time period.

[0037] Then, based on the inflow process of the downstream power station and the power characteristic curve of the downstream power station's units, the actual power generation of the downstream power station is calculated. First, similarly, the reservoir capacity change of the downstream power station is recursively derived based on the water balance equation given above, and the reservoir capacity of the downstream power station at each time period is calculated. Then, the downstream power output is calculated based on the power characteristic curve of the downstream power station's units, thus obtaining the power output of the nth scenario at time period t. At this point, the actual power generation of the downstream power station in time period t is the product of the duration of that time period and its output.

[0038] Additionally, if limited by the unit's maximum output, the output calculated over time period t... Greater than the unit's maximum output Then, the maximum output of the generating unit is taken as the output for time period t. Additionally, if water needs to be discharged due to reservoir capacity limitations, the actual power generation flow will be less than the inflow. In this case, a more accurate calculation requires consideration of reservoir scheduling rules. However, to focus on risk assessment, this embodiment can use the following simplified formula to calculate the actual power generation: In the above formula, This represents the actual amount of electricity that can be generated in the nth scenario during time period t. This represents the overall efficiency coefficient of the unit during time period t. This represents the hydropower head of the downstream power station during time period t. This formula ignores the time-varying effect of reservoir regulation on the hydropower head and is applicable to run-of-river power stations or power stations with weak daily regulation capacity. Specific calculation methods for other types of power stations will not be listed one by one.

[0039] Finally, the actual generateable power is compared with the market-contracted power volume of downstream power plants in each time period to calculate the deviation power volume sequence of downstream power plants in this scenario. That is, the difference between the actual generateable power volume of downstream power plants and the market-contracted power volume of downstream power plants in each time period can be obtained. Then, the deviation power volume sequence is mapped to the economic loss sequence in this scenario. Usually, according to the deviation settlement rules of the electricity spot market, the deviation power volume is mapped to economic loss. In this embodiment, the market adopts a punitive purchase price for the shortfall in power volume and treats the excess power volume (water abandonment) with a lower compensation price or zero compensation. Therefore, the economic loss includes at least the purchase penalty cost caused by the power shortage and the resource waste loss caused by water abandonment. The following economic loss function is set for calculation: In the above formula, This represents the economic loss of scenario n in time period t. Let be the penalty price for the shortfall in time period t, and the loss calculated based on this is the cost of the electricity purchase penalty. This represents the deviation in electricity consumption for scenario n during time period t. This represents the discounted price of water wastage during time period t, reflecting the value of the wasted water resources. The economic loss calculated based on this is the resource waste loss, typically expressed as the grid-connected electricity price or opportunity cost. This represents the total economic loss corresponding to scenario n. This is the sum of economic losses at each time point in the scenario.

[0040] S4. After obtaining the economic loss sequences under N scenarios, it is necessary to extract statistical indicators with risk management significance from these discrete loss data. Traditional methods usually focus on expected loss (i.e., average value), but the average value cannot reflect extreme risks. For example, the loss may be small in 95% of cases, but huge in 5% of cases. If downstream power plants only focus on the average loss, they are likely to suffer huge economic losses when extreme events occur. Therefore, this invention analyzes multiple sets of economic loss sequences to obtain the probability distribution function of economic loss, and introduces the Conditional Value at Risk (CVaR) model to calculate the average loss that downstream power plants may suffer under a preset confidence level. The specific steps are as follows: S41. Calculate the total economic loss for each scenario based on the economic loss sequences of multiple scenarios, and then construct an empirical distribution function of economic loss using the total economic loss values ​​of N scenarios. Its expression is: In the above formula, This represents the cumulative probability that the economic loss does not exceed a threshold x. Let n be the total economic loss in the nth scenario. This indicates an indicator function that takes the value 1 when the condition within the curly braces is true, and 0 otherwise. R represents the set of real numbers.

[0041] S42. Arrange the total economic loss values ​​of each scenario in ascending order (i.e., from smallest to largest) to obtain the ordinal statistics.

[0042] S43, at the preset confidence level Below, calculate the ordinal statistics. Quantiles are used to obtain value at risk. Among them, confidence level Typically, values ​​of 0.90, 0.95, or 0.99 are used, indicating that... The probability guarantees that economic losses will not exceed the value at risk. Based on ordinal statistics, The estimated value is: ,in, The rounding up symbol, Indicates not less than The smallest integer after sorting, i.e., the smallest integer after sorting. Each loss value is used as Here, N represents the total number of scenes.

[0043] S44, at the same confidence level Below, the total economic loss is calculated to exceed the value at risk. The conditional expected value at time, in order to obtain the conditional value of risk. , For losses exceeding the value at risk The average loss over time, i.e. the expected tail loss, is estimated based on the order statistic in actual calculations. Its calculation expression is as follows: In the above formula, This represents the k-th value in the ordinal statistic, ordered from smallest to largest. The formula applies to all values ​​greater than or equal to the Value at Risk. The loss value is taken as the arithmetic mean.

[0044] S45. In order to adjust the conditional risk value As the optimization objective is embedded in the computational model of subsequent step S5, it needs to be expressed in a mathematical form that is easy to solve. Therefore, it is necessary to construct conditional value of risk. The equivalent form of linear programming proposed by Rockafellar and Uryasev for CVaR linear programming is as follows: In the above formula, To assist in optimizing variables, their optimal value is... , This indicates that the economic loss in the nth scenario exceeds... The part of this form is convex and can be linearized by introducing auxiliary variables, making it easier to solve efficiently using a linear programming solver.

[0045] S46. The calculated value of risk Conditional Value at Risk A statistical stability test is performed. If the test passes, proceed to step S5; otherwise, return to step S2 and increase the number of scenarios. Specifically, in this embodiment, considering that the choice of the number of scenarios N in the Monte Carlo simulation directly affects the stability of the risk indicators, if N is too small, the estimates of VaR and CVaR may fluctuate significantly. This embodiment uses the bootstrap resampling method for stability testing. The specific steps are as follows: N samples with replacement are drawn from the original N economic loss values ​​to form the bootstrap sample set. The sampling is repeated B times (e.g., 1000 times), and the conditional value of risk is calculated for each sampling. Then calculate the standard deviation and relative error of the B CVaR estimates, where the relative error is the ratio of the standard deviation to the conditional value at risk. If the relative error is less than the preset threshold (e.g., 5%), the indicator is considered stable and can proceed to step S5; otherwise, the number of scenes N needs to be increased (e.g., the number of scenes is increased by a certain multiple, such as 2 times), and then the process returns to step S2 to regenerate the scene and calculate.

[0046] S5. With the goal of minimizing conditional risk value, the recommended contracted electricity volume decomposition scheme for downstream power plants and risk hedging strategies with upstream power plants are output through an optimized solution algorithm. In this embodiment, the solution can be performed according to the following steps: First, a decision variable vector can be constructed based on the contracted electricity volume schemes of downstream power plants in multiple future time periods: In the above formula, For the decision variable vector, This refers to the contracted electricity volume that downstream power plants intend to declare or sign during the time period t.

[0047] Then, as mentioned earlier, Conditional Value at Risk The objective function is constructed by minimizing the equivalent form of the linear programming problem, and its expression is as follows: In the above formula, Given a contract power consumption scheme in the nth scenario The total economic loss at that time, as described in step S3, can be calculated using the following formula: ,

[0048]

[0049] In the above formula, This represents the actual power generation capacity of the downstream power plants in scenario n. The contracted electricity volume won by the downstream power plant in the market during time period t is the contracted electricity volume that the downstream power plant intends to declare or sign in time period t.

[0050] In solving the above objective function, constraints need to be added according to the actual situation, such as: 1. The total contracted electricity volume of the downstream power station should match the expected power generation capacity; 2. The contracted electricity volume in each time period is limited by the technical output of the generating unit; 3. The rate of change of the contracted electricity volume in adjacent time periods is limited by the ramp-up capability of the generating unit; the execution of the contracted electricity volume must meet the reservoir capacity limit of the downstream reservoir, because the contracted electricity volume determines the power generation water use plan, which in turn affects the reservoir water level, and it is necessary to ensure that the reservoir water level does not exceed the limit in all possible scenarios; 4. The constraint is that the number is non-negative. The specific constraint is determined based on the actual situation, and will not be listed in detail in this embodiment.

[0051] In addition, this embodiment takes into account the objective function. This term is non-linear, therefore, by introducing an auxiliary variable... Perform exact linearization:

[0052]

[0053] Transform the objective function into a linear form:

[0054] because yes The piecewise linear function, by As can be seen from the expression, it is actually The broken line function is essentially a V-shaped function composed of two linear segments. Therefore, the entire optimization problem is a linear programming problem, which can be efficiently solved using standard algorithms such as the simplex method or interior point method. This yields the optimal contract power allocation scheme, the corresponding risk value, and the corresponding conditional risk value. Then, based on the optimal contract power allocation scheme and the actual power generation under all scenarios, system risk can be analyzed and hedging strategies generated. First, the average deviation power of each time period t under all scenarios can be calculated. If this average deviation power is less than 0, it indicates a systemic shortage risk in time period t, meaning actual power generation may be insufficient. Conversely, if it is greater than 0, it indicates a systemic water curtailment risk in time period t, meaning there may be excessive water inflow. After identifying the source of risk, the following hedging strategies can be adopted: 1. Contract for Difference (CfD) strategy: For periods with systemic shortage risk, it is recommended that downstream power plants sign CfD contracts with upstream power plants to lock in the purchase cost of a portion of the power. The contract parameters in this embodiment can be determined using the following formula: Contract power:

[0055] Contract Price: In the above formula, This is the risk coverage ratio, with a value between 0 and 1, used to control the hedging ratio. A value of 0.5 indicates that half of the systemic deficit is hedged. This is a discount factor used to reflect the revenue compensation required by the upstream power plant for bearing the risk. A value of 0.1 indicates a 10% discount on the price. This represents the average deviation of electricity consumption for each time period t under all scenarios calculated previously.

[0056] 2. Physical power generation rights transfer strategy: For periods with systemic water curtailment risks, it is recommended that downstream power plants sign physical power generation rights transfer agreements with upstream power plants to transfer excess power generation capacity to upstream power plants or third parties.

[0057] Downstream power plants can incorporate the above hedging strategies into their considerations, and then recalculate the economic losses under each scenario. For example, after signing a CfD agreement, the cost of purchasing electricity for the shortfall is reduced, and the risk value after hedging is recalculated. Conditional Value at Risk This allows for verification of the risk reduction effect, ultimately leading to the output of a recommended contract power allocation scheme and the value of risk before and after optimization. Conditional Value at Risk and the value at risk after implementing a hedging strategy Conditional Value at Risk And the specific parameters of the hedging strategy used, etc.

[0058] The present invention also provides an electricity balance assessment system, including a processor and a memory, wherein the memory is used to store a computer program, and the computer program is executed by the processor to implement an electricity balance assessment method.

[0059] Those skilled in the art will understand that all or part of the steps in the methods of the above embodiments can be implemented by a program instructing related hardware. Therefore, this application can take the form of a completely hardware embodiment, a completely software embodiment, or an embodiment combining software and hardware aspects. Moreover, this application can take the form of a computer program product implemented on one or more computer-usable storage media (including but not limited to disk storage, CD-ROM, optical storage, etc.) containing computer-usable program code.

[0060] The various embodiments in this specification are described in a progressive manner, with each embodiment focusing on its differences from other embodiments. Similar or identical parts between embodiments can be referred to interchangeably. Since the above embodiments are substantially similar to the method embodiments, their descriptions are relatively simple; relevant parts can be referred to the descriptions of the method embodiments.

[0061] The above embodiments provide a detailed description of the present invention. Specific examples have been used to illustrate the principles and implementation methods of the present invention. The descriptions of the above embodiments are only for the purpose of helping to understand the method and core ideas of the present invention. At the same time, for those skilled in the art, there will be changes in the specific implementation methods and application scope based on the ideas of the present invention. Therefore, the content of this specification should not be construed as a limitation of the present invention.

Claims

1. A method for assessing the power balance of a cascade hydropower station, applied to scenarios where multiple-owner cascade hydropower stations participate in the electricity spot market, characterized in that... The method specifically includes the following steps: S1. Collect historical auction clearing data of upstream power plants, inflow forecast data of downstream power plants, and historical clearing price data of the spot market respectively; S2. Based on the historical bidding clearing data and the interval water inflow forecast data, generate multiple sets of scenarios. Each set of scenarios includes the bidding volume and price curve sequence of the upstream power station in multiple time periods and the interval water inflow sequence of the downstream power station. S3. For each scenario, the spot market clearing is simulated based on the declared quantity and price curve sequence of the upstream power station to obtain the winning bid output process of the upstream power station, and the downstream discharge flow process of the upstream power station is calculated. The inflow flow process of the downstream power station is calculated by combining the interval water inflow sequence. Then, the actual power generation of the downstream power station is calculated by combining the unit power characteristic curve of the downstream power station. The actual power generation is compared with the market winning bid contract power of the downstream power station in each time period to calculate the deviation power generation sequence of the downstream power station in this scenario. The deviation power generation sequence is mapped to the economic loss sequence in this scenario. The economic loss includes at least the power purchase penalty cost caused by the power shortage and the resource waste loss caused by water abandonment. S4. Analyze multiple sets of economic loss sequences to obtain the probability distribution function of economic loss, and introduce the Conditional Value at Risk (CVaR) model to calculate the average loss that downstream power plants may suffer under the preset confidence level. S5. With the goal of minimizing conditional risk value, the algorithm optimizes the solution to output the recommended contract power allocation scheme for downstream power plants and the risk hedging strategy with upstream power plants.

2. The power balance assessment method according to claim 1, characterized in that, The specific steps for generating multiple sets of the aforementioned scenes are as follows: S21. Extract features from historical auction clearing data and construct an auction behavior feature vector for each trading day and time period, including base price, peak premium, price elasticity coefficient, and risk preference factor. S22. Based on the bidding behavior feature vectors of multiple historical trading days in various time periods, calculate the mean vector and covariance matrix for each time period; S23. Construct a four-variable normal distribution based on the mean vector and covariance matrix, and sample from it multiple times to obtain the bidding behavior feature vectors for multiple consecutive time periods. S24. Reconstruct the bidding volume and price curve of the upstream power station in that period according to the bidding behavior feature vector obtained in each sampling, and arrange them in order to obtain the bidding volume and price curve sequence. S25. Process the historical water inflow forecast data of the downstream power station, extract multiple uncertainty statistical parameters to characterize the uncertainty of water inflow, and perform time series modeling to obtain an interval water inflow prediction model to characterize the water inflow value of the downstream power station. S26. Calculate the interval water inflow values ​​that are equal to the number of declared quantity and price curves according to the interval water inflow prediction model, and arrange them in order to obtain the interval water inflow sequence. S27. Construct a scenario consisting of a sequence of declared volume and price curves of upstream power stations in multiple consecutive time periods and a sequence of water inflow of downstream power stations in intervals. S28. Repeat steps S21-S27 to obtain multiple scenes.

3. The power balance assessment method according to claim 2, characterized in that, The reconstructed volume-price curve is generated based on the exponential decay volume-price curve reconstruction function, and its expression is: In the above formula, This represents the bid volume and price curve for the nth scenario in time period t, where p is the bid price variable. To maximize the technical output of the upstream power station during time period t, This is the base pricing parameter; when the price is lower than this value, the power plant will declare all available capacity. This is the price elasticity coefficient, which has a value greater than 0. It is used to control the rate at which the bid power decreases as the price increases.

4. The power balance assessment method according to claim 3, characterized in that, The interval water inflow prediction model is an AR(1) model, which includes at least the seasonal component sequence, the autoregressive coefficient sequence, and the white noise standard deviation sequence as its uncertainty statistical parameters. The interval water inflow sequence is generated recursively based on the AR(1) model. The specific steps are as follows: S261. Initialize the value of the random disturbance term for the initial time period; S262. Starting from the initial time period, perform the following operations sequentially for each time period to generate random disturbance terms: independently sample white noise from the standard normal distribution, and then calculate the random disturbance terms for the current time period according to the AR(1) model; S263. Add the random disturbance term of each time period to the seasonal component to obtain the interval inflow rate for that time period; S264. Physical constraint verification of the inflow rate of the interval is performed to ensure compliance with the basic laws of hydrology.

5. The power balance assessment method according to claim 1, characterized in that, Step S4 specifically includes the following steps: S41. Construct an empirical distribution function of economic loss based on economic loss sequences from multiple scenarios. Its expression is: In the above formula, This represents the cumulative probability that the economic loss does not exceed a threshold x. Let n be the total economic loss in the nth scenario. This indicates an indicator function that takes the value 1 when the condition inside the curly braces is true, and 0 otherwise. R represents the set of real numbers. S42. Arrange the total economic loss values ​​of each scenario in ascending order to obtain the ordinal statistics; S43, at the preset confidence level Below, calculate the ordinal statistics. Quantiles are used to obtain value at risk. ; S44, at the same confidence level Below, the total economic loss is calculated to exceed the value at risk. The conditional expected value at time, in order to obtain the conditional value of risk. ; S45. Constructing Conditional Value at Risk The equivalent form of linear programming; S46. The calculated value of risk Conditional Value at Risk Perform a statistical stability test. If the test passes, proceed to step S5; otherwise, return to step S2 and increase the number of scenarios.

6. A system for implementing the power balance assessment method according to any one of claims 1-5, characterized in that, It includes a processor and a memory, the memory being used to store a computer program, which, when executed by the processor, implements the power balance assessment method as described in any one of claims 1-5.

Images