New energy output fluctuation-oriented multi-time scale transient stability control method and device

Abstract

The invention discloses a new energy output fluctuation-oriented multi-time scale transient stability control method and device. The method comprises the steps of determining a new energy historical output sequence, meteorological features, time features, historical fluctuation features of new energy output and a feature matrix; determining a new energy output predicted value at each moment in a preset time period based on the feature matrix; the new energy output predicted value after the target moment is corrected based on the new energy real output value at the target moment, and a corrected new energy output predicted value is obtained; performing system transient risk assessment based on the corrected new energy output predicted value and the fluctuation characteristics of the new energy output at the target moment to obtain a risk assessment result; and performing intra-day resource adjustment based on a risk assessment result to obtain an intra-day resource adjustment scheme, and determining a corresponding control measure based on the intra-day resource adjustment scheme and the new energy fluctuation scene at the target moment. Therefore, transient instability can be effectively inhibited while resources are efficiently utilized.

Description

TECHNICAL FIELD

[0001] The embodiment of the present application relates to the technical field of power systems, in particular to a multi-time scale transient stability control method and device for new energy output fluctuation. BACKGROUND

[0002] With the continuous increase of the penetration rate of wind power, photovoltaic and other new energy, the randomness and volatility of new energy output pose a serious challenge to the transient stability of the power system. The volatility of new energy output easily leads to transient instability. The existing technical solutions are difficult to balance between quickly solving transient instability and efficient resource utilization. SUMMARY

[0003] The embodiment of the present application provides a multi-time scale transient stability control method and device for new energy output fluctuation, which can effectively cope with the problem of transient instability caused by new energy output fluctuation, and can effectively suppress transient instability while efficiently utilizing resources.

[0004] In a first aspect, the embodiment of the present application provides a multi-time scale transient stability control method for new energy output fluctuation, comprising:

[0005] Determine the historical output sequence of new energy, weather characteristics, time characteristics and historical fluctuation characteristics of new energy output, and construct a feature matrix based on the historical output sequence of new energy, the weather characteristics, the time characteristics and the historical fluctuation characteristics;

[0006] Determine the new energy output prediction value of each time in the preset period based on the feature matrix;

[0007] If the deviation of the new energy output prediction value of the target time and the new energy real output value of the target time does not meet the preset condition, correct the new energy output prediction value after the target time based on the new energy real output value of the target time to obtain the corrected new energy output prediction value;

[0008] Perform system transient risk assessment based on the corrected new energy output prediction value and the fluctuation characteristics of the new energy output of the target time to obtain a risk assessment result;

[0009] Perform intraday resource adjustment based on the risk assessment result to obtain an intraday resource adjustment scheme, and determine the corresponding control measures based on the intraday resource adjustment scheme and the new energy fluctuation scenario of the target time.

[0010] In a second aspect, the embodiment of the present application provides a multi-time scale transient stability control device for new energy output fluctuation, comprising:

[0011] The data preprocessing module is configured to determine a new energy historical output sequence, meteorological features, time features, and historical fluctuation features of new energy output, and construct a feature matrix based on the new energy historical output sequence, the meteorological features, the time features, and the historical fluctuation features.

[0012] The new energy output prediction module is configured to determine a new energy output prediction value at each time in a preset time period based on the feature matrix.

[0013] The correction module is configured to, if a deviation between the new energy output prediction value at the target time and the actual new energy output value at the target time does not satisfy a preset condition, correct the new energy output prediction value after the target time based on the actual new energy output value at the target time to obtain a corrected new energy output prediction value.

[0014] The transient risk assessment module is configured to perform system transient risk assessment based on the corrected new energy output prediction value and the fluctuation features of the new energy output at the target time to obtain a risk assessment result.

[0015] The adjustment control module is configured to perform intraday resource adjustment based on the risk assessment result to obtain an intraday resource adjustment scheme, and determine a corresponding control measure based on the intraday resource adjustment scheme and the new energy fluctuation scenario at the target time.

[0016] In a third aspect, an electronic device is provided, including a memory and a processor, the memory storing a computer program, and the processor executing the computer program to implement the method provided in the embodiments of the present application.

[0017] In a fourth aspect, a computer readable storage medium is provided, storing a computer program, and when the computer program is executed in a computer, the computer is caused to execute the method provided in the embodiments of the present application.

[0018] The technical scheme provided in the embodiments of the present application can improve the prediction accuracy, effectively solve the problem of insufficient sensitivity to new energy fluctuation, perform transient risk assessment based on the corrected new energy prediction value and the new energy output fluctuation feature of the target moment, obtain a risk assessment result, perform intraday resource adjustment based on the risk assessment result, obtain an intraday resource adjustment scheme, determine the corresponding control measures based on the intraday resource adjustment scheme and the new energy fluctuation scenario of the target moment, realize efficient resource utilization, and effectively solve the problem of transient instability, thereby realizing dual optimization of transient stability control and resource utilization. BRIEF DESCRIPTION OF DRAWINGS

[0019] Figure 1 A multi-time scale transient stability control method flow chart for new energy output fluctuation is provided in the embodiments of the present application.

[0020] Figure 2 A multi-time scale transient stability control device structure block diagram for new energy output fluctuation is provided in the embodiments of the present application.

[0021] Figure 3 An electronic device structure schematic diagram is provided in the embodiments of the present application. DETAILED DESCRIPTION

[0022] The present application will be further described in detail through the accompanying drawings and specific embodiments.

[0023] Figure 1 A multi-time scale transient stability control method flow chart for new energy output fluctuation is provided in the embodiments of the present application. The method can be executed by a multi-time scale transient stability control device for new energy output fluctuation. The device can be realized by software and / or hardware, and can be configured in a computer or other electronic device.

[0024] As shown in Figure 1 The technical scheme provided in the embodiments of the present application includes the following steps:

[0025] S110: Determine the new energy historical output sequence, the meteorological feature, the time feature and the historical fluctuation feature of the new energy output, and construct a feature matrix based on the new energy historical output sequence, the meteorological feature, the time feature and the historical fluctuation feature.

[0026] In this embodiment, data can be acquired from multiple data sources through the data acquisition module. These data sources include: New energy source side: wind power output can be acquired through wind power grid connection point power sensors; photovoltaic power output can be acquired through photovoltaic grid connection point power sensors; wind speed can be acquired through anemometers; and sunlight intensity can be acquired through light sensors. Grid side: node voltage can be acquired through bus node voltage transformers; line current can be acquired through current transformers; and system frequency can be acquired through synchronous phasor measurement devices. Environmental side: temperature, cloud cover, etc., can be acquired from meteorological stations. Historical side: records of new energy output, load, and transient events (such as historical maximum fluctuations) from the database can be retrieved over the past year. By acquiring raw data such as new energy output, grid operation data, and environmental data from multiple data sources, historical new energy output sequences, meteorological characteristics, temporal characteristics, and historical fluctuation characteristics of new energy output can be extracted from the raw data, forming a feature matrix.

[0027] In this embodiment, steps S110-S130 can be determined using a new energy output prediction model. Specifically, the four types of features mentioned above can be filtered through the data preprocessing module in the new energy output prediction model to form a feature matrix. .in: The historical power output sequence of new energy sources is obtained by selecting power output data at the 15-minute level for the first 7 days to form a continuous time series sample. Meteorological characteristics (including environmental parameters that affect the output of new energy sources, such as wind speed, light intensity, temperature, and cloud cover). Time characteristics (including hourly and weekday / weekend identifiers, used to capture intraday periodic patterns); Historical fluctuation characteristics (extracting the maximum drop in the corresponding period of the previous 7 days to enhance sensitivity to fluctuations).

[0028] Among these methods, meteorological characteristics can be normalized using formulas. The data is compressed to the interval [0, 1] to eliminate the interference of dimensional differences on the training of the system energy output prediction model, where: This is the original meteorological characteristic data; , These are the historical minimum and maximum values ​​for this meteorological characteristic, respectively. These are the standardized meteorological characteristics.

[0029] Among them, the hour feature in the time feature uses sine and cosine coding, through the formula and ( (For hours, 0-23) This transforms discrete time information into a continuous vector, more accurately capturing intraday periodicity. a sine encoding value of the hour feature, reflecting the periodic variation trend of the hour in a 24-hour cycle, a cosine encoding value of the hour feature, and together constitute a two-dimensional continuous vector, avoiding "distance distortion" between discrete hours.

[0030] S120: determining a new energy output prediction value of each time in a preset period based on the feature matrix.

[0031] In this embodiment, the output prediction module in the new energy output prediction model can be used to obtain the new energy output prediction value of each time in the preset period based on the feature matrix, where the output prediction module can be an improved Long Short-Term Memory (LSTM) prediction module. The preset period can be a future period starting from the prediction time point. For example, if the prediction starts at 11 o'clock, the prediction time point is 11 o'clock, and the preset period can be the next 4 hours starting from 11 o'clock, i.e. 11 o'clock-15 o'clock.

[0032] In this embodiment, the attention mechanism is integrated into the traditional LSTM to improve the ability of the model to capture key time sequence features and build a network structure suitable for the output prediction module.

[0033] In this embodiment, for the basic LSTM layer, an LSTM unit containing an input gate, a forget gate, a cell state, and an output gate can be used to realize long-term memory and short-term processing of time sequence features through a formula group:

[0034]

[0035] where, is a sigmoid activation function, various weight matrices (such as , , , , , ) and bias terms (such as , , , ) are optimized through subsequent training; is the input feature vector at time t, which is composed of the row vector corresponding to time t in the feature matrix; is the input feature weight matrix of the input gate, which is used to quantify the influence degree of the input feature vector output by the input gate on the output of the input gate; is the hidden state of the LSTM unit at time t-1; is the hidden state weight matrix of the input gate, which is used to quantify the hidden state at time t-1 the degree of influence on the output of the input gate; a bias vector of the input gate, used to adjust the output reference of the input gate; an output vector of the input gate at time t, with a value range of [0, 1], used to control the proportion of new information entering the cell state; an input feature weight matrix of the forget gate, used to quantify the input features at time t the degree of influence on the output of the forget gate; a hidden state weight matrix of the forget gate, used to quantify the hidden state at time t-1 the degree of influence on the output of the forget gate; a bias vector of the forget gate, used to adjust the output reference of the forget gate; an output vector of the forget gate (Forget Gate), with a value range of [0, 1], used to control the proportion of historical information retained in the cell state; an input feature weight matrix of the cell state, used to quantify the input features at time t the degree of influence on the update of the cell state; a hidden state weight matrix of the cell state, used to quantify the hidden state at time t-1 the degree of influence on the update of the cell state; a bias vector of the cell state, used to adjust the update reference of the cell state; tanh is a hyperbolic tangent activation function, with a value range of [-1, 1], used to perform nonlinear transformation on the candidate update value of the cell state; a cell state vector at time t-1, storing the long-term memory information at time t-1; a cell state vector at time t, updated through the synergistic effect of the forget gate and the input gate, storing the long-term memory information at time t; an input feature weight matrix of the output gate, used to quantify the input features at time t the degree of influence on the output of the output gate; a hidden state weight matrix of the output gate, used to quantify the hidden state at time t-1 the degree of influence on the output of the output gate; a bias vector of the output gate, used to adjust the output reference of the output gate; an output vector of the output gate at time t, with a value range of [0, 1], used to control the output proportion of the cell state to the hidden state; a hidden state of the LSTM unit at time t, comprehensively reflecting the short-term key information at time t.

[0036] For attention mechanism embedding, in order to improve the attention of the new energy output prediction model to the recent data (the recent data has greater influence on short-term prediction), the attention mechanism is added to the hidden state sequence output by the LSTM unit. The attention weight of the hidden state at time t is calculated by formula The attention weight of the hidden state at time t is calculated by formula ; wherein, is the energy value of the hidden state at time t; is the energy value, , is the weight matrix, is the bias term, is the sequence length; is the energy value of the hidden state at time k, and k is in the range of [1, T]; and the weighted and fused hidden state is generated by to strengthen the effect of key time sequence features.

[0037] For output layer construction, the hidden state fused with the attention mechanism is input into the full connection layer, and the new energy output prediction value at time t is generated by formula is the output layer weight matrix, is the bias term), to complete the conversion from features to prediction results.

[0038] S130: If the deviation of the new energy output prediction value at the target time from the corresponding new energy real output value does not meet the preset condition, the new energy output prediction value after the target time is corrected based on the new energy real output value at the target time, to obtain a corrected new energy output prediction value.

[0039] In this embodiment, to cope with the deviation between the new energy real output and the new energy prediction value, the future short-term prediction result can be corrected based on the latest data every hour, to improve the prediction timeliness. Specifically, the new energy real output value at the target time can be compared with the new energy output prediction value at the corresponding time to calculate the deviation. When the deviation exceeds the preset threshold, the prediction correction mechanism is triggered. For example, a feature matrix can be constructed based on the historical data of 7 days (time granularity is 15 minutes) to obtain the new energy output prediction value at each time in the future 4 hours, for example, the new energy output prediction value from 11 o'clock to 15 o'clock. If the current time is the target time, which is 12 o'clock, the new energy real output value at the current time is called, and the new energy output prediction value from 13 o'clock to 15 o'clock in the future is corrected based on the new energy real output value at the current time.​​​

[0040] In the embodiment, the future new energy output prediction value can be updated by using the weighted fusion method. Optionally, the new energy real output value based on the target time point corrects the new energy output prediction value after the target time point, including:

[0041] The new energy output prediction value is corrected based on the following formula:

[0042]

[0043] wherein, is the corrected new energy output prediction value at t time point, is the new energy real output value at t time point; is the new energy real output value at t time point; is the weight coefficient of the new energy output prediction value at t time point; is the weight coefficient of the new energy real output value at t time point; is the weight coefficient of the new energy real output value at t time point; is the time interval; wherein, may be greater than wherein, may be 0.7, may be 0.3, wherein the weight coefficient can be dynamically adjusted according to the actual deviation (the greater the deviation, the higher the weight of the new energy real output value); is the time interval parameter, which can be 1 hour.

[0044] In the embodiment, before step S120, the output prediction module can be iteratively trained by using the training set data, the parameters of the output prediction module are adjusted based on the loss function; the output prediction module is verified by using the verification set data, if the deviation between the loss function value corresponding to the verification set data and the loss function value corresponding to the training set data exceeds the preset deviation, the feature matrix is adjusted; wherein, the loss function is:

[0045]

[0046] wherein, is the loss function value; is the new energy output prediction value at t time point; is the new energy real output value at t time point; is the new energy output prediction value at t+1 time point; is the fluctuation penalty coefficient.

[0047] Specifically, the parameters of the new energy output prediction model can be trained and adjusted by the model training and optimization module in the new energy output prediction model to reduce the prediction error and avoid overfitting, and ensure the generalization ability of the model in the actual scene. The loss function adopts a composite loss function containing a volatility penalty term, which controls the overall prediction bias and enhances the ability to capture output volatility. In the loss function, the first term is the sum of squared errors between the new energy output prediction value and the corresponding actual new energy output value, which can be used to control the overall bias, and the second term is the sum of absolute values of the volatility of the new energy output prediction value, which can be used to enhance the sensitivity of the model to output mutations through a volatility penalty coefficient (need to be set according to the scene) to improve the sensitivity of the model to output mutations; 96 is the number of 15-minute granularity time points within 24 hours, which ensures coverage of the entire daily cycle.

[0048] In this embodiment, the parameters of the output prediction module can be iteratively updated (learning rate and iteration number are adjusted according to training effect) using the Adam optimizer, and an early stopping mechanism is introduced: if the loss function value of the validation set does not decrease for 10 consecutive iterations, the training can be stopped to avoid overfitting the training data. In the training process, the error changes of the training set and the validation set are monitored in real time, and if the deviation between the two is too large, the feature matrix input in the feature preprocessing link is returned to be adjusted, until the output prediction module converges and the generalization ability meets the standard. Specifically, the features in the feature matrix can be increased, and the increased features can be real-time weather warning signals; or the feature granularity can be refined, for example, the data granularity can be split into 10-minute granularity; or cross-features can be added, for example, "hour x season x weather type" can be added to increase the information quantity. It should be noted that the above training process is mainly for parameter adjustment of the output prediction module (improved LSTM module) in the new energy output prediction model.

[0049] S140: Based on the corrected new energy output prediction value and the fluctuation characteristics of the new energy output at the target time, the system transient risk assessment is performed to obtain a risk assessment result.

[0050] In this embodiment, optionally, the system transient risk assessment based on the corrected new energy output prediction value and the fluctuation characteristics of the new energy output at the target time includes: determining the power angle stability margin and the voltage stability margin based on the corrected new energy output prediction value; determining the risk level based on the power angle stability margin, the voltage stability margin, and the fluctuation characteristics of the new energy output at the target time.

[0051] In the embodiment, the transient risk assessment module can quantify the system transient stability state and divide the risk level based on the corrected new energy output prediction value, provide basis for subsequent resource optimization. Specifically, 1) prediction data receiving and preprocessing: the relevant data output by the new energy output prediction model can be received and preliminarily processed to lay a foundation for subsequent evaluation. 2) Data receiving: obtain the corrected new energy output prediction value , and the fluctuation characteristics of the target moment (for example, the current moment) new energy output, including the maximum drop amount (unit: MW) and the fluctuation rate (unit: MW / min). 3) Data verification: check the integrity of the data, if there is a missing (such as the missing of part of the time period ), use linear interpolation method to fill in; at the same time, check the rationality of the data, for example, the maximum drop amount should not exceed 50% of the total installed capacity of new energy, if it exceeds, it is marked as abnormal and fed back to the new energy output prediction model. 4) Data format conversion: convert the corrected new energy output prediction value into a format matched with the power grid topology data, to ensure that the subsequent power flow calculation and other links can be normally called.

[0052] In the embodiment, the system power angle stability state can be evaluated through equivalent simplification and energy calculation. Among them, the system equivalent simplification can use the extended equal-area criterion to simplify the multi-machine power system into an equivalent two-machine model composed of "critical units" and "remaining units", highlighting the units that play a key role in transient stability. Among them, the energy area calculation can calculate the acceleration area (unit: MW s) absorbed by the critical unit in the acceleration stage and the deceleration area (unit: MW s) released in the deceleration stage. Among them, the acceleration area reflects the excess energy obtained by the unit due to the disturbance, and the deceleration area reflects the ability of the system to suppress acceleration. Among them, the power angle stability margin

[0053] (unit: %) can be calculated according to the formula: The greater the value, the better the system power angle stability performance, when , , the system transient power angle stability.

[0054]

[0055]

[0056] Among them, is the predicted new energy output fluctuation amount, is the acceleration time; is the new energy fluctuation prediction model, the predicted output value 1s after the disturbance occurs;​ Available energy storage power based on predicted risk configuration Let be the deceleration time. Wherein, the predicted output value of the new energy source at time t... Forecast value of new energy power output at time t+1 Predicting fluctuations in new energy output .

[0057] In this embodiment, the system voltage stability level is assessed by combining the grid topology with the corrected renewable energy output forecast. Specifically, the assessment can be based on the corrected renewable energy output forecast. Load data and acquired power grid topology data (such as line impedance, transformer parameters, etc.) are used to perform power flow calculations to obtain the voltage at each bus node. (Unit: pu). From the calculated node voltages, the lowest voltage value is selected. (Unit: pu). Based on the formula

[0058] Calculate voltage stability margin (Unit: %), of which pu represents the rated voltage. The larger the value, the more sufficient the system voltage stability reserve.

[0059] Specifically, the risk level of the system can be determined by comprehensively considering the power angle stability margin, voltage stability margin, and fluctuation characteristics of renewable energy output. The power angle stability margin can be used as a starting point. Voltage stability margin The maximum drop in new energy output is used as the core basis for risk level classification.

[0060] Specifically, when , And the maximum drop in power output of new energy sources When the risk level is MW (Modular Motion), the system is considered low-risk, indicating good transient stability and an extremely low probability of instability. When the 12% threshold is met... 15% Or 10MW Maximum drop in power output of new energy sources If any of the MW (Modular, Variable, and Variable) indicators is classified as medium risk, the system has a certain risk of transient instability and requires appropriate reserves of stability resources; when , When the maximum power output drop of new energy sources exceeds 20MW, it is considered high-risk, indicating a high risk of system transient instability. Priority should be given to allocating stable resources and preparing control measures. The risk level, , and corresponding judgment basis is sorted into a risk report. The risk assessment result can include power angle stability margin, voltage stability margin, and corresponding risk level, etc.

[0061] S150: Based on the risk assessment result, intra-day resource adjustment is performed to obtain an intra-day resource adjustment scheme, and a corresponding control strategy is determined based on the intra-day resource adjustment scheme and the new energy fluctuation scenario of the target time.

[0062] In this embodiment, the evaluation result is transmitted to an optimization model, and relevant data is stored for subsequent analysis. The power angle stability margin , the voltage stability margin , and the risk level can be transmitted to the optimization model. Various data generated in this evaluation process, including original prediction data, calculated energy area, voltage value, and risk level, etc., are stored in a database to provide data support for post-evaluation and iteration.

[0063] In this embodiment, the risk assessment result can be used to perform intra-day resource adjustment to obtain an intra-day resource adjustment scheme, including defining the power angle stability margin, the voltage stability margin, the maximum new energy output drop, the energy storage reserve capacity, and the load shedding reserve to define the state of the power system, defining the energy storage reserve adjustment capacity and the load shedding reserve adjustment amount to define the execution action of the power system, and establishing a reward function based on the power angle stability margin, the voltage stability margin, and the control cost of the intra-day resource adjustment scheme. The state of the power system at the target time and the reward function are used to determine the optimal execution action of the power system as the intra-day resource adjustment scheme. The energy storage reserve adjustment capacity satisfies the constraint within a first preset range, and the load shedding reserve adjustment amount satisfies the constraint within a second preset range.

[0064] The reward function is

[0065] The power angle stability margin is The voltage stability margin is The control cost of the intra-day resource adjustment scheme is The maximum allowed control cost is The reward value is

[0066] ​​In this embodiment, an optimization model can be used to optimize the allocation of schedulable resources based on risk assessment results, reducing costs while ensuring system stability. Specifically, the types of resources available for stability control and related parameters can be identified first, providing basic data for resource optimization. Resource type determination involves identifying schedulable stable resources, primarily including energy storage systems and controllable loads. Key parameters for various resources are obtained from the data acquisition module. The parameters for energy storage systems may include: total capacity. (Unit: MWh), Maximum charge / discharge power (Unit: MW) and standby cost coefficient (Unit: Yuan / kWh). The parameters for controllable load may include: total shelvable load. (Unit: MW), Unit Compensation Cost (Unit: Yuan / MW) h). Resource status verification: Check the current status of various resources. For example, the state of charge (SOC) of the energy storage system should be within a reasonable range (e.g., 20%-80%). If it exceeds the range, record it and consider it in subsequent optimizations.

[0067] In this embodiment, the power system status is collected once every hour to obtain the latest status variables from relevant modules or models and verify their validity, providing reliable input for subsequent adjustments, as shown in Table 1.

[0068] Table 1 Power System Status Data Table

[0069]

[0070] In this embodiment, if any of the above variables exceeds a reasonable range, an exception handling mechanism is triggered. Specifically, if... or If the risk is identified as "emergency instability risk," skip routine adjustments and directly trigger the emergency measures of the real-time collaborative control module (such as emergency energy storage discharge); if the energy storage SOC < 20%, mark it. Given the "low capacity" status, this adjustment only allows for increases in capacity. (Reduction is prohibited), and the maximum adjustment amount is ≤ 10% of the total capacity; if Data missing: Use linear interpolation (based on the fluctuation trend of the previous 30 minutes) to fill in the missing data; the error must be ≤5%. The largest drop in energy output.

[0071] In this embodiment, the current transient risk level (consistent with the transient risk assessment module level standard) can be mapped first based on the real-time state variables mentioned above, and then the priority of resource adjustment can be determined.

[0072] Among them, combining 3 core indicators ( , , Risks are determined according to the rules shown in Table 2.

[0073] Table 2 Risk Level Information Table

[0074]

[0075] In this embodiment, the actions performed by the power system have a certain priority; specifically, the actions to be performed... ( Adjusting capacity for energy storage backup The priority rule for adjusting the available load reserve capacity can be: prioritize adjusting the energy storage reserve capacity. Because "the unit stabilization cost of energy storage is lower than the load shedding cost" (e.g., energy storage cost is 0.5 yuan / kWh, and load shedding compensation is 200 yuan / MW). h), only when Reaching capacity limit ( Only when the SOC is insufficient should the load be adjusted for standby. The adjustment range strictly adheres to the ±5% constraint. , That is, the energy storage reserve adjustment capacity meets the constraint of a first preset range, and the load shelving reserve adjustment amount meets the constraint of a second preset range. The target time is the energy storage reserve capacity, which is the current time. The available load is reserved for the target time, i.e., the current time.

[0076] In this embodiment, the reward value can be calculated based on the real-time status of the power system. The optimal resource adjustment amount is determined by "maximizing the reward value," ensuring that the intraday resource adjustment plan simultaneously satisfies "stable improvement" and "controllable cost." In the above reward function, the parameter 0.6 is the weight coefficient of stability margin, and since transient stability is the core objective, its weight is higher than that of cost (0.4). Considering both stability margins, the minimum values ​​of the power angle stability margin and voltage stability margin are chosen to avoid situations where one indicator fails while another becomes unstable (e.g., ...). but (The overall stability margin is calculated at 12%). The parameter 0.4 is the weighting coefficient for the cost ratio: taking into account economic efficiency and avoiding excessive resource allocation that could lead to a surge in costs. Cost control for intraday resource adjustment plan (RMB): ;in, For energy storage backup costs, The cost of load shedding compensation all comes from the data acquisition module. The maximum allowed control cost (yuan) can be determined based on the day-ahead optimized day control cost upper limit, which is usually 1 / 24 of the day control cost budget.

[0077] In this embodiment, for different risk levels, the corresponding of different resource adjustment amounts is calculated by enumeration method (with only ±5% adjustment range, small enumeration amount and high real-time performance) The largest scheme is selected:

[0078] Among them, the high-risk level scenario only enumerates the adjustment amount of "increasing reserve" ; The low-risk level scenario only enumerates the adjustment amount of "reducing reserve" ; , and the adjustment amount of "small increase / maintenance" is enumerated in the medium-risk level scenario ; , to avoid excessive adjustment. Thus, different resource adjustment amounts are enumerated under different risk level scenarios, the corresponding reward values are calculated by the reward function, the resource adjustment amount corresponding to the maximum reward value is selected, and the intra-day resource adjustment scheme is obtained.

[0079] In this embodiment, after the intra-day resource optimization adjustment, the following constraints need to be met:

[0080] Energy storage reserve capacity constraint: ( is the total capacity of energy storage );

[0081] Cuttable load reserve constraint: ( is the total amount of cuttable load );

[0082] Energy storage SOC constraint: the adjusted should correspond to SOC in 20%-80%, for example, if MWh, SOC=30% corresponds to MWh, and the adjusted cannot exceed 15 MWh.

[0083] In the embodiment, the corresponding control strategy is determined based on the intraday resource adjustment scheme and the new energy fluctuation scenario at the target moment, including: if the new energy fluctuation scenario at the target moment is a small fluctuation scenario, the virtual inertia of the wind power / photovoltaic converter is calculated, and the wind power / photovoltaic converter is controlled based on the virtual inertia; if the new energy fluctuation scenario at the target moment is a medium fluctuation scenario, the energy storage discharges at a first rate based on the energy storage reserve capacity and the load shedding is performed based on the load shedding reserve; if the new energy fluctuation scenario at the target moment is a high fluctuation scenario, the energy storage discharges at a second rate based on the energy storage reserve capacity and the load shedding is performed based on the load shedding reserve.

[0084] Specifically, the control model can be cooperated to take corresponding control measures according to the intraday resource adjustment scheme when the actual output of new energy fluctuates, and transient instability is suppressed. The fluctuation of the actual output of new energy can be monitored in real time, and the fluctuation scenario is classified. Real-time data is collected by a phasor measurement unit (PMU) and the like, the sampling frequency is 100 Hz, and the real output of new energy and the like are obtained. The formula for calculating the real output fluctuation of new energy within 100 ms is , wherein is the real output fluctuation of new energy (unit: MW, negative value represents output drop), is the real output value of new energy at t moment (unit: MW), is the real output value of new energy at t-1 moment (unit: MW). The fluctuation scenario is classified based on the above real output fluctuation of new energy.

[0085] For a small fluctuation scenario, when the real output fluctuation of new energy is monitored MW, and the system is in a low / medium risk level, the fluctuation can be suppressed by the new energy itself adjustment, and the core adopts a virtual inertia control strategy: the system frequency deviation ( Hz, f is the real-time frequency of the system), the virtual inertia ( is the basic inertia, for example, it can be 2s MW, is the frequency deviation coefficient) is calculated according to the formula , and the virtual inertia control instruction is issued to the wind power converter or the photovoltaic converter, the system inertia is improved by the additional virtual inertia of the converter, and the frequency fluctuation is quickly suppressed.

[0086] For a medium disturbance scenario, when the real output fluctuation of new energy is When the system has a capacity of MW and is classified as medium / high risk, energy storage can be used as the core approach: energy storage backup capacity can be obtained. Can be switched off for standby and maximum charge / discharge power Calculate the available power of energy storage (10 represents the maximum duration of emergency discharge, for example, it could be 10 seconds).

[0087] according to Determine the energy storage discharge power and prioritize allocating all energy storage power to the node with the lowest voltage as calculated by power flow calculation; if the energy storage reserve capacity cannot fully fill the power gap ( If the load is cut off, no more than 30% of the available load will be reserved (prioritizing the cut-off of low-importance industrial interruptible loads), while virtual inertia control will be retained as an auxiliary measure. Control commands will be issued in the order of "energy storage → load shedding" with an interval of 50ms to ensure that the execution is completed within 300ms, balancing the stabilization effect and resource cost.

[0088] Specifically, for scenarios with large fluctuations, when the actual output of new energy fluctuates... When the system is at high risk (MW), multi-resource joint control is adopted to quickly curb the instability trend: the first step is to trigger emergency discharge of energy storage, according to the discharge power of the energy storage system. The energy storage discharge rate is set at the maximum rate (1.2 is the overload factor), the duration of which does not exceed 10 seconds, and the total discharge amount does not exceed the energy storage reserve capacity. The third step is if a power deficit still exists after the energy storage discharges ( Then calculate the amount of load that needs to be removed. And strictly control the standby load that can be switched off. Within the specified range, priority is given to cutting off marked low-priority loads; instructions are issued in the order of "energy storage → load shedding" with an interval of 50ms, and the voltage and frequency recovery status is monitored in real time to ensure that the voltage is maintained above 0.9pu and the frequency deviation is controlled within ±0.15Hz within 1s, with a response delay of no more than 300ms, so as to minimize the risk of transient instability.

[0089] In this embodiment, control commands of the control strategy can be sent to execution nodes such as energy storage and load through the scheduling data network, requiring a response delay of ≤300ms, and real-time monitoring of the command execution status of each execution node, such as the actual discharge power of energy storage and the actual load shedding amount, as well as collecting feedback information from the execution nodes, including whether the command execution was successful and the actual execution parameters.

[0090] The technical scheme provided in the embodiments of the present application can improve the prediction accuracy, effectively solve the problem of insufficient sensitivity to new energy fluctuation, perform transient risk assessment based on the corrected new energy prediction value and the new energy output fluctuation feature of the target moment, obtain a risk assessment result, perform intraday resource adjustment based on the risk assessment result, obtain an intraday resource adjustment scheme, determine the corresponding control measures based on the intraday resource adjustment scheme and the new energy fluctuation scenario of the target moment, realize efficient resource utilization, and effectively solve the problem of transient instability, and realize dual optimization of transient stability control and resource utilization.

[0091] In the embodiments of the present application, the intraday resource adjustment based on the risk assessment result to obtain an intraday resource adjustment scheme includes: adjusting the intraday resource based on the risk assessment result through an optimization model to obtain an intraday resource adjustment scheme.

[0092] The corresponding control strategy is determined based on the intraday resource adjustment scheme and the new energy fluctuation scenario of the target moment, including: the corresponding control strategy is determined based on the intraday resource adjustment scheme and the new energy fluctuation scenario of the target moment through a cooperative control model.

[0093] The method further includes:

[0094] The control effect evaluation index is obtained, and the parameters of the new energy output prediction model, the parameters of the optimization model and the parameters of the cooperative control model are adjusted based on the control effect evaluation index; wherein the control effect evaluation index includes a stability evaluation index and an economic evaluation index.

[0095] Specifically, the control effect can be evaluated by the post-evaluation and iteration module, and the model parameters and the strategy can be optimized according to the evaluation result. The control effect can be evaluated from two dimensions of stability and economy.

[0096] The stability index evaluation includes a frequency index and a voltage index. Specifically, the frequency index: the maximum frequency deviation is calculated, and the requirement is ≤±0.15Hz, wherein is the maximum frequency deviation (unit: Hz) in the control process. The voltage index: the minimum voltage is counted, and the requirement is ≥0.9p.u., wherein is the actual minimum voltage (unit: p.u.) in the control process.

[0097] wherein the economic index evaluation: calculate the unit stability cost, the formula is wherein is the unit stability cost (unit: yuan / (MW s) ), is the control cost of the intra-day resource adjustment scheme (unit: yuan), is the stability margin after control (including voltage stability margin and power angle stability margin) (unit: %), is the stability margin before control (unit: %), and the requirement is yuan / (MW s). Through the evaluation results of the stability evaluation index and the economic evaluation index, a control effect evaluation report can be generated.

[0098] In this embodiment, the related model parameters can be adjusted and optimized according to the evaluation results. Specifically, for new energy output prediction model parameter adjustment, if the error between the new energy output prediction value and the actual new energy output value is greater than 10%, the related data at the target time is added to the training set, and the attention weight of the LSTM module is adjusted and improved to increase the weight of recent data (for example, the recent data is increased by 0.1).

[0099] On the basis of the above embodiment, based on the control effect evaluation results and the model parameter adjustment, strategy optimization suggestions can be formed, and the related information can be fed back to the associated modules to promote the continuous optimization of the overall strategy.

[0100] Specifically, the resource configuration related suggestions can be generated in combination with the correlation analysis of “fluctuation-resource-cost”. For example, when the unit stability cost supported by the energy storage is lower than the load shedding in the fluctuation scenario of more than 10 MW, the suggestion of “preferentially calling energy storage resources to reduce the amount of load shedding” is proposed; if the utilization rate of a certain type of resource is less than 30% for a long time, it is suggested to adjust the standby ratio of this type of resource.

[0101] Specifically, control strategy adjustment suggestions are proposed for the control effect of different fluctuation scenarios. For example, in the large fluctuation (large impact) scenario, if the voltage recovery speed does not meet the expectation, it is suggested to optimize the response speed parameter of the unit reactive power regulation; if the virtual inertia control has poor suppression effect on frequency in the small fluctuation scenario, it is suggested to adjust the basic inertia or coefficient of the virtual inertia calculation.

[0102] Specifically, related optimization suggestions can be proposed according to the prediction error and the evaluation accuracy. If the prediction error significantly increases under a specific weather condition (such as heavy rain), it is suggested to increase the feature weight of this type of weather condition in the new energy output prediction model; if the risk level judgment deviates greatly from the actual control demand, it is suggested to recalibrate the threshold value of the risk level division.

[0103] On the basis of the above-mentioned embodiments, the iteration information can also be fed back.

[0104] Specifically, model parameter feedback: the adjusted attention weight of the LSTM module, the cost weight, etc. can be fed back to the new energy output prediction model, the optimization model, and the collaborative control model respectively, to ensure that each part uses the updated parameters for subsequent operation.

[0105] Strategy suggestion feedback: the generated strategy optimization suggestions are sorted and fed back to the control strategy formulation link of the relevant part. For example, the resource configuration optimization suggestions are fed back to the optimization model as a reference for it to formulate a resource configuration scheme; the control strategy optimization suggestions are fed back to the collaborative control model to guide it to adjust the scene-adaptive control strategy.

[0106] Data support feedback: the key data generated in this evaluation and iteration process, such as the stability index comparison before and after optimization, the expected effect analysis of the strategy suggestions, etc. are stored in the database and shared with each module or model, providing data support for subsequent optimization.

[0107] Exemplarily, a certain power grid contains wind power (installed capacity 200 MW) and photovoltaic power (installed capacity 150 MW), with a new energy penetration rate of 35%. In the afternoon (14:00-16:00) of summer, affected by strong convective weather, the wind power output may experience a second-level sudden drop (maximum drop 30 MW), and the photovoltaic power may experience a 10-15 MW fluctuation due to cloud cover, which needs to be realized by the control method provided by the embodiments of the present application to achieve transient stability and new energy consumption coordination optimization.

[0108] Implementation steps (according to the closed-loop control process)

[0109] I. Data acquisition module: fusion of real-time data and historical data

[0110] Data source acquisition

[0111] New energy side: wind power grid connection point power sensor (real output 120 MW), photovoltaic grid connection point power sensor (real output 80 MW), anemometer (10 m / s, expected to drop to 5 m / s after 30 minutes), light sensor (800 W / ㎡, expected to drop to 400 W / ㎡ after 10 minutes);

[0112] Grid side: bus node voltage (1.02 p.u.), line current (450 A), system frequency (50.02 Hz);

[0113] Environment side: weather station data (temperature 32℃, cloud cover 30%, expected to rise to 80% after 15 minutes);

[0114] History side: Call the maximum wind power drop (25MW) and the photovoltaic fluctuation record (±12MW) in the same period (summer afternoon) in the past year.

[0115] Data synchronization storage

[0116] Synchronize data to real-time database through power dispatching network, timestamp accurate to second level (such as 14:00:00, 14:00:01), provide high timeliness input for subsequent prediction.

[0117] Second, new energy fluctuation prediction model:

[0118] Data preprocessing module for data preprocessing:

[0119] Screen core features: 15-minute granularity output data in the past 7 days (historical fluctuation characteristics), real-time wind speed / illumination (meteorological characteristics), 14 o'clock (hour characteristics, encoded into continuous vector by sine and cosine);

[0120] Standardization: compress the illumination intensity (400-1000W / ㎡) to the [0,1] interval through the normalization formula to eliminate the dimension effect.

[0121] Improved LSTM module prediction:

[0122] Embedded attention mechanism: give higher attention weight (α=0.6) to the data in the past 30 minutes, and strengthen the capture of short-term fluctuations;

[0123] Prediction results: output 14:00-16:00 output trajectory, predict that wind power will drop by 28MW (120MW→92MW) at 14:15, and photovoltaic will drop to 65MW due to cloud cover at 14:30.

[0124] Intraday rolling module correction:

[0125] At 14:05, the deviation between the actual output value of new energy and the predicted value of new energy output reached 6% (triggering the correction threshold), and the subsequent prediction was corrected through a weighted fusion algorithm (70% prediction value + 30% actual value): the maximum wind power drop was corrected to 30MW, and the minimum photovoltaic output was corrected to 62MW.

[0126] Third, transient risk assessment module: quantitative risk level

[0127] Data preprocessing and verification

[0128] Receive the corrected new energy output prediction value, fill in the missing data from 14:20 to 14:22 through linear interpolation, and verify the maximum drop (30MW≤200MW×50%, reasonable and effective).

[0129] Stability margin calculation

[0130] Power angle stability margin: calculated by extended equal area criterion, accelerating area =80MW s, decelerating area =100MW s, =1-80 / 100=20%;

[0131] Voltage stability margin: minimum bus voltage calculated by power flow =0.88 p.u., =1-0.88 / 1.0=12%.

[0132] Risk level determination

[0133] Due to voltage stability margin =12% < 15%, determined as high risk, need to prioritize the allocation of stability resources.

[0134] Four, stability resource optimization module: dynamic allocation of resources

[0135] Adjustable resource parameters

[0136] Energy storage: total capacity 50MWh, maximum charge and discharge power 20MW, standby cost 0.5 yuan / kWh;

[0137] Controllable load: total amount of 30MW, compensation cost 200 yuan / MW h;

[0138] Synchronous unit: maximum reactive reserve 50000kvar, parameter adjustment cost 0.1 yuan / kvar.

[0139] Intraday resource adjustment (based on real-time risk)

[0140] State s: =20%, =12%, maximum drop amount 30MW, current energy storage reserve 10MWh;

[0141] Execute action a: according to the reward function (0.6 x stability margin-0.4 x cost ratio), increase the energy storage reserve capacity by 5% (to 10.5MWh), and the controllable load reserve by 5% (to 1.5MW).

[0142] Five, collaborative control model: scenario adaptive control

[0143] Fluctuation scenario identification

[0144] 14:15 real-time monitoring of wind power output sudden drop 30MW (|ΔP|=30MW>20MW), determined as a large impact scenario.

[0145] Control measure implementation

[0146] Emergency discharge of energy storage: discharge at a maximum rate of 1.2 x 20 MW = 24 MW for 10 s;

[0147] Contingency control: 30 MW of power gap needs to be compensated, 24 MW of energy storage is deducted, and 6 MW of load shedding is implemented. = 30 - 24 = 6 MW).

[0148] Instruction execution and feedback

[0149] Control instructions are issued through the dispatch data network, with a response delay of 250 ms (≤ 300 ms). The feedback information shows that the actual discharge of energy storage is 23.8 MW, the load shedding is 5.9 MW, and the execution is successful.

[0150] Six, post-evaluation and iteration module: optimization strategy

[0151] Control effect evaluation

[0152] Stability evaluation index: maximum frequency deviation 0.1 Hz (≤ ± 0.15 Hz), minimum voltage 0.92 p.u. (≥ 0.9 p.u.);

[0153] Economic evaluation index: control cost of intra-day resource adjustment scheme 25,000 yuan, stability margin improved by 8% (from 12% to 20%), unit stability cost = 25,000 yuan / (8% x corresponding energy) = 3.1 yuan / (MW s) (≤ 8 yuan / (MW s)).

[0154] Model parameter iteration

[0155] Due to the prediction error (30 MW actual drop vs. 28 MW initial prediction) ≤ 10%, there is no need to adjust the LSTM attention weight; the control effect meets the standard, and the current parameters are maintained.

[0156] Strategy optimization suggestion

[0157] In the large impact scenario, the cost of energy storage support is lower than that of load shedding, and it is recommended to prioritize the use of energy storage in subsequent similar scenarios (which can reduce the amount of load shedding by 10%).

[0158] Thus, the embodiment of the present application constructs a closed-loop control system of "data collection-prediction-evaluation-optimization-control-iteration", and through the coordinated linkage of each part, the whole process of accurate response to the transient instability problem caused by new energy output fluctuation is realized, and the limitation of fragmentation of each link in traditional control is broken. Through the new energy output prediction mechanism based on the improved LSTM module, the attention mechanism and the intra-day rolling correction strategy are fused to improve the prediction ability of the output fluctuation trend and characteristics, and the problem of insufficient sensitivity of traditional prediction model to random fluctuation is solved. In the embodiment of the present application, a multi-dimensional transient risk evaluation and resource optimization coordination mechanism is established, the risk level is divided in combination with the power angle stability margin and the voltage stability margin, the dynamic configuration of stable resources is realized, and the waste of resources caused by indiscriminate control is avoided. Through the real-time coordinated control and closed-loop iteration strategy of scene adaptation, corresponding control measures are taken according to the new energy fluctuation scene, and the model parameters and control strategy are continuously optimized through post-evaluation, thereby enhancing the adaptability and economy of the control scheme.

[0159] Figure 2 is a multi-time scale transient stability control device structure block diagram provided by the embodiment of the present application for new energy output fluctuation, as shown in Figure 2 The device comprises:

[0160] The data preprocessing module 210 is configured to determine a new energy historical output sequence, weather features, time features, and historical fluctuation features of the new energy output, and construct a feature matrix based on the new energy historical output sequence, the weather features, the time features, and the historical fluctuation features.

[0161] The output prediction module 220 is configured to determine a new energy output prediction value at each time in a preset time period based on the feature matrix.

[0162] The correction module 230 is configured to correct the new energy output prediction value after the target time based on the actual new energy output value at the target time to obtain a corrected new energy output prediction value, if the deviation between the new energy output prediction value at the target time and the actual new energy output value at the target time does not satisfy a preset condition.

[0163] The transient risk evaluation module 240 is configured to perform system transient risk evaluation based on the corrected new energy output prediction value and the fluctuation features of the new energy output at the target time to obtain a risk evaluation result.

[0164] The adjustment control module 250 is configured to perform intra-day resource adjustment based on the risk evaluation result to obtain an intra-day resource adjustment scheme, and determine a corresponding control measure based on the intra-day resource adjustment scheme and the new energy fluctuation scene at the target time.

[0165] In an optional embodiment, the determining, based on the feature matrix, of the new energy output prediction value at each time point in the preset time period comprises: The feature matrix, the new energy output prediction value at each time point in the preset time period, and the corrected new energy output prediction value are determined by a new energy output prediction model.

[0166] The device model iterative optimization module is configured to:

[0167] The new energy output prediction model is iteratively trained using the training set data, and the parameters of the new energy output prediction model are adjusted based on a loss function.

[0168] The new energy output prediction model is verified using the validation set data, and if the deviation between the loss function value corresponding to the validation set data and the loss function value corresponding to the training set data exceeds a preset deviation, the feature matrix is adjusted. The loss function is:

[0169]

[0170] wherein, is the loss function value; is the new energy output prediction value at time t; is the actual new energy output value at time t; is the new energy output prediction value at time t+1; is a fluctuation penalty coefficient.

[0171] In an optional embodiment, the correcting, based on the actual new energy output value at the target time point, of the new energy output prediction value after the target time point comprises:

[0172] The new energy output prediction value is corrected based on the following formula:

[0173]

[0174] wherein, is the corrected new energy output prediction value at time t, is the new energy output prediction value at time t; is the actual new energy output value at time t; is the actual new energy output value at time t; is a weight coefficient of the new energy output prediction value at time t; is a weight coefficient of the actual new energy output value at time t; is a weight coefficient of the actual new energy output value at time t; is a time interval.

[0175] In an optional embodiment, the system transient risk assessment based on the corrected new energy output prediction value and the fluctuation characteristics of the new energy output at the target time point comprises:

[0176] determine the power angle stability margin and the voltage stability margin based on the corrected new energy output prediction value;

[0177] determine the risk level based on the power angle stability margin and the voltage stability margin, and the fluctuation characteristics of the new energy output at the target time.

[0178] In an optional embodiment, based on the risk assessment result, an intraday resource adjustment is performed to obtain an intraday resource adjustment scheme, which includes:

[0179] The power angle stability margin, the voltage stability margin, the maximum drop amount of new energy output, the energy storage reserve capacity, and the load shedding reserve define the state of the power system, the energy storage reserve adjustment capacity and the load shedding reserve adjustment amount define the execution action of the power system, and the power angle stability margin, the voltage stability margin, and the control cost of the intraday resource adjustment scheme establish the reward function;

[0180] determine the optimal execution action of the power system based on the state of the power system at the target time and the reward function as the intraday resource adjustment scheme; wherein the energy storage reserve adjustment capacity satisfies the constraint within the first preset range; and the load shedding reserve adjustment amount satisfies the constraint within the second preset range.

[0181] The reward function is: ;

[0182] wherein, is the power angle stability margin; is the voltage stability margin; is the control cost of the intraday resource adjustment scheme; is the maximum allowed control cost, is the reward value.

[0183] In an optional embodiment, based on the intraday resource adjustment scheme and the new energy fluctuation scenario at the target time, a corresponding control measure is determined, which includes:

[0184] If the new energy fluctuation scenario at the target time is a small fluctuation scenario, the virtual inertia of the wind power converter or the photovoltaic converter is calculated, and the wind power converter or the photovoltaic converter is controlled based on the virtual inertia;

[0185] If the new energy fluctuation scenario at the target time is a medium fluctuation scenario, the energy storage is discharged at a first rate based on the energy storage reserve capacity, and the load shedding is performed based on the load shedding reserve;

[0186] If the new energy fluctuation scenario of the target moment is a high fluctuation scenario, the energy storage is discharged at a second rate based on the energy storage reserve capacity, and the load shedding is performed based on the load shedding reserve; wherein the second rate is greater than the first rate.

[0187] In an optional embodiment, the intraday resource adjustment based on the risk assessment result comprises:

[0188] The intraday resource is adjusted based on the risk assessment result through an optimization model to obtain an intraday resource adjustment scheme;

[0189] Based on the intraday resource adjustment scheme and the new energy fluctuation scenario of the target moment, a corresponding control strategy is determined, which comprises:

[0190] Based on the intraday resource adjustment scheme and the new energy fluctuation scenario of the target moment, a corresponding control strategy is determined through a collaborative control model;

[0191] The device further comprises a parameter adjustment module for:

[0192] Obtaining a control effect evaluation index, and adjusting the parameters of the new energy output prediction model, the parameters of the optimization model and the parameters of the collaborative control model based on the control effect evaluation index; wherein the control effect evaluation index comprises a stability evaluation index and an economic evaluation index.

[0193] As shown in Figure 3 The embodiments of the present application provide an electronic device, which comprises a processor 111, a communication interface 112, a memory 113 and a communication bus 114, wherein the processor 111, the communication interface 112 and the memory 113 complete mutual communication through the communication bus 114,

[0194] The memory 113 is used for storing a computer program;

[0195] In an embodiment of the present application, the processor 111 is used for executing the program stored in the memory 113, and realizes the method provided by any one of the preceding method embodiments, which comprises:

[0196] Determine the new energy historical output sequence, the meteorological characteristics, the time characteristics and the historical fluctuation characteristics of the new energy output, and construct a feature matrix based on the new energy historical output sequence, the meteorological characteristics, the time characteristics and the historical fluctuation characteristics;

[0197] Determine the new energy output prediction value of each moment in a preset period based on the feature matrix;

[0198] If the deviation between the new energy output prediction value at the target moment and the new energy actual output value at the target moment does not satisfy the preset condition, the new energy output prediction value after the target moment is corrected based on the new energy actual output value at the target moment, to obtain a corrected new energy output prediction value;

[0199] System transient risk assessment is performed based on the corrected new energy output prediction value and the fluctuation characteristics of the new energy output at the target moment, to obtain a risk assessment result;

[0200] Intraday resource adjustment is performed based on the risk assessment result, to obtain an intraday resource adjustment scheme, and corresponding control measures are determined based on the intraday resource adjustment scheme and the new energy fluctuation scenario at the target moment.

[0201] The embodiment of the application further provides a computer readable storage medium, which has a computer program stored thereon, and the computer program is executed by a processor to implement the steps of the method provided in any one of the foregoing method embodiments.

[0202] The device embodiments described above are merely illustrative, wherein the units described as separated components can or can not be physically separated, and the components displayed as units can or can not be physical units, that is, can be located in one place, or can be distributed on multiple network units. Part or all of the modules can be selected according to actual needs to achieve the purpose of the embodiment scheme.

[0203] Through the description of the foregoing embodiments, those skilled in the art can clearly understand that each embodiment can be realized by means of software plus a general hardware platform, and of course can also be realized by hardware. Based on such understanding, the foregoing technical solutions can be embodied in the form of a software product, which can be stored in a computer readable storage medium, such as a ROM / RAM, a magnetic disk, an optical disk, etc., and includes a plurality of instructions to make a computer device (which can be a personal computer, a server, or a network device, etc.) execute the methods described in each embodiment or some parts of the embodiments.

[0204] The foregoing embodiments are merely examples for clearly illustrating the application, and are not intended to limit the application. Based on the foregoing description, other different forms of changes or modifications can be made by those skilled in the art. Here, all the embodiments are not required to be exhausted, and the obvious changes or modifications derived therefrom are still within the protection scope of the application.

Claims

1. A method for multi-time scale transient stability control oriented to new energy output fluctuation, characterized in that, The method comprises the following steps: determining a new energy historical output sequence, meteorological characteristics, time characteristics, and historical fluctuation characteristics of new energy output, and constructing a feature matrix based on the new energy historical output sequence, the meteorological characteristics, the time characteristics, and the historical fluctuation characteristics; determining a new energy output prediction value at each time in a preset period based on the feature matrix; if the deviation between the new energy output prediction value at a target time and the actual new energy output value at the target time does not satisfy a preset condition, correcting the new energy output prediction value after the target time based on the actual new energy output value at the target time to obtain a corrected new energy output prediction value; performing system transient risk assessment based on the corrected new energy output prediction value and the fluctuation characteristics of the new energy output at the target time to obtain a risk assessment result; performing intraday resource adjustment based on the risk assessment result to obtain an intraday resource adjustment scheme, and determining a corresponding control measure based on the intraday resource adjustment scheme and the new energy fluctuation scenario at the target time.

2. The method of claim 1, wherein, The method comprises the following steps: determining a new energy output prediction value at each time in a preset period based on the feature matrix; The method further comprises the following steps: iteratively training the output prediction module using training set data and adjusting the parameters of the output prediction module based on a loss function; verifying the output prediction module using verification set data, and if the deviation between the loss function value corresponding to the verification set data and the loss function value corresponding to the training set data exceeds a preset deviation, adjusting the feature matrix; wherein the loss function is: ； wherein, is a loss function value; is a new energy output prediction value at time t; is a new energy true output value at time t; is a new energy output prediction value at time t+1; is a fluctuation penalty coefficient.

3. The method of claim 1, wherein, The method further comprises the following steps: correcting the new energy output prediction value based on the following formula: ； wherein, is a new energy output prediction value after correction at time t, is a new energy output prediction value at time t; is a new energy output prediction value at time t, is a new energy real output value at time t; is a weight coefficient of the new energy output prediction value at time t; is a new energy output prediction value at time t, is a weight coefficient of the new energy real output value at time t; is a time interval.

4. The method of claim 1, wherein, The method further comprises the following steps: determining the power angle stability margin and the voltage stability margin based on the corrected new energy output prediction value; determining the risk level based on the power angle stability margin, the voltage stability margin, and the fluctuation characteristics of the new energy output at the target time.

5. The method of claim 4, wherein, The method further comprises the following steps: defining the state of the power system based on the power angle stability margin, the voltage stability margin, the maximum new energy output drop, the energy storage reserve capacity, and the load shedding reserve, defining the execution action of the power system based on the energy storage reserve adjustment capacity and the load shedding reserve adjustment amount, and establishing a reward function based on the power angle stability margin, the voltage stability margin, and the control cost of the intraday resource adjustment scheme; determining the optimal execution action of the power system based on the state of the power system at the target time and the reward function as the intraday resource adjustment scheme; wherein the energy storage reserve adjustment capacity satisfies the constraint within a first preset range, and the load shedding reserve adjustment amount satisfies the constraint within a second preset range. wherein the reward function is; ; wherein, is the power angle stability margin; is the voltage stability margin; is the control cost of the intra-day resource adjustment scheme; is the maximum allowed control cost, is the reward value.

6. The method of claim 1, wherein, The control measures corresponding to the day-ahead resource adjustment scheme and the new energy fluctuation scenario at the target time are determined, including: If the new energy fluctuation scenario at the target time is a small fluctuation scenario, a virtual inertia of a wind power converter or a photovoltaic converter is calculated, and the wind power converter or the photovoltaic converter is controlled based on the virtual inertia; If the new energy fluctuation scenario at the target time is a medium fluctuation scenario, energy storage discharge is performed at a first rate based on an energy storage reserve capacity, and load shedding is performed based on a cuttable load reserve; If the new energy fluctuation scenario at the target time is a high fluctuation scenario, energy storage discharge is performed at a second rate based on an energy storage reserve capacity, and load shedding is performed based on a cuttable load reserve; wherein the second rate is greater than the first rate.

7. The method of claim 2, wherein, The day-ahead resource adjustment is performed based on the risk assessment result, and a day-ahead resource adjustment scheme is obtained, including: The day-ahead resource is adjusted by an optimization model based on the risk assessment result, and a day-ahead resource adjustment scheme is obtained; The control strategies corresponding to the day-ahead resource adjustment scheme and the new energy fluctuation scenario at the target time are determined, including: The control strategies corresponding to the day-ahead resource adjustment scheme and the new energy fluctuation scenario at the target time are determined by a coordinated control model; The method further includes: Control effect evaluation indexes are obtained, and parameters of the new energy output prediction model, parameters of the optimization model, and parameters of the coordinated control model are adjusted based on the control effect evaluation indexes; wherein the control effect evaluation indexes include stability evaluation indexes and economic evaluation indexes.

8. A multi-time scale transient stability control device for new energy output fluctuation, characterized in that, It includes: A data preprocessing module is configured to determine a new energy historical output sequence, meteorological characteristics, time characteristics, and historical fluctuation characteristics of new energy output, and construct a feature matrix based on the new energy historical output sequence, the meteorological characteristics, the time characteristics, and the historical fluctuation characteristics; An output prediction module is configured to determine new energy output prediction values at each time in a preset time period based on the feature matrix; A correction module is configured to correct new energy output prediction values after a target time based on a new energy actual output value at the target time if a deviation between the new energy output prediction value at the target time and the new energy actual output value at the target time does not satisfy a preset condition, and obtain corrected new energy output prediction values; A transient risk assessment module is configured to perform system transient risk assessment based on the corrected new energy output prediction values and fluctuation characteristics of new energy output at the target time, and obtain a risk assessment result; An adjustment control module is configured to perform day-ahead resource adjustment based on the risk assessment result, obtain a day-ahead resource adjustment scheme, and determine corresponding control measures based on the day-ahead resource adjustment scheme and the new energy fluctuation scenario at the target time.

9. An electronic device, comprising: It includes a memory and a processor, the memory stores a computer program, and the processor executes the computer program to implement the method of any one of claims 1-7.

10. A computer-readable storage medium, characterized in that, It has a computer program stored thereon, which makes the computer execute the method of any one of claims 1-7 when executed in the computer.

Images