Power grid supply guarantee situation quantitative analysis method based on multi-time scale flexibility demand identification

Abstract

The invention discloses a power grid supply guarantee situation quantitative analysis method based on multi-time scale flexibility demand identification. The method comprises the following steps: 1) constructing a net load quantile prediction sequence satisfying energy closing and quantile monotonicity; 2) based on the net load quantile prediction sequence, calculating a short-time flexibility demand and a long-time flexibility demand, and calculating a flexibility balance and risk assessment index; 3) taking the net load quantile prediction sequence, the short-time flexibility demand and the long-time flexibility demand as input, solving the scheduling optimization model based on time sequence production simulation, and obtaining a flexibility relaxation amount and a supply guarantee gap at each moment; and 4) calculating the supply guarantee demand and the flexibility gap rate based on the flexibility slack amount and the supply guarantee gap at each moment. The dynamic evaluation of the supply guarantee capability of the power grid under different time scales can be realized, the high-risk time period and key bottleneck factors can be accurately identified, and a quantitative decision basis is provided for the operation scheduling and planning of the power grid.

Description

Technical Field

[0001] This invention relates to the field of power grid operation, aiming to solve problems such as the disconnect between forecast results and supply capacity assessment, imperfect identification of flexibility demand, and lack of multi-timescale comprehensive analysis in power grids with a high proportion of clean energy. Specifically, it is a quantitative analysis method for power grid supply situation based on multi-timescale flexibility demand identification. The core of this invention includes: (1) establishing an energy-shape consistency model for medium- and long-term electricity and short-term power forecasts to achieve multi-timescale forecast fusion; (2) introducing interval forecasting and quantile feature extraction to transform forecast uncertainty into a flexibility demand curve; (3) using time-series production simulation methods, combined with system constraints, to construct multiple types of slack variables to quantify supply gaps such as balance, reserve, and ramp-up. This invention can realize dynamic assessment of power grid supply capacity at different time scales, accurately identify high-risk periods and key bottleneck factors, and provide quantitative decision-making basis for power grid operation scheduling and planning. Background Technology

[0002] With the continuous integration of a high proportion of renewable energy into the power system, the power grid is gradually evolving from a deterministic system dominated by conventional thermal power to an uncertain system dominated by new energy sources such as wind and solar power. The output of such systems exhibits significant randomness, volatility, and intermittency, significantly increasing the difficulty of balancing power supply and demand. Traditional dispatching methods relying on planning and experience are no longer sufficient to meet the requirements for safe and stable grid operation. Particularly in regions with a high proportion of clean energy, such as Southwest China, wind and solar power output is severely affected by weather conditions, and hydropower plays a crucial role in regulation and supply assurance. Therefore, accurately identifying the flexibility needs at different stages of the future and quantifying the grid's supply assurance pressure has become a core scientific issue in grid operation and planning.

[0003] Existing research generally employs two technical approaches to address the risks posed by the uncertainty of new energy sources. One type of research focuses on the prediction and correction of new energy power output, utilizing methods such as statistical learning, machine learning, or deep learning to improve short-term prediction accuracy, enabling dispatch centers to grasp power output trends earlier. However, these methods often remain at the "prediction" level, only providing output curves or probability distributions, lacking a mechanism to translate prediction results into system operation indicators, and failing to intuitively reflect the power grid's supply security risks at different time scales. The other type of research focuses on the quantitative assessment of flexibility demand and supply, analyzing the adequacy of system flexibility resources by establishing flexibility indicator systems such as peak shaving, frequency regulation, and reserve. While this type of method can assess flexibility gaps from the resource side, it often assumes fixed input data and fails to consider prediction errors and the propagation effect of uncertainty, leading to discrepancies between the assessment results and actual operational risks.

[0004] Furthermore, existing flexibility analysis methods generally suffer from fragmented time scales. Short-term studies typically assess intraday balance in hourly or 15-minute increments, while medium- to long-term studies target monthly or annual power balance, lacking a data-model mapping relationship between the two. Due to the lack of a multi-timescale fusion mechanism, forecast results cannot maintain consistency across different time levels, leading to insufficient systematic assessment of flexibility requirements and supply assurance. Simultaneously, existing assessment frameworks are mostly static or single-scenario analyses, making it difficult to quantify system tail risks under extreme weather and forecast biases, and failing to provide quantitative risk indicators for resource planning.

[0005] In summary, current technologies lack a comprehensive analytical method that can integrate forecast results with system regulation constraints across multiple time scales, failing to effectively reflect the supply security situation of high-proportion clean energy power grids under different operating scenarios. Particularly when considering renewable energy range forecasts, flexibility demand fluctuations, and various grid constraints, existing methods struggle to systematically transform forecast results into supply security risk indicators. Therefore, there is an urgent need to propose a quantitative analysis method for supply security that considers multi-time scale forecasts, consistency constraints, and time-series scheduling simulations, enabling a comprehensive assessment and early warning of the operational safety of high-proportion clean energy power grids. Summary of the Invention

[0006] The purpose of this invention is to provide a quantitative analysis method for power grid supply security based on multi-timescale flexibility demand identification, comprising the following steps:

[0007] Step 1) Construct a net load quantile prediction sequence that satisfies energy closure and quantile monotonicity;

[0008] Step 2) Based on the net load quantile forecast sequence, calculate the short-term flexibility requirement and the long-term flexibility requirement, and calculate the flexibility balance and risk assessment index.

[0009] Step 3) Using the net load quantile forecast sequence, short-term flexibility demand, and long-term flexibility demand as inputs, solve the scheduling optimization model based on time-series production simulation to obtain the flexibility slack and supply gap at each time point.

[0010] Step 4) Calculate the supply demand and flexibility gap rate based on the flexibility slack and supply gap at each moment;

[0011] The overall flexibility level of the power grid is determined by flexibility balance and risk assessment indicators, and the supply guarantee demand of the power grid during time-series operation is determined by supply guarantee demand and flexibility gap rate.

[0012] Furthermore, in step 1, the steps for constructing the net load quantile prediction sequence that satisfies energy closure and quantile monotonicity include:

[0013] Step 1.1) Under a unified time resolution, acquire input data, including historical measured data of wind power, photovoltaic, hydropower and load, meteorological forecast data and operating status information;

[0014] Step 1.2) Based on the input data, use the short-term power prediction model to predict the short-term power and obtain the short-term power curve;

[0015] Based on the input data, the medium- and long-term electricity consumption forecasting model is used to predict the medium- and long-term electricity consumption, and the medium- and long-term electricity consumption curve is obtained.

[0016] Step 1.3) Use energy-shape consistency to perform cross-scale alignment and quantile monotonic correction on short-term power curves and medium- to long-term charge curves to form a net load quantile prediction sequence that satisfies energy closure and quantile monotonicity.

[0017] Furthermore, meteorological data includes wind speed, irradiance, temperature, relative humidity, precipitation, and water flow.

[0018] Operating status information includes unit output, equipment availability, installed capacity, and load level;

[0019] The input data also undergoes preprocessing, including time interpolation, missing value imputation, and normalization preprocessing;

[0020] The preprocessed input data is represented as follows:

[0021] (1)

[0022] in, Indicates the first Energy-like energy at all times Standardized meteorological driving variables, The raw meteorological data, These are the mean and standard deviation of the variable, respectively.

[0023] Furthermore, the short-term power interval prediction model includes a CNN module, an LSTM module, and an output layer with a quantile regression branch;

[0024] The CNN module is used to extract local temporal features from meteorological sequences, and the LSTM module is used to characterize the dynamic changes of power over time.

[0025] The output of the CNN module is shown below:

[0026] (2)

[0027] in, For the first Each convolutional channel at time... The output characteristics, The kernel length is 1. As a meteorological feature dimension, For convolution channels; For convolution kernel and bias term;

[0028] The output of the LSTM module is shown below:

[0029] (3)

[0030] in, This represents the input gate, forget gate, and output gate of an LSTM. Candidate memory states, The current memory state, Indicates a hidden state. It is the Sigmoid activation function. Multiply by Hadamard; This is the concatenated input; W and b represent the weights and biases, respectively.

[0031] The output layer of the short-term power range prediction model is shown below:

[0032] (4)

[0033] in, For linear layer weights and bias terms, For the first Energy-like energy at all times Short-term predicted power;

[0034] The loss function for the short-term power range prediction model is shown below:

[0035] (5)

[0036] in, For actual measured power, This is the quantile error function. The loss is multiplied by the method when the actual error is negative, thus obtaining the upper and lower bounds of the prediction under different confidence intervals. For the first Quantile prediction.

[0037] Furthermore, the medium- and long-term power prediction model adopts the XGBoost model;

[0038] The medium- and long-term electricity forecasting model outputs the medium- and long-term electricity volume. As shown below:

[0039] (6)

[0040] (7)

[0041] in, For the first A tree of return, It is a monthly aggregate vector of meteorological and macroscopic characteristics (such as hydrology, climate, load growth rate, etc.); This is the fitting loss (such as squared loss). To measure the actual power consumption, For complexity regularization; The number of leaf nodes. For node weights, is the regularization coefficient. , The loss function and the actual electricity consumption are given.

[0042] Furthermore, the steps for cross-scale alignment and quantile monotonic correction of the short-term power curve and the medium-to-long-term energy curve include:

[0043] Step 1.3.1) According to the constraint relationship (8), perform energy alignment on the short-term power curve and the medium- and long-term energy curve at different time scales;

[0044] (8)

[0045] in, For time scale, This is the scaling factor. This refers to the quantile power after energy alignment;

[0046] Step 1.3.2) Monotonicity correction is performed using the quantile projection method, resulting in:

[0047] (9)

[0048] Step 1.3.3) Construct the net load quantile prediction sequence, i.e.:

[0049] (10)

[0050] in, The load quantile power prediction results, These are the power generation quantiles after energy alignment and monotonic correction. .

[0051] Furthermore, step 2, which involves calculating short-term and long-term flexibility requirements and determining the flexibility balance and risk assessment metrics, includes:

[0052] Step 2.1) Based on the net load quantile prediction sequence, construct the net load interval, i.e.:

[0053] (11)

[0054] Step 2.2) Calculate the short-term flexibility requirements for uplink and downlink, i.e.:

[0055] (12)

[0056] in, Indicates time arrive The maximum upward adjustment demand between; This indicates a decrease in demand for adjustment;

[0057] Step 2.3) Calculate the long-term flexibility requirements for uplink and downlink, i.e.:

[0058] (13)

[0059] (14)

[0060] In the formula, the time window ; , This indicates the need for long-term flexibility in both upward and downward movements;

[0061] Step 2.4) Calculate the flexibility balance and risk assessment index to evaluate the flexibility and security level of the system;

[0062] Among them, the indicators for balancing flexibility and risk assessment include flexibility margin (FM) and conditional value at risk (CVaR);

[0063] The flexibility margin (FM) is shown below:

[0064] (15)

[0065] (16)

[0066] (17)

[0067] (18)

[0068] In the formula, It refers to the flexibility that the system can provide in the uplink direction; It refers to the flexibility that the system can provide in the downlink direction; and This refers to the flexibility requirements derived from the aforementioned calculations. If or This indicates a lack of flexibility in that direction. System flexibility and redundancy; when The system has a flexibility gap. When This indicates that the system has sufficient flexibility; when This indicates that the system lacks overall flexibility.

[0069] The conditional value at risk (CVaR) is shown below:

[0070] First, define the value at risk at confidence level α:

[0071] (19)

[0072] (20)

[0073] In the formula, It is the quantile threshold of the flexibility gap at confidence level α; This is the expected flexibility gap under the most unfavorable 1-α percentile scenario; if If so, the system has insufficient average flexibility at confidence level α.

[0074] Furthermore, the objective function of the scheduling optimization model based on time-series production simulation is as follows:

[0075] (twenty one)

[0076] in, For unit operating costs; Cost of energy storage charging and discharging; The penalty coefficient for the flexibility gap; The penalty coefficient for supply gaps.

[0077] Furthermore, the constraints of the scheduling optimization model based on time-series production simulation include power balance constraints, unit and energy storage output constraints, and flexibility constraints.

[0078] The power balance constraints are as follows:

[0079] (twenty two)

[0080] In the formula, For the unit At any moment Power output (MW); It is the energy storage system at all times The charging and discharging power; It is the slack in the supply and demand balance;

[0081] The power output constraints of the generating unit and energy storage are as follows:

[0082] (twenty three)

[0083] (twenty four)

[0084] in, The energy state of the energy storage system; , These are the upper and lower limits of energy storage for the energy storage system; , For the unit At any moment The upper and lower limits of output;

[0085] The flexibility constraints are as follows:

[0086] (25)

[0087] (26)

[0088] in, This refers to the uplink and downlink regulation capabilities that the generator set can provide. It refers to the uplink and downlink regulation capabilities that energy storage systems can provide; This represents the flexibility gap between uplink and downlink, and is a non-negative variable. When flexibility resources are insufficient, A positive value indicates that standby capacity or purchased electricity is required. This indicates insufficient uplink flexibility, requiring additional issuance or backup. This indicates insufficient downlink flexibility, which may lead to curtailment of solar and wind power. This indicates that the supply-demand balance cannot be maintained, resulting in a supply gap.

[0089] Furthermore, the supply demand represents the weighted compensation amount required by the system to maintain supply-demand balance during the simulation period, i.e.:

[0090] (27)

[0091] In the formula, As weight; It is the slack in the supply and demand balance;

[0092] The flexibility gap ratio represents the proportion of the system's unmet flexibility requirements across all time periods, out of the total demand.

[0093] (28)

[0094] In the formula, For the flexibility gap rate.

[0095] The technical effectiveness of this invention is undeniable. It constructs a unified technical framework from panoramic power generation forecasting to quantification of supply security risks: by establishing an energy-shape consistency model for medium- and long-term power generation and short-term power forecasting, it achieves forecast fusion across multiple time scales; by introducing interval forecasting and quantile feature extraction, it transforms forecast uncertainty into a flexibility demand curve; and by utilizing time-series production simulation methods, combined with system constraints, it constructs multiple types of relaxation variables to quantify supply security gaps such as balance, reserve, and ramp-up. This invention enables dynamic assessment of the power grid's supply security capacity at different time scales, accurately identifies high-risk periods and key bottleneck factors, and provides quantitative decision-making basis for power grid operation scheduling and planning. Attached Figure Description

[0096] Figure 1 This is a flowchart of the method. Detailed Implementation

[0097] The present invention will be further described below with reference to embodiments, but it should not be construed that the scope of the present invention is limited to the following embodiments. Various substitutions and modifications made based on ordinary technical knowledge and common practices in the art without departing from the above-described technical concept of the present invention should be included within the scope of protection of the present invention.

[0098] Example 1:

[0099] A quantitative analysis method for power grid supply security based on multi-timescale flexibility demand identification includes the following steps:

[0100] Step 1) Construct a net load quantile prediction sequence that satisfies energy closure and quantile monotonicity;

[0101] Step 2) Based on the net load quantile forecast sequence, calculate the short-term flexibility requirement and the long-term flexibility requirement, and calculate the flexibility balance and risk assessment index.

[0102] Step 3) Using the net load quantile forecast sequence, short-term flexibility demand, and long-term flexibility demand as inputs, solve the scheduling optimization model based on time-series production simulation to obtain the flexibility slack and supply gap at each time point.

[0103] Step 4) Calculate the supply demand and flexibility gap rate based on the flexibility slack and supply gap at each moment;

[0104] The overall flexibility level of the power grid is determined by flexibility balance and risk assessment indicators, and the supply guarantee demand of the power grid during time-series operation is determined by supply guarantee demand and flexibility gap rate.

[0105] Example 2:

[0106] A quantitative analysis method for power grid supply security based on multi-timescale flexibility demand identification, with the same technical content as Embodiment 1, further comprising the following steps in step 1: constructing a net load quantile prediction sequence that satisfies energy closure and quantile monotonicity.

[0107] Step 1.1) Under a unified time resolution, acquire input data, including historical measured data of wind power, photovoltaic, hydropower and load, meteorological forecast data and operating status information;

[0108] Step 1.2) Based on the input data, use the short-term power prediction model to predict the short-term power and obtain the short-term power curve;

[0109] Based on the input data, the medium- and long-term electricity consumption forecasting model is used to predict the medium- and long-term electricity consumption, and the medium- and long-term electricity consumption curve is obtained.

[0110] Step 1.3) Use energy-shape consistency to perform cross-scale alignment and quantile monotonic correction on short-term power curves and medium- to long-term charge curves to form a net load quantile prediction sequence that satisfies energy closure and quantile monotonicity.

[0111] Example 3:

[0112] A quantitative analysis method for power grid supply guarantee based on multi-timescale flexibility demand identification, with the same technical content as any one of Examples 1-2, further including meteorological data such as wind speed, irradiance, temperature, relative humidity, precipitation and inflow rate;

[0113] Operating status information includes unit output, equipment availability, installed capacity, and load level;

[0114] The input data also undergoes preprocessing, including time interpolation, missing value imputation, and normalization preprocessing;

[0115] The preprocessed input data is represented as follows:

[0116] (1)

[0117] in, Indicates the first Energy-like energy at all times Standardized meteorological driving variables, The raw meteorological data, These are the mean and standard deviation of the variable, respectively.

[0118] Example 4:

[0119] A quantitative analysis method for power grid supply situation based on multi-timescale flexibility demand identification, with the same technical content as any one of embodiments 1-3. Further, the short-term power interval prediction model includes a CNN module, an LSTM module, and an output layer with quantile regression branches.

[0120] The CNN module is used to extract local temporal features from meteorological sequences, and the LSTM module is used to characterize the dynamic changes of power over time.

[0121] The output of the CNN module is shown below:

[0122] (2)

[0123] in, For the first Each convolutional channel at time... The output characteristics, The kernel length is 1. As a meteorological feature dimension, For convolution channels; For convolution kernel and bias term;

[0124] The output of the LSTM module is shown below:

[0125] (3)

[0126] in, This represents the input gate, forget gate, and output gate of an LSTM. Candidate memory states, The current memory state, Indicates a hidden state. It is the Sigmoid activation function. Multiply by Hadamard; This is the concatenated input; W and b represent the weights and biases, respectively.

[0127] The output layer of the short-term power range prediction model is shown below:

[0128] (4)

[0129] in, For linear layer weights and bias terms, For the first Energy-like energy at all times Short-term predicted power;

[0130] The loss function for the short-term power range prediction model is shown below:

[0131] (5)

[0132] in, For actual measured power, This is the quantile error function. The loss is multiplied by the method when the actual error is negative, thus obtaining the upper and lower bounds of the prediction under different confidence intervals. For the first Quantile prediction.

[0133] Example 5:

[0134] A quantitative analysis method for power grid supply guarantee based on multi-timescale flexibility demand identification, with the same technical content as any one of embodiments 1-4, further wherein the medium- and long-term power forecasting model adopts the XGBoost model.

[0135] The medium- and long-term electricity forecasting model outputs the medium- and long-term electricity volume. As shown below:

[0136] (6)

[0137] (7)

[0138] in, For the first A tree of return, It is a monthly aggregate vector of meteorological and macroscopic characteristics (such as hydrology, climate, load growth rate, etc.); This is the fitting loss (such as squared loss). To measure the actual power consumption, For complexity regularization; The number of leaf nodes. For node weights, is the regularization coefficient. , The loss function and the actual electricity consumption are given.

[0139] Example 6:

[0140] A quantitative analysis method for power grid supply security based on multi-timescale flexibility demand identification, with technical content identical to any one of Examples 1-5, further comprising the steps of cross-scale alignment and quantile monotonic correction of short-term power curves and medium-to-long-term energy curves, including:

[0141] Step 1.3.1) According to the constraint relationship (8), perform energy alignment on the short-term power curve and the medium- and long-term energy curve at different time scales;

[0142] (8)

[0143] in, For time scale, This is the scaling factor. This refers to the quantile power after energy alignment;

[0144] Step 1.3.2) Monotonicity correction is performed using the quantile projection method, resulting in:

[0145] (9)

[0146] Step 1.3.3) Construct the net load quantile prediction sequence, i.e.:

[0147] (10)

[0148] in, The load quantile power prediction results, These are the power generation quantiles after energy alignment and monotonic correction. .

[0149] Example 7:

[0150] A quantitative analysis method for power grid supply security based on multi-timescale flexibility demand identification, with technical content identical to any one of embodiments 1-6, further comprising the following steps in step 2: calculating short-term and long-term flexibility demands, and calculating flexibility balance and risk assessment indicators:

[0151] Step 2.1) Based on the net load quantile prediction sequence, construct the net load interval, i.e.:

[0152] (11)

[0153] Step 2.2) Calculate the short-term flexibility requirements for uplink and downlink, i.e.:

[0154] (12)

[0155] in, Indicates time arrive The maximum upward adjustment demand between; This indicates a decrease in demand for adjustment;

[0156] Step 2.3) Calculate the long-term flexibility requirements for uplink and downlink, i.e.:

[0157] (13)

[0158] (14)

[0159] In the formula, the time window ; , This indicates the need for long-term flexibility in both upward and downward movements;

[0160] Step 2.4) Calculate the flexibility balance and risk assessment index to evaluate the flexibility and security level of the system;

[0161] Among them, the indicators for flexibility balance and risk assessment include flexibility margin (FM) and conditional value-at-risk (CVaR);

[0162] The flexibility margin (FM) is shown below:

[0163] (15)

[0164] (16)

[0165] (17)

[0166] (18)

[0167] In the formula, It refers to the flexibility that the system can provide in the uplink direction; It refers to the flexibility that the system can provide in the downlink direction; and This refers to the flexibility requirements derived from the aforementioned calculations. If or This indicates a lack of flexibility in that direction. System flexibility and redundancy; when The system has a flexibility gap. When This indicates that the system has sufficient flexibility; when This indicates that the system lacks overall flexibility.

[0168] The conditional value at risk (CVaR) is shown below:

[0169] First, define the value at risk at confidence level α:

[0170] (19)

[0171] (20)

[0172] In the formula, It is the quantile threshold of the flexibility gap at confidence level α; This is the expected flexibility gap under the most unfavorable 1-α percentile scenario; if If so, the system has insufficient average flexibility at confidence level α.

[0173] Example 8:

[0174] A quantitative analysis method for power grid supply security based on multi-timescale flexibility demand identification is provided, with the technical content being the same as any one of Examples 1-7. Further, the objective function of the scheduling optimization model based on time-series production simulation is as follows:

[0175] (twenty one)

[0176] in, For unit operating costs; Cost of energy storage charging and discharging; The penalty coefficient for the flexibility gap; The penalty coefficient for supply gaps.

[0177] Example 9:

[0178] A quantitative analysis method for power grid supply guarantee based on multi-timescale flexibility demand identification, with the same technical content as any one of Examples 1-8. Further, the constraints of the scheduling optimization model based on time-series production simulation include power balance constraints, unit and energy storage output constraints, and flexibility constraints.

[0179] The power balance constraints are as follows:

[0180] (twenty two)

[0181] In the formula, For the unit At any moment Power output (MW); It is the energy storage system at all times The charging and discharging power; It is the slack in the supply and demand balance;

[0182] The power output constraints of the generating unit and energy storage are as follows:

[0183] (twenty three)

[0184] (twenty four)

[0185] in, The energy state of the energy storage system; , These are the upper and lower limits of energy storage for the energy storage system; , For the unit At any moment The upper and lower limits of output;

[0186] The flexibility constraints are as follows:

[0187] (25)

[0188] (26)

[0189] in, This refers to the uplink and downlink regulation capabilities that the generator set can provide. It refers to the uplink and downlink regulation capabilities that energy storage systems can provide; This represents the flexibility gap between uplink and downlink, and is a non-negative variable. When flexibility resources are insufficient, A positive value indicates that standby capacity or purchased electricity is required. This indicates insufficient uplink flexibility, requiring additional issuance or backup. This indicates insufficient downlink flexibility, which may lead to curtailment of solar and wind power. This indicates that the supply-demand balance cannot be maintained, resulting in a supply gap.

[0190] Example 10:

[0191] A quantitative analysis method for power grid supply guarantee situation based on multi-timescale flexibility demand identification, with technical content the same as any one of embodiments 1-9, further wherein the supply guarantee demand represents the weighted compensation amount required by the system to maintain supply and demand balance within the simulation period, i.e.:

[0192] (27)

[0193] In the formula, As weight; It is the slack in the supply and demand balance;

[0194] The flexibility gap ratio represents the proportion of the system's unmet flexibility requirements across all time periods, out of the total demand.

[0195] (28)

[0196] In the formula, For the flexibility gap rate.

[0197] Example 11:

[0198] A quantitative analysis method for power grid supply security based on multi-timescale flexibility demand identification, comprising the following steps:

[0199] Step 1: Multi-timescale power generation forecasting and quantile alignment

[0200] This step involves acquiring historical measured and meteorological forecast data for wind power, solar power, hydropower, and load at a uniform time resolution (15 minutes recommended). A short-term power prediction model (CNN-LSTM) and a medium-to-long-term power prediction model (XGBoost) are then established. Based on this, the P10 / P50 / P90 intervals are obtained through quantile regression (Pinball loss). Energy-shape consistency is then used for cross-scale alignment and quantile monotonicity correction, ultimately forming a net load quantile curve that satisfies both energy closure and quantile monotonicity.

[0201] Standard data acquisition

[0202] First, we obtain historical measured power data for wind power, solar power, hydropower, and loads. Meteorological forecast data and operational status information. Meteorological data includes wind speed, irradiance, temperature, relative humidity, precipitation, and inflow rate; operational information includes unit output, equipment availability, installed capacity, and load level. To ensure temporal consistency and spatial representativeness of the input data, temporal interpolation, missing data imputation, and normalization preprocessing are performed, with a unified time resolution of 15 minutes, resulting in a standardized input sequence:

[0203] (1)

[0204] in, Indicates the first Energy-like energy at all times Standardized meteorological driving variables, The raw meteorological data, These are the mean and standard deviation of the variable, respectively.

[0205] Short-term power range prediction based on CNN-LSTM

[0206] Secondly, a CNN-LSTM model combining convolutional and recurrent structures is used to predict short-term power. The CNN module is used to extract local temporal features from the meteorological sequence, the LSTM module is used to characterize the dynamic changes of power over time, and finally, linear regression is performed to obtain the power.

[0207] CNN layers:

[0208] (2)

[0209] in, For the first Each convolutional channel at time... The output characteristics, The kernel length is 1. As a meteorological feature dimension, For convolution channels; For the convolution kernel and bias term.

[0210] LSTM layer: Let Given the concatenated input, the LSTM recursion is as follows:

[0211] (3)

[0212] in, This represents the input gate, forget gate, and output gate of an LSTM. Candidate memory states, The current memory state, Indicates a hidden state. It is the Sigmoid activation function. Multiply by Hadamard; This is the concatenated input; W and b represent the weights and biases, respectively.

[0213] Output layer:

[0214] (4)

[0215] in, For linear layer weights and bias terms, For the first Energy-like energy at all times Short-term predicted power.

[0216] To characterize prediction uncertainty, this invention sets a quantile regression branch in the output layer of the CNN-LSTM backbone network, and outputs the predicted values ​​of power at different confidence levels (P10, P50, P90). The loss function used is Pinball loss:

[0217] (5)

[0218] in, For actual measured power, This is the quantile error function. The loss is multiplied by the method when the actual error is negative, thus obtaining the upper and lower bounds of the prediction under different confidence intervals. For the first Quantile prediction.

[0219] Medium and long-term electricity forecast

[0220] To ensure energy conservation over medium- to long-term timescales, this step introduces an XGBoost model based on gradient boosting trees to analyze monthly or quarterly electricity consumption. To make predictions, the core calculation expression is:

[0221] (6)

[0222] (7)

[0223] in, For the first A tree of return, It is a monthly aggregate vector of meteorological and macroscopic characteristics (such as hydrology, climate, load growth rate, etc.); This is the fitting loss (such as squared loss). To measure the actual power consumption, For complexity regularization ( The number of leaf nodes. For node weights, (where is the regularization coefficient).

[0224] Energy shape constraint and correction

[0225] To ensure consistency between short-term power curves and medium- to long-term energy levels in terms of energy dimension, energy alignment is performed on results from different time scales. The constraints are as follows:

[0226] (8)

[0227] in, , This is the scaling factor. This represents the quantile power after energy alignment.

[0228] To prevent quantile curves from intersecting, monotonicity correction is performed using the quantile projection method:

[0229] (9)

[0230] This method ensures through interval projection This improves the physical plausibility of the prediction interval.

[0231] Finally, based on the above results, the net load quantile curve is calculated, providing input for subsequent flexibility requirement identification:

[0232] (10)

[0233] in, The load quantile power prediction results, These are the power generation quantiles after energy alignment and monotonic correction. .

[0234] Step 2: Identification and Calculation of Multi-Time-Scale Flexibility Requirements Based on Interval Forecasting

[0235] This step aims to establish a framework for calculating and decomposing flexibility demand across multiple time scales using the net load quantile forecast results output in Step 1. By quantifying the system's uplink and downlink adjustment capacity demand at short-term (15min), medium-term (1h, 4h), and daily (24h) scales, and further combining this with the characteristics of flexibility supply for hierarchical allocation, the system's flexibility balance can be identified and risks measured.

[0236] 2.1. Constructing the Net Load Range

[0237] First, input the net load quantile prediction sequence from step 1, and select... These represent the lower confidence bound, median, and upper confidence bound, respectively. From this, the net load interval can be constructed:

[0238] (11)

[0239] Time of day The width of the net load uncertainty interval is used for subsequent flexibility requirement quantification.

[0240] 2.2. Short-term flexibility requirements

[0241] At a high-resolution (15-minute) scale, flexibility requirements primarily reflect the system's ability to cope with rapid fluctuations in net load. Uplink and downlink flexibility requirements are defined as follows:

[0242] (12)

[0243] in, Indicates time arrive The maximum upward adjustment demand between; These represent the downward adjustment demand, and both reflect the most unfavorable changes in the upper and lower bounds of the system's predictive uncertainty.

[0244] 2.3. Flexibility requirements over longer timescales

[0245] On longer timescales (e.g., 1 hour, 4 hours, 24 hours), uplink / downlink flexibility requirements should reflect the worst-case scenario from earlier to later times (i.e., explicitly consider the "order of occurrence"), and remain non-negative. Let the time window be... .

[0246] Uplink flexibility requirements: in all cases where they are met Among the paired time points, the one with the largest difference between the "upper bound of the later time point" and the "lower bound of the earlier time point" is selected:

[0247] (13)

[0248] Downlink flexibility requirements: in all cases where they are met Among the paired time points, the one with the largest difference between the "upper bound of the earlier time point" and the "lower bound of the later time point" is selected:

[0249] (14)

[0250] 2.4. Calculation of Flexibility Balance and Risk Assessment Indicators

[0251] After obtaining the uplink and downlink flexibility requirements, this invention establishes a system flexibility supply and demand balance model and introduces indicators such as flexibility margin (FM) and conditional value at risk (CVaR) to quantitatively assess the system's flexibility security level.

[0252] Define the flexibility balance between the uplink and downlink directions of the system as follows:

[0253] (15)

[0254] (16)

[0255] in, It refers to the flexibility that the system can provide in the uplink direction; It refers to the flexibility that the system can provide in the downlink direction; and This refers to the flexibility requirements derived from the aforementioned calculations. If or This indicates a lack of flexibility in that direction.

[0256] The overall system flexibility balance is defined as:

[0257] (17)

[0258] when System flexibility and redundancy; when The system has a lack of flexibility.

[0259] Flexibility margin (FM) is used to measure the overall level of flexibility supply and demand matching:

[0260] (18)

[0261] Where the numerator represents the redundancy of flexibility supply; the denominator is the total flexibility requirement of the system; when This indicates that the system has sufficient flexibility; when This indicates that the system has a risk of insufficient overall flexibility.

[0262] To characterize the tail risk of the flexibility gap, we introduce conditional value of risk (CVaR).

[0263] First, define the value at risk at confidence level α:

[0264] (19)

[0265] Then, conditional value at risk is defined:

[0266] (20)

[0267] in, It is the quantile threshold of the flexibility gap at confidence level α; This is the expected flexibility gap under the most unfavorable 1-α percentile scenario; if If the system exhibits insufficient average flexibility at confidence level α, the absolute value of which reflects the gap strength.

[0268] Step 3: Quantitative Assessment Method for Supply Demand Based on Time-Series Production Simulation

[0269] In step 2, this invention obtains the system flexibility supply and demand situation and its statistical characteristics through flexibility demand identification and risk assessment (FM, CVaR). However, the above indicators can only reflect the overall flexibility level of the system in a statistical sense and cannot directly quantify the supply pressure under a specific time series.

[0270] Therefore, this invention further constructs a scheduling optimization model based on Sequential Production Simulation (SPS). By introducing flexibility slack and supply-demand balance slack, it quantitatively evaluates the supply guarantee requirements of the system during time-series operation, realizing the transformation from "statistical evaluation" to "operational quantification".

[0271] 3.1 Overall Approach to Time-Sequence Production Simulation

[0272] Time-series production simulation is a widely used technique in power system analysis, used to simulate the dynamic matching relationship between system load, power output, and constraints over time series. Its core idea is to determine the output decisions of each unit and the system balance state at each moment by minimizing operating costs or flexibility relaxation costs under a given time-series forecasting scenario.

[0273] This invention extends the traditional time-series production simulation (SPS) model by introducing flexibility constraints and the variable of "supply slack," enabling the model to reflect both economic operation and the supply pressure caused by insufficient system flexibility.

[0274] 3.2. Model Inputs and Variable Definitions

[0275] Input from the multi-timescale net load forecast in step 1 ; Flexibility requirements in step 2 Data such as operating constraints of various adjustable power supplies (maximum / minimum output, ramp rate, start / stop limits, etc.). Some decision variables are defined as follows: For the unit At any moment Power output (MW); It is the energy storage system at all times The charging and discharging power (discharging is positive, charging is negative); It is the uplink and downlink flexibility slack (MW), representing unmet flexibility needs; It is the supply-demand balance slack (MW), which represents the supply gap in the system.

[0276] 3.3. Model Constraint Construction

[0277] Power balance constraints:

[0278] (twenty one)

[0279] Unit and energy storage output constraints:

[0280] (twenty two)

[0281] (twenty three)

[0282] in, This refers to the energy state of the energy storage system.

[0283] Flexibility constraints:

[0284] (twenty four)

[0285] (25)

[0286] in, This refers to the uplink and downlink regulation capabilities that the generator set can provide. It refers to the uplink and downlink regulation capabilities that energy storage systems can provide; This represents the flexibility gap (slack) between uplink and downlink, and is a non-negative variable. When flexibility resources are insufficient, A positive value indicates that measures such as activating standby capacity or purchasing external power are needed.

[0287] when This indicates insufficient uplink flexibility, requiring additional issuance or backup; when This indicates insufficient downlink flexibility, which may lead to curtailment of solar and wind power; when This indicates that the supply and demand balance cannot be maintained, resulting in a supply gap.

[0288] 3.4. Optimize the objective function

[0289] Taking into account the costs of power generation, energy storage, and flexibility risks, this invention sets the following objectives:

[0290] (26)

[0291] in, For unit operating costs; Cost of energy storage charging and discharging; The penalty coefficient for the flexibility gap; This is the penalty coefficient for supply gaps. By using a high-weight relaxation penalty, the model prioritizes utilizing existing flexible resources to maintain balance, thereby approximating the operating limits of the real system.

[0292] 3.5. Quantitative Indicators for Supply Guarantee

[0293] After solving the above model, the flexibility slack and supply gap at each time step can be obtained: Based on this, let's define two key quantitative indicators:

[0294] Supply demand:

[0295] (27)

[0296] This represents the weighted compensation amount required by the system to maintain supply and demand balance during the simulation period.

[0297] Flexibility gap rate:

[0298] (28)

[0299] This indicates the proportion of the overall demand that the system's flexibility is not met across all time periods.

[0300] when or A sustained increase indicates increased pressure on system flexibility and heightened risks to supply security.

[0301] This invention introduces the quantile interval prediction method into the field of power grid flexibility analysis. By extracting the net load quantile envelope at different time scales, it identifies the characteristics of uplink and downlink flexibility demand changes at multiple time scales such as 15min, 1h, 4h, and 24h.

[0302] Compared with traditional flexibility calculation methods based on single-point prediction, this mechanism can characterize the dynamic adjustment pressure brought about by prediction uncertainty, and improve the accuracy and robustness of flexibility demand identification.

[0303] (2) Calculation method for non-negative flexibility requirements under time-ordered constraints: In view of the directional ambiguity in the calculation of flexibility requirements at different time scales, this invention proposes a time-ordered, non-negative constraint-based flexibility requirement identification formula.

[0304] By explicitly introducing time sequence conditions during the calculation process The non-negative operator max(0,⋅) ensures that both upward and downward demands always have clear temporal logic and physical meaning. This method can uniformly define flexibility requirements at any time scale, achieving consistent cross-scale modeling from short-term fluctuations to medium- to long-term upslopes.

[0305] This invention constructs a flexibility supply-demand balance model and introduces two types of indicators: Flexibility Margin (FM) and Conditional Value at Risk (CVaR), forming a two-dimensional security assessment system that can simultaneously characterize both "average level" and "tail risk." FM measures the overall system flexibility supply-demand matching degree, while CVaR is used to capture extreme risk scenarios of flexibility gaps. This system compensates for the lack of risk characterization in traditional flexibility assessments, enabling power grid flexibility analysis to move from static mean evaluation to dynamic probabilistic description.

[0306] This invention innovatively introduces two types of variables, "flexibility slack" and "supply-demand balance slack," into the traditional time-series production simulation (SPS) framework. This is achieved by adding upstream and downstream flexibility gap terms to the constraint equations. Supply gap items This improvement enables a feasible solution for the system's operating state under conditions of insufficient flexibility supply. It not only allows the model to reflect the limits of the system's adjustment capacity but also quantifies the flexibility gap and supply pressure by adjusting the slack, providing data for subsequent compensation and optimization.

[0307] This invention establishes a quantitative indicator chain from flexibility demand identification to supply guarantee demand assessment, including the flexibility gap rate. Supply demand The invention includes indicators such as flexibility margin (FM) and conditional risk index (CVaR). Through the synergistic calculation of these indicators, the invention can quantitatively characterize the system's flexibility and security level and the evolution trend of supply guarantee under multiple time scales and scenarios.

Claims

1. A quantitative analysis method for power grid supply security based on multi-timescale flexibility demand identification, characterized in that, Includes the following steps: Step 1) Construct a net load quantile prediction sequence that satisfies energy closure and quantile monotonicity; Step 2) Based on the net load quantile forecast sequence, calculate the short-term flexibility requirement and the long-term flexibility requirement, and calculate the flexibility balance and risk assessment index. Step 3) Using the net load quantile forecast sequence, short-term flexibility demand, and long-term flexibility demand as inputs, solve the scheduling optimization model based on time-series production simulation to obtain the flexibility slack and supply gap at each time point. Step 4) Calculate the supply demand and flexibility gap rate based on the flexibility slack and supply gap at each moment; The overall flexibility level of the power grid is determined by flexibility balance and risk assessment indicators, and the supply guarantee demand of the power grid during time-series operation is determined by supply guarantee demand and flexibility gap rate.

2. The method for quantitative analysis of power grid supply situation based on multi-timescale flexibility demand identification as described in claim 1, characterized in that, Step 1, the steps for constructing the net load quantile prediction sequence that satisfies energy closure and quantile monotonicity, include: Step 1.1) Under a unified time resolution, acquire input data, including historical measured data of wind power, photovoltaic, hydropower and load, meteorological forecast data and operating status information; Step 1.2) Based on the input data, use the short-term power prediction model to predict the short-term power and obtain the short-term power curve; Based on the input data, the medium- and long-term electricity consumption forecasting model is used to predict the medium- and long-term electricity consumption, and the medium- and long-term electricity consumption curve is obtained. Step 1.3) Use energy-shape consistency to perform cross-scale alignment and quantile monotonic correction on short-term power curves and medium- to long-term charge curves to form a net load quantile prediction sequence that satisfies energy closure and quantile monotonicity.

3. The method for quantitative analysis of power grid supply situation based on multi-timescale flexibility demand identification as described in claim 2, characterized in that, Meteorological data include wind speed, irradiance, temperature, relative humidity, precipitation, and water flow. Operating status information includes unit output, equipment availability, installed capacity, and load level; The input data also undergoes preprocessing, including time interpolation, missing value imputation, and normalization preprocessing; The preprocessed input data is represented as follows: （1） in, Indicates the first Energy-like energy at all times Standardized meteorological driving variables, This is the raw meteorological data. These are the mean and standard deviation of the variable, respectively.

4. The method for quantitative analysis of power grid supply situation based on multi-timescale flexibility demand identification as described in claim 2, characterized in that, The short-term power interval prediction model includes a CNN module, an LSTM module, and an output layer with a quantile regression branch. The CNN module is used to extract local temporal features from meteorological sequences, and the LSTM module is used to characterize the dynamic changes of power over time. The output of the CNN module is shown below: （2） in, For the first Each convolutional channel at time... The output characteristics, The kernel length is 1. As a meteorological feature dimension, For convolution channels; For convolution kernel and bias term; The output of the LSTM module is shown below: （3） in, This represents the input gate, forget gate, and output gate of an LSTM. Candidate memory states, The current memory state, Indicates a hidden state. It is the Sigmoid activation function. Multiply by Hadamard; This is the concatenated input; W and b represent the weights and biases, respectively. The output layer of the short-term power range prediction model is shown below: （4） in, For linear layer weights and bias terms, For the first Energy-like energy at all times Short-term predicted power; Loss function of short-term power range prediction model As shown below: （5） in, For actual measured power, This is the quantile error function; For the first Quantile prediction.

5. The method for quantitative analysis of power grid supply situation based on multi-timescale flexibility demand identification as described in claim 2, characterized in that, The medium- and long-term electricity prediction model adopts the XGBoost model. The medium- and long-term electricity forecasting model outputs the medium- and long-term electricity volume. As shown below: （6） （7） in, For the first A tree of return, This is a monthly aggregate vector representing meteorological and macroscopic characteristics; For fitting loss, To measure the actual power consumption, For complexity regularization; The number of leaf nodes. For node weights, The regularization coefficient; , The loss function and the actual electricity consumption are given.

6. The method for quantitative analysis of power grid supply situation based on multi-timescale flexibility demand identification according to claim 2, characterized in that, The steps for cross-scale alignment and quantile monotonic correction of short-term power curves and medium- to long-term energy curves include: Step 1.3.1) According to the constraint relationship (8), perform energy alignment on the short-term power curve and the medium- and long-term energy curve at different time scales; （8） in, For time scale, This is the scaling factor. This refers to the quantile power after energy alignment; Step 1.3.2) Monotonicity correction is performed using the quantile projection method, resulting in: （9） Step 1.3.3) Construct the net load quantile prediction sequence, i.e.: （10） in, The load quantile power prediction results, These are the power generation quantiles after energy alignment and monotonic correction. .

7. The method for quantitative analysis of power grid supply situation based on multi-timescale flexibility demand identification as described in claim 1, characterized in that, Step 2, which involves calculating short-term and long-term flexibility requirements and determining the flexibility balance and risk assessment metrics, includes the following steps: Step 2.1) Based on the net load quantile prediction sequence, construct the net load interval, i.e.: （11） Step 2.2) Calculate the short-term flexibility requirements for uplink and downlink, i.e.: （12） in, Indicates time arrive The maximum upward adjustment demand between; This indicates a decrease in demand for adjustment; Step 2.3) Calculate the long-term flexibility requirements for uplink and downlink, i.e.: （13） （14） In the formula, the time window ; , This indicates the need for long-term flexibility in both upward and downward movements; Step 2.4) Calculate the flexibility balance and risk assessment index to evaluate the flexibility and security level of the system; Among them, the indicators for balancing flexibility and risk assessment include flexibility margin and conditional value of risk; The flexibility margin (FM) is shown below: （15） （16） （17） （18） In the formula, It refers to the flexibility that the system can provide in the uplink direction; It refers to the flexibility that the system can provide in the downlink direction; and It is a need for flexibility; if or This indicates a lack of flexibility in that direction; System flexibility and redundancy; when The system has a flexibility gap; when This indicates that the system has sufficient flexibility; when This indicates that the system lacks overall flexibility. The conditional value at risk (CVaR) is shown below: First, define the value at risk at confidence level α: （19） （20） In the formula, It is the quantile threshold of the flexibility gap at confidence level α; This is the expected flexibility gap under the most unfavorable 1-α percentile scenario; if If so, the system has insufficient average flexibility at confidence level α.

8. The method for quantitative analysis of power grid supply situation based on multi-timescale flexibility demand identification as described in claim 1, characterized in that, The objective function of the scheduling optimization model based on time-series production simulation is shown below: （21） in, For unit operating costs; Cost of energy storage charging and discharging; The penalty coefficient for the flexibility gap; The penalty coefficient for supply gaps; This represents a gap in flexibility for both upward and downward traffic.

9. The method for quantitative analysis of power grid supply situation based on multi-timescale flexibility demand identification as described in claim 1, characterized in that, The constraints of the scheduling optimization model based on time-series production simulation include power balance constraints, unit and energy storage output constraints, and flexibility constraints. The power balance constraints are as follows: （22） In the formula, For the unit At any moment contribution; It is the energy storage system at all times The charging and discharging power; It is the slack in the supply and demand balance; The power output constraints of the generating unit and energy storage are as follows: （23） （24） in, The energy state of the energy storage system; , These are the upper and lower limits of energy storage for the energy storage system; , For the unit At any moment The upper and lower limits of output; The flexibility constraints are as follows: （25） （26） in, This refers to the uplink and downlink regulation capabilities that the generator set can provide. It refers to the uplink and downlink regulation capabilities that energy storage systems can provide; The uplink and downlink flexibility gap is represented by a non-negative variable; when flexibility resources are insufficient, A positive value indicates that standby capacity or purchased electricity is required. This indicates insufficient uplink flexibility, requiring additional issuance or backup. This indicates insufficient downlink flexibility and a probability of curtailment of solar and wind power. This indicates that the supply-demand balance cannot be maintained, resulting in a supply gap.

10. A method for quantitative analysis of power grid supply situation based on multi-timescale flexibility demand identification as described in claim 1, characterized in that, The supply demand quantity represents the weighted compensation amount required by the system to maintain supply and demand balance during the simulation period. ,Right now: （27） In the formula, As weight; It is the slack in the supply and demand balance; The flexibility gap ratio represents the proportion of the system's unmet flexibility requirements across all time periods, out of the total demand. （28） In the formula, For the flexibility gap rate.

Images