A method and device for generating a power grid power imbalance risk defense pre-control scheme based on extreme weather

Abstract

The application relates to the technical field of power grid safe operation control, and discloses a power grid power imbalance risk defense pre-control scheme generation method and device based on extreme weather, which has the technical scheme as follows: extreme weather forecast data, meteorological real-time monitoring data and power grid real-time operation data are collected to construct meteorological and power grid coupling time sequence data sets; a forecast entropy is calculated according to multi-dimensional deviation weighting, a risk potential is constructed by fusing a power grid vulnerability index, and stepless smooth transition of a warning window period and a disaster correction window period is realized; in the warning window period, economy is given priority to, and a basic pre-control strategy is solved through double-layer master-slave game; in the disaster correction window period, safety is switched to be given priority to, a rolling period is adaptively adjusted according to a forecast entropy change rate, and a strategy is dynamically corrected; finally, power balance closed-loop verification is carried out, and a forecast entropy weight is dynamically corrected through attribution backtracking.

Description

Technical Field

[0001] This invention relates to the field of power grid safe operation control technology, and more specifically, to a method and apparatus for generating a power grid power imbalance risk defense and pre-control scheme based on extreme weather. Background Technology

[0002] In recent years, extreme weather events such as blizzards, freezing rain, and ice storms have occurred frequently, posing a serious threat to the safe and stable operation of the power system. Extreme weather often leads to a sharp drop in the output of new energy power generation, ice accumulation and galloping of transmission lines, and an increase in equipment failure rates, which in turn causes power imbalance in the power grid and may lead to large-scale power outages in severe cases.

[0003] Currently, the main technical shortcomings of power grid pre-control methods under extreme weather conditions are as follows: Most existing methods directly use single weather forecast values ​​for pre-control decisions, or simply employ a binary classification based on confidence thresholds, such as "reliable" and "unreliable." However, extreme weather forecasts exhibit significant biases in time, space, and intensity, and the impact mechanisms of these biases on the power grid differ across dimensions. Traditional methods cannot comprehensively characterize this multidimensional uncertainty, leading to either inaccurate forecasts resulting in failure or the waste of substantial backup resources.

[0004] Extreme weather is a dynamic evolutionary process, with its impact range and intensity changing continuously over time. Most existing methods employ fixed pre-control strategy generation cycles and fixed optimization target weights, which cannot adapt to the varying pace of weather changes. For example, when the weather deteriorates rapidly, high-frequency rolling corrections are required; when the weather is relatively stable, excessively high correction frequencies lead to wasted computational resources. Furthermore, the transition from economical to safety-oriented strategies is often a hard switch, which can easily cause abrupt jumps in control commands. Summary of the Invention

[0005] The purpose of this invention is to provide a method and apparatus for generating a power grid power imbalance risk prevention and control scheme based on extreme weather, which can effectively improve the power balance capability and operational safety of the power grid under extreme weather conditions.

[0006] The above-mentioned technical objective of the present invention is achieved through the following technical solution: 1. A method for generating a power grid power imbalance risk defense and control scheme based on extreme weather, comprising the following steps: S1. Collect extreme weather forecast data, real-time meteorological monitoring data, and real-time power grid operation data continuously according to the corresponding preset time intervals; and align all collected data in time series, construct a dynamic influence factor matrix through the dual mapping of distance between meteorological grids and power grid node impedance matrices, and form a coupled time series dataset of meteorology and power grid. S2. Based on the meteorological and power grid coupled time series dataset, and combined with the multi-dimensional deviation of extreme weather forecasts, the forecast entropy is obtained by weighted calculation of the normalized deviations of each dimension. By integrating the predicted entropy with the power grid vulnerability index, the risk potential of the power grid is constructed. Based on a preset disaster arrival time threshold, the risk potential energy is transformed through a continuous monotonic mapping function to obtain the operational stage judgment coefficient, thereby achieving a stepless smooth change in the operational state. The coefficient is used to classify and judge the early warning window period and different levels of pre-disaster correction window period. S3. When it is determined that the warning window period has been entered, the optimization objective is to minimize the weighted sum of scheduling operation cost and power imbalance risk, where the weight of scheduling operation cost is higher than that of power imbalance risk. The basic pre-control strategy is obtained by constructing a two-layer master-slave game optimization model and solving the Nash equilibrium. S4. When it is determined that the pre-disaster correction window period has been entered, the optimization objective will be switched to prioritizing the safe operation of the power grid. At the same time, the rolling correction cycle will be dynamically adjusted according to the forecast entropy change rate. The cycle will be automatically shortened when the change is drastic, and the baseline cycle will be maintained otherwise. The latest real-time meteorological data and real-time power grid operation data will be introduced in each rolling cycle to resolve the optimization model and dynamically correct the pre-control strategy. S5. After each strategy generation or correction, verify the hard constraints of power grid operation. If they are not met, automatically identify and mark the dominant deviation factors that cause the constraints to fail to meet the standards. Feed back the marking results to S2 to dynamically correct the weight coefficients corresponding to each deviation dimension of the forecast entropy. At the same time, backtrack and adjust the allocation of control resources until all hard constraints meet the standards.

[0007] As a preferred technical solution of the present invention, in S2: the multi-dimensional deviation includes the time deviation, spatial vector deviation, and intensity deviation of extreme weather forecasts; in addition to the basic three dimensions, it also includes extended dimensions such as load fluctuation and abnormal equipment failure rate added according to the characteristics of the power grid. The continuous monotonic mapping function adopts the Sigmoid function. When the expected arrival time of the disaster is greater than or equal to a preset threshold, it is determined to enter the early warning window period. Otherwise, the degree of matching of the pre-disaster correction window period is determined according to the operational stage judgment coefficient.

[0008] As a preferred technical solution of the present invention, in S3: the structure of the optimization model of the two-layer master-slave game is as follows: the upper layer is decided by the dispatch center on the output of the unit, the charging and discharging power of the energy storage and the load adjustment amount, and the lower layer sets up a virtual adversary. Within the power disturbance range defined by the forecast deviation, the fluctuation mode that causes the most severe power imbalance of the power grid is selected. When the upper and lower layers reach Nash equilibrium, the basic pre-control strategy is obtained; the power disturbance range adopts asymmetric multi-cell constraint, and its expansion degree is proportional to the forecast entropy.

[0009] As a preferred technical solution of the present invention, in S3: the execution actions of the basic pre-control strategy include: maintaining the baseline output of conventional units, reserving an adjustable capacity margin within a preset percentage range for energy storage equipment, locking flexible load resources in the region and keeping them on standby, controlling new energy equipment to operate in a conservative output range, and not performing large-scale load shedding operations; the weight of the operating cost item is greater than the weight of the power risk item during the early warning window period.

[0010] As a preferred technical solution of the present invention, the adjustment method of the rolling correction cycle in S4 is as follows: when the absolute value of the forecast entropy change rate exceeds the set threshold, the rolling cycle is shortened to a preset short cycle, and the reference rolling cycle is maintained for other operating conditions.

[0011] As a preferred technical solution of the present invention, the modified pre-control strategy in S4 includes: adding standby units as needed, using the dynamic charging and discharging level of energy storage devices to suppress power fluctuations, implementing graded flexible load control for high-risk areas based on the risk potential energy field, adjusting power flow distribution to avoid risks in weak branches; and automatically releasing redundant standby resources when the real-time risk is lower than the predicted value.

[0012] As a preferred technical solution of the present invention, in S5: the hard constraints of power grid operation include: the active power balance error of the whole domain and each zone is controlled within a first preset error range, and the power grid frequency fluctuation range is controlled within a second preset error range; the dominant deviation factors marked by the attribution backtracking include at least one of time deviation, spatial deviation, intensity deviation, load fluctuation, and abnormal equipment failure rate; after identifying the dominant deviation factors, the weight of the corresponding deviation dimension increases according to a preset step size, and the sum of the weights of each dimension is always 1. After the weight of any dimension increases, the weights of the other dimensions decrease proportionally, and an upper limit for the weight coefficient is set.

[0013] As a preferred technical solution of the present invention, in S1: the method for constructing the dynamic influence factor matrix is ​​as follows: the meteorological grid data and the power grid node impedance matrix are mapped in both geographical distance and electrical distance. The matrix is ​​reshaped in real time with the movement of extreme weather or updated according to a preset period. When the rate of change of forecast entropy exceeds a preset threshold, the matrix is ​​triggered to be updated immediately.

[0014] As a preferred technical solution of the present invention, in S2: the power grid vulnerability index is calculated and synthesized by weighting at least two types of sub-indicators among the critical load ratio, the weakness of the grid structure, the adequacy of reserve capacity, and the historical extreme weather failure rate. Each sub-indicator has been normalized to the [0,1] interval before weighting. In S4, the failure probability of power grid equipment is updated in real time according to the real-time meteorological intensity, equipment type, and equipment operating conditions. Different failure probability calculation rules are configured for different types of extreme weather.

[0015] A device for generating a power grid power imbalance risk prevention and control scheme based on extreme weather includes: a processor and a memory, wherein the memory stores a computer program executable by the processor, and the processor implements the above method when executing the computer program.

[0016] In summary, this invention offers the following advantages: It proposes the concept of forecast entropy, integrating information from extreme weather forecasts across three dimensions—time deviation, spatial vector deviation, and intensity deviation—and calculating a normalized index through weighted summation. This comprehensively characterizes the reliability of forecasts: lower entropy values ​​indicate more reliable forecasts, while higher entropy values ​​represent greater forecast uncertainty. Furthermore, this invention integrates forecast entropy with a power grid vulnerability index to construct risk potential energy, and converts it into operational phase decision coefficients using continuous monotonic mapping functions such as the Sigmoid function. Unlike traditional methods that use fixed confidence thresholds for binary hard switching, this invention achieves a seamless transition from the early warning window to the pre-disaster correction window. As the weather approaches, the conservatism of the pre-control strategy continuously changes with the risk potential energy, preventing abrupt jumps in control commands and making it more user-friendly for power grid equipment and dispatchers.

[0017] This invention constructs a master-slave game-theoretic two-layer optimization model: the upper layer, determined by the dispatch center based on current forecast entropy, includes decisions on generator output, energy storage charging / discharging power, and load adjustment; the lower layer, with a virtual adversary, selects the fluctuation mode that causes the most severe power imbalance in the grid within a power disturbance range defined by forecast deviation. When the upper and lower layers reach Nash equilibrium, the resulting pre-control scheme considers both the most adverse effects of extreme weather and avoids unnecessary over-reserve. The power disturbance range employs asymmetric multi-cell constraints, the expansion of which is proportional to the forecast entropy; that is, the more uncertain the forecast, the larger the safety margin; the more reliable the forecast, the closer the strategy is to the economic optimum. This design enables the invention to output reasonable pre-control schemes under different forecast accuracies.

[0018] This invention dynamically adjusts the rolling correction cycle based on the rate of change of forecast entropy: when the weather changes rapidly and the absolute value of the rate of change of forecast entropy exceeds a set threshold, the system automatically shortens the rolling cycle to a preset short cycle to achieve high-frequency tracking; when the weather is relatively stable, the system maintains a longer baseline rolling cycle to avoid unnecessary computational overhead. This adaptive mechanism achieves a dynamic balance between computational accuracy and efficiency, making it particularly suitable for scheduling centers with limited computing resources. Simultaneously, the latest real-time meteorological data and real-time power grid operation data are incorporated into each rolling cycle to ensure that the pre-control strategy is always based on the latest current conditions.

[0019] This invention performs a power balance closed-loop verification after each pre-control strategy is generated or modified. When the active power balance error or frequency fluctuation exceeds the preset range, the system automatically identifies and marks the dominant deviation factors causing the non-compliance, such as excessive time deviation, excessive spatial deviation, excessive intensity deviation, abnormal load fluctuation, or abnormal equipment failure rate. The marking results are then fed back to the forecast entropy calculation stage to dynamically adjust the weight coefficients of the corresponding deviation dimensions. For example, if a certain area repeatedly fails to achieve pre-control due to spatial deviation, the system will gradually increase the weight of spatial deviation in the forecast entropy, making subsequent pre-control strategies more sensitive to the positioning error of that area.

[0020] This invention continuously collects extreme weather forecast data, real-time meteorological monitoring data, and real-time power grid operation data at different preset time intervals, and constructs a dynamic influence factor matrix through dual mapping of geographical distance and electrical distance. This matrix is ​​reshaped in real time according to the movement of extreme weather or updated at preset cycles, and can also be triggered for immediate updates when the rate of change of forecast entropy exceeds a preset threshold. Compared with traditional methods that only focus on a single data source or fixed coupling relationship, this invention achieves deep integration of meteorological information and power grid topology, enabling the pre-control scheme to perceive the differences in the impact of extreme weather on different electrical nodes.

[0021] This invention classifies power grid zones into risk levels based on real-time risk potential energy fields during the disaster correction window: areas with risk potential energy above a first threshold are considered high-risk areas, those between the first and second thresholds are medium-risk areas, and those below the second threshold are low-risk areas. High-risk areas prioritize load regulation, while low-risk areas retain all electrical load. This tiered regulation strategy ensures power supply reliability in critical areas while fully utilizing the adjustment capabilities of flexible loads to participate in power balancing, avoiding the excessive impact of direct load shedding. Attached Figure Description

[0022] Figure 1 This is a flowchart of the method of the present invention. Detailed Implementation

[0023] It is readily understood that, based on the technical solution of this invention, various embodiments of the invention can be conceived by those skilled in the art without altering the essential spirit of the invention. Therefore, the following detailed embodiments and accompanying drawings are merely illustrative examples of the technical solution of this invention and should not be considered as the entirety of the invention or as limitations or restrictions on the technical solution of this invention. Rather, these embodiments are provided to enable those skilled in the art to gain a more thorough understanding of the invention. Preferred embodiments of the invention are described below in conjunction with the accompanying drawings, which form part of this application and, together with the embodiments of the invention, serve to illustrate the innovative concept of the invention.

[0024] like Figure 1As shown in the figure, this embodiment provides a method for generating a power grid power imbalance risk prevention and control scheme based on extreme weather, applicable to medium and high voltage distribution networks or transmission networks affected by extreme weather such as blizzards, freezing rain, etc. The following detailed explanation is provided with specific numerical values ​​and scenarios.

[0025] The system implementing this method includes: a data acquisition module, a data processing and calculation module, a pre-control scheme generation module, and a communication and command issuance module.

[0026] In this embodiment, the preset parameter configurations are shown in Table 1 below: Table 1 ; The specific implementation method of S1 is as follows: S1.1 Multi-source data acquisition.

[0027] Taking a power grid in a certain region of Heilongjiang Province as an example, the power grid includes two 220kV substations, eight 66kV substations, a total load of approximately 600MW, and grid-connected wind power capacity of 250MW and photovoltaic capacity of 80MW. The region experiences severe winters with frequent blizzards and freezing weather, posing multiple threats to power grid equipment, including icing, galloping, and flashover.

[0028] Extreme weather forecast data: Snow / freezing weather forecast data is obtained from meteorological departments and updated every 15 minutes, including snowfall, freezing rain intensity, temperature, wind speed, humidity and other factors for the next 72 hours.

[0029] Real-time meteorological monitoring data: collected from meteorological monitoring terminals distributed in various substations and online monitoring devices on transmission lines, reported every 5 minutes, including real-time temperature, wind speed, snowfall intensity, and line icing thickness.

[0030] Real-time power grid operation data: Acquired in real time from the SCADA / EMS system, including the output of each unit, the voltage of each node, the power flow of each line, the load power, the energy storage SOC, etc. The data refresh frequency is 5 seconds, and the data is aggregated every 5 minutes to obtain the average value, maximum value, and minimum value.

[0031] S1.2 Timing Alignment.

[0032] Because the three types of data have different time intervals—15 minutes, 5 minutes, and 5 seconds / 5 minutes—they need to be aligned on the timestamps. This embodiment adopts a downward alignment strategy: using 5 minutes as the basic time window, the 15-minute forecast data is linearly interpolated to each 5-minute time point, and the 5-second real-time power grid operation data is aggregated into a 5-minute average. Each aligned sample contains the forecast value, real-time value, and power grid status at the same moment.

[0033] S1.3 Construction of dynamic influence factor matrix.

[0034] A dual mapping is performed between meteorological raster data and the power grid node impedance matrix: Geographic distance mapping: Calculate the Euclidean distance from each power grid node to the center of the meteorological grid. The closer the distance, the greater the influence of the grid's meteorological elements on that node. For the Heilongjiang region, special attention needs to be paid to the vertical relationship between the east-west trending lines and the southward movement of cold air.

[0035] Electrical distance mapping: Based on the power grid topology, calculate the electrical distance between nodes, taking into account line impedance, transformer turns ratio, etc., and the propagation of meteorological effects along the electrical path.

[0036] By weighted fusion of two distances, a dynamic influence factor matrix M(t) is constructed, whose elements m ij (t) represents the influence weight of the j-th meteorological grid on the i-th power grid node at time t. This matrix is ​​reshaped in real time as the blizzard weather system moves: the weight of the affected node gradually increases when the cold front approaches; and the weight gradually decreases after the weather system passes.

[0037] S1.4 Weather and Power Grid Coupled Time Series Dataset.

[0038] The time-aligned data from the three categories are combined with the dynamic influence factor matrix to form a multidimensional dataset D(t) = {extreme weather forecast data, real-time meteorological monitoring data, real-time power grid operation data, M(t)}. This dataset serves as the input for subsequent steps.

[0039] The specific implementation method of step S2 is as follows: S2.1 Calculation of forecast entropy.

[0040] In this embodiment, the construction of the forecast entropy draws on the idea of ​​information entropy in measuring uncertainty. Since a larger deviation in each dimension indicates a lower reliability of the forecast in that dimension, this application normalizes the deviation value δ. i It is considered a measure of the uncertainty of this dimension.

[0041] The formula for calculating forecast entropy is: H=-Σ{i=1}^{n}w i ·δ i ·ln(δ i ); Where n is the total number of deviation dimensions, w i Let δ be the weight coefficient for the i-th dimension. i Let f(δ) = -δ·lnδ be the normalized bias value of the i-th dimension. The function f(δ) = -δ·lnδ has the following property on δ∈[0,1]: when δ→0... + When f(δ) → 0, when δ = 1, f(δ) = 0, and when δ = e -1The maximum value is found at approximately 0.368. This means that the uncertainty contribution tends to zero when the deviation is extremely small or extremely large, and the uncertainty contribution is greatest when the deviation is in an intermediate state, which meets the practical need for measuring the uncertainty of forecast deviations. When δ i When δ = 0, according to the limit definition... i ·lnδ i =0.

[0042] Assuming the current time t=0, the blizzard weather system is expected to affect the power grid in the region in the next 5 hours. Forecast data shows that the blizzard is expected to start at t=5.5h (time deviation Δt=+0.5h), the center of heavy snowfall is expected to be in County B (spatial vector deviation Δs=15km), and the maximum snowfall is expected to be 15mm (intensity deviation Δp=3mm, compared with the climate average).

[0043] This embodiment uses the following method to calculate the normalized measure of the deviation in each dimension: Time deviation measure δ t =min(|Δt| / T max ,1), where T max =2 hours is the preset maximum tolerance time deviation.

[0044] Spatial vector deviation δ s =min(Δs / S max ,1), where S max =30km is the preset maximum tolerance space deviation.

[0045] Strength deviation measure δ p =min(|Δp| / P max ,1), where P max =6mm is the preset maximum tolerance intensity deviation (in the blizzard classification, 24-hour snowfall ≥10mm is a blizzard, and 6mm is the acceptable deviation range).

[0046] In this embodiment: δ t =0.5 / 2=0.25, δ s =15 / 30=0.5, δ p =3 / 6=0.5.

[0047] The initial weights are configured with equal weights: w t =w s =w p =1 / 3. Substituting the weights into the prediction entropy formula: H=-[(1 / 3)×0.25×ln(0.25)+(1 / 3)×0.5×ln(0.5)+(1 / 3)×0.5×ln(0.5)] =-(1 / 3)×[0.25×ln(0.25)+0.5×ln(0.5)+0.5×ln(0.5)]; The calculations are: -0.25×(-1.3863)=0.3466, -0.5×(-0.6931)=0.3466, -0.5×(-0.6931)=0.3466. The sum of the three terms is 1.0398, which is 0.3466 when divided by 3.

[0048] That is, H≈0.35, which is at a moderately low level, indicating that the current forecast has a certain degree of reliability but also contains biases. In this embodiment, the calculated forecast entropy is truncated so that it falls within the [0,1] interval.

[0049] S2.2 Calculation of the power grid vulnerability index.

[0050] The power grid vulnerability index V is a weighted composite of four sub-indicators. Each sub-indicator has been normalized to the [0,1] interval before weighting, as shown in Table 2 below: Table 2 ; V=0.35×0.3+0.45×0.3+0.25×0.2+0.30×0.2=0.105+0.135+0.05+0.06=0.35.

[0051] S2.3 Risk Potential Construction.

[0052] With α=0.6 and β=0.4, the values ​​of α and β reflect the dispatchers' relative emphasis on forecast uncertainty and the inherent vulnerability of the power grid. In this embodiment, these values ​​were adjusted based on expert experience. The risk potential energy Φ=0.6×0.35+0.4×0.35=0.21+0.14=0.35. Since both H and V fall within the interval [0,1] and α+β=1, Φ naturally falls within the interval [0,1].

[0053] S2.4 Sigmoid Mapping and Stage Determination.

[0054] Using Sigmoid parameters: steepness k=10, median Φ0=0.5. Calculate the decision coefficient for the operational phase: F(Φ)=1 / (1+e^{-10×(0.35-0.5)})=1 / (1+e^{1.5})=1 / (1+4.4817)≈0.182.

[0055] The decision logic is as follows: First, determine the estimated arrival time T of the disaster. arrival With preset threshold T th (This example uses 3 hours) Relationship: When T arrival ≥Tth When this happens, it is directly determined that the warning window period has begun; When T arrival <T th Then, based on the operational stage judgment coefficient F(Φ), a graded judgment is made: when F(Φ) < 0.3, the early warning window period strategy is maintained; when 0.3 ≤ F(Φ) < 0.6, the early stage of the pre-disaster correction window period is entered; when F(Φ) ≥ 0.6, the in-depth stage of the pre-disaster correction window period is entered. Different stages correspond to different strategy adjustment ranges.

[0056] The Sigmoid function achieves a continuous mapping from risk potential energy to decision coefficients, ensuring a smooth mathematical evolution of the system state. At the decision level, response levels are divided based on decision coefficients, facilitating the step-by-step execution of actual control commands.

[0057] In this embodiment, let T be the estimated time T for the disaster to arrive. arrival =5 hours, preset threshold T th =3 hours. Due to T arrival =5h≥3h, indicating the start of the warning window period.

[0058] The specific implementation method of step S3 is as follows: S3.1 Construction of a two-layer master-slave game optimization model.

[0059] Upper-level model (scheduling center): Decision variable: Output P of conventional units g (Unit: MW), Energy storage charging and discharging power P b (Positive for discharging, negative for charging), flexible load adjustment ΔL f (A positive value indicates a reduction).

[0060] Objective function: min[0.7×C] cost +0.3×R risk ]; in: C cost =Σ(c g ×P g )+c b ×|P b |+c f ×ΔL f c g =0.45 yuan / kWh is the power generation cost of thermal power units (most of the power plants in Heilongjiang are thermal power plants, and combined heat and power (CHP) constraints also apply during the heating season), c b =0.35 yuan / kWh is the energy storage depreciation cost, c f =0.7 yuan / kWh is the load regulation compensation cost.

[0061] R risk =Σ(λi ×ΔP i ²), ΔP i λ represents the power deficit of the i-th partition. i The risk weights are assigned to different zones (high λ for critical load zones including hospitals, government facilities, and heating facilities, and low λ for general load zones). Heilongjiang's winter heating load falls under the category of critical loads related to people's livelihoods, hence its relatively high λ value.

[0062] Lower-level model (virtual adversary): Decision variable: Power perturbation vector d=[d wind ,d pv ,d load The values ​​represent wind power output deviation (heavy wind power generation in Heilongjiang during winter, but may drop sharply due to icing), photovoltaic power output deviation (limited output due to short daylight hours in winter), and load fluctuation deviation (surge in residential electricity consumption during blizzards).

[0063] Constraints: d∈D(H), where D(H) is the asymmetric polytope perturbation interval. The polytope is a convex polyhedron bounded by a finite number of linear inequalities, specifically in the form of: Let the nominal disturbance interval be D0 = [-d̄, d̄], then D(H) = [-(1+γ] H) d̄,(1+γ H) d̄], where γ is a preset expansion coefficient (γ=0.5 in this embodiment), that is, the expansion degree of the perturbation interval is proportional to the forecast entropy H - the more uncertain the forecast (the larger H is), the larger the perturbation interval. When H=0.35, the radius of the perturbation interval is approximately 1.175 times the nominal value.

[0064] Lower-level objective: Maximize the risk term R in the upper-level objective function. risk (Even when the power grid imbalance is most severe).

[0065] S3.2 Solving for Nash equilibrium.

[0066] The iterative dual ascent method is used to solve the problem: after the upper layer gives the decision variables, the lower layer selects the worst-case perturbation; the upper layer adjusts the decision based on the feedback from the lower layer; this process is repeated iteratively until convergence. In this embodiment, after 6 iterations, the change in the objective function is less than the preset convergence threshold, indicating that Nash equilibrium has been reached.

[0067] S3.3 Basic Pre-control Strategy Output.

[0068] The basic pre-control strategy for the early warning window period obtained from the solution is as follows: Table 3 ; The specific implementation method of step S4 is as follows: S4.1 Time progression and window period transition.

[0069] Two hours later (t=2h), the blizzard system continued to approach. New forecast data showed that the time deviation decreased to 0.2h, the spatial vector deviation decreased to 8km, and the intensity deviation increased to 4.5mm. The forecast entropy was recalculated: δ t =0.1, δ s ≈0.267, δ p =0.75, new H≈0.48, risk potential energy Φ≈0.44, F(Φ)≈0.35.

[0070] The estimated arrival time of the disaster is T. arrival =3h, still ≥3h, continues to be in the warning window period.

[0071] After another hour (t=3h), T arrival =2h < 3h, new forecast data shows that the blizzard path is basically determined (spatial vector deviation 3km, time deviation 0.1h), but the snowfall intensity may be stronger than expected (intensity deviation 5mm, belonging to blizzard to heavy blizzard level). Calculated: δ s ≈0.1, δ t =0.05, δ p ≈0.833, new H≈0.38 (due to the significant reduction in spatiotemporal deviation and the increase in intensity deviation, the overall entropy value has increased slightly), risk potential energy Φ≈0.46, F(Φ)≈0.42.

[0072] Because of T arrival =2h<3h, enter the pre-disaster correction window period determination process. F(Φ)=0.42 is in the 0.3-0.6 range, and is determined to be in the initial stage of the pre-disaster correction window period.

[0073] S4.2 Adaptive rolling correction cycle adjustment.

[0074] The forecast entropy change rate was monitored: from t=2h to t=3h, H changed from 0.48 to 0.38, a change of -0.10, with a time interval of 1 hour, and an average change rate of approximately -0.00167 / minute. The absolute value is much smaller than the set threshold of 0.1 / minute. Therefore, the rolling cycle remains at a 5-minute baseline cycle.

[0075] Another 0.5 hours passed (t=3.5h), and the leading edge of the cold front began to affect the power grid area. The temperature plummeted to -28℃, the wind speed increased to 12m / s, and the snowfall intensity rapidly increased from the predicted 15mm to the measured 22mm. The forecast data was updated rapidly: the intensity deviation jumped from 5mm to 9mm, and the forecast entropy surged from 0.38 to 0.72, with a change rate as high as (0.72-0.38) / 0.5=0.68 / minute. The absolute value exceeded the set threshold of 0.1 / minute, and the system automatically switched the rolling cycle from 5 minutes to 1-2 minutes (1 minute in this embodiment). In this embodiment, the rolling cycle adjustment is mainly based on the forecast entropy change rate. In practical applications, it can be further combined with real-time meteorological intensity, power grid load rate, and other factors for joint decision-making.

[0076] S4.3 Rolling Correction Strategy Generation.

[0077] During each rolling cycle, perform the following operations: Data refresh: Read the latest 5-minute real-time meteorological data (actual temperature -28℃, wind speed 12m / s, snowfall intensity 22mm, ice thickness on some lines has reached 8mm) and the latest 1-minute real-time power grid operation data (the load rate of a certain 66kV line has risen to 78%, energy storage SOC=0.58, and 2 lines have triggered icing alarms).

[0078] Resolve the optimization model: Substitute the latest data into the S3 two-layer master-slave game optimization model. At this time, the objective function weights have been switched to power imbalance risk term weight 0.8 and scheduling operation cost term weight 0.2.

[0079] In this embodiment, after entering the pre-disaster correction window, all flexible load resources locked during the early warning window are released and re-participate in this round of optimization.

[0080] The revised strategy is shown in Table 4 below: Table 4 ; Example of graded flexible load control in S4.4.

[0081] This embodiment divides the power grid into existing electrical zones, and the risk potential energy of each zone is taken as the weighted average of the risk potential energy of all nodes within that zone. Based on the real-time risk potential energy field calculation results: Area A (including hospitals, government offices, and heating stations): Risk potential energy 0.22 → Low-risk area → No load regulation will be implemented to ensure heating and power supply for residents; Area B (Business Center): Risk potential 0.50 → Medium risk area → Level II control, reduce non-essential loads by 8%; Region C (Industrial Park): Risk potential 0.75 → High-risk area → Level 1 control, reduce non-essential loads by 30%; S4.5 device failure probability is updated in real time.

[0082] Based on real-time weather intensity (temperature -28℃, wind speed 12m / s, line icing thickness 8mm), the preset "Blizzard / Freezing - Line Fault Probability Table" was consulted: for icing thickness between 5-10mm, the fault probability of a typical overhead line in this area is 0.15 (per hour). Considering the line's operating years (15 years, moderate aging), current load rate (78%), and historical galloping records, the fault probability was adjusted to 0.22. This fault probability is converted into N-1 failure scenarios for the corresponding line, which are input into the optimization model as one of the optional disturbance modes for the lower-level virtual adversary, influencing the lower-level virtual adversary's disturbance selection (preferring to choose the scenario where the line trips due to icing galloping).

[0083] The specific implementation method of step S5 is as follows: S5.1 Power balance verification.

[0084] After each policy correction, the system automatically performs a power balance check.

[0085] Example scenario: After a certain correction, the global active power balance calculation is as follows: Total power generation: 400MW from conventional units (including an additional 50MW heating unit) + 175MW actual wind power output (79.5% of the predicted 220MW) + 25MW photovoltaic power (lower output in winter) + 15MW energy storage discharge = 615MW; Total load: Rigid load 550MW (including residential electric heating load, which increased during blizzard weather) + interruptible load actually cut off 10MW (12MW mandated, execution rate 83%) = 560MW; The calculation yields: 615MW - 560MW = 55MW (surplus); The surplus to load ratio is approximately 9.82% (55 / 560), which far exceeds the first preset error range of 1%, thus failing the verification. This embodiment does not distinguish between power deficit and power surplus scenarios, and verifies the error uniformly based on absolute value. In practical applications, when there is a power surplus, the output of conventional units is preferentially reduced or energy storage charging is increased; when there is a power deficit, standby units are preferentially added or load regulation is initiated.

[0086] S5.2 Attribution Retrospection.

[0087] The system initiates the attribution backtracking module to check for possible deviation factors item by item, as shown in Table 5 below: Table 5 ; The dominant deviation factors were identified as: intensity deviation and load fluctuation.

[0088] S5.3 Dynamic weight correction.

[0089] In this embodiment, the forecast entropy of S2 is based on a three-dimensional bias structure, namely, time bias, spatial vector bias, and intensity bias. Load fluctuations and equipment failure rate anomalies identified in the attribution backtracking are not inherent bias dimensions of extreme weather forecasts, but rather power grid-side response factors. To maintain conceptual clarity, this embodiment does not directly incorporate load fluctuations into the dimensional expansion of the forecast entropy. Instead, a separate load fluctuation risk term is added to the objective function of the optimization model and controlled through an independent penalty coefficient. If it is necessary to incorporate the forecast entropy system in subsequent practical applications, the dimensions can be pre-expanded in S2, but this embodiment maintains the clarity of the three-dimensional structure of the forecast entropy.

[0090] The sum of the weights for each dimension is always 1. As the weight of any dimension increases, the weights of the other dimensions decrease proportionally. The intensity deviation weight w of the predicted entropy in step S2... p The preset step size is 0.05, and the maximum is 0.55. After this backtracking, w p As it increases to 0.383, w decreases accordingly. t and w s (Keep the sum to 1). The corrected weights are: w t =0.308, w s =0.309, w p =0.383.

[0091] The effect of this correction is that the sensitivity to intensity deviation is increased in subsequent forecast entropy calculations, prompting the pre-control strategy to reserve more safety margin to cope with the uncertainty of blizzard intensity.

[0092] S5.4 Retrospectively adjust and regulate resource allocation.

[0093] In cases dominated by intensity deviation, the system increases the energy storage reserve capacity margin from 28% to 35% and decreases the wind power conservative output factor from 0.75 to 0.70 (to cope with the risk of wind turbine icing shutdown caused by stronger snowfall).

[0094] After adjustment and re-verification: The energy storage discharge strategy was adjusted (reduced from 15MW to 8MW), and wind power was arranged with more conservative values ​​(reduced from 175MW to 165MW). The actual surplus decreased from 55MW to 38MW, and the error ratio decreased from 9.82% to 6.8%. This still did not meet the target, so iterative adjustments continued until the surplus ratio was ≤1%. In this embodiment, the maximum number of attribution backtracking iterations was set to 10. If the target was still not met after reaching the maximum number of iterations, the scheme with the smallest constraint deviation was selected as the final strategy.

[0095] S5.5 offline self-evolution.

[0096] After each extreme weather event, the system can collect the following data: the forecast entropy sequence for each rolling cycle, the actual power deficit for each cycle, the reserve capacity reserved by the pre-control strategy for each cycle, and the actual equipment failures and line icing conditions. By comparing and analyzing the relationship between actual errors and forecast entropy, and using historical error samples within the rolling time window, the morphological parameters of the asymmetric multicell disturbance interval are updated with the goal of minimizing the mean square error, making the characterization of the uncertainty set increasingly closer to the actual characteristics of the power grid under winter blizzard weather.

[0097] The above description is merely a preferred embodiment of the present invention, but the scope of protection of the present invention is not limited thereto. Any variations or substitutions that can be easily conceived by those skilled in the art within the scope of the technology disclosed in this invention should be included within the protection scope of this invention.

[0098] It should be understood that, in order to simplify the present invention and help those skilled in the art understand its various aspects, in the above description of exemplary embodiments of the present invention, various features of the present invention are sometimes described in a single embodiment or with reference to a single figure. However, the present invention should not be construed as implying that all features included in the exemplary embodiments are essential technical features of the claims of the present invention.

[0099] Those skilled in the art will understand that all or part of the processes of the methods described in the above embodiments can be implemented by a computer program instructing related hardware, and the program can be stored in a computer-readable storage medium. The computer-readable storage medium may be a disk, optical disk, read-only memory, or random access memory, etc.

[0100] It should be understood that the modules, units, components, etc., included in the device of one embodiment of the present invention can be adaptively changed to be placed in a device different from that embodiment. Different modules, units, or components included in the device of the embodiment can be combined into a single module, unit, or component, or they can be divided into multiple sub-modules, sub-units, or sub-components.

[0101] The modules, units, or components in the embodiments of the present invention can be implemented in hardware, in software running on one or more processors, or in a combination thereof. Those skilled in the art should understand that... In practice, microprocessors or digital signal processors (DSPs) can be used to implement embodiments of the invention. The invention can also be implemented on computer program products or computer-readable media for performing some or all of the methods described herein.

Claims

1. A method for generating a power grid power imbalance risk prevention and control scheme based on extreme weather, characterized by: Includes the following steps: S1. Collect extreme weather forecast data, real-time meteorological monitoring data, and real-time power grid operation data continuously according to the corresponding preset time intervals; and align all collected data in time series, construct a dynamic influence factor matrix through the dual mapping of distance between meteorological grids and power grid node impedance matrices, and form a coupled time series dataset of meteorology and power grid. S2. Based on the meteorological and power grid coupled time series dataset, and combined with the multi-dimensional deviation of extreme weather forecasts, the forecast entropy is obtained by weighted calculation of the normalized deviations of each dimension. By integrating the predicted entropy with the power grid vulnerability index, the risk potential of the power grid is constructed. Based on a preset disaster arrival time threshold, the risk potential energy is transformed through a continuous monotonic mapping function to obtain the operational phase judgment coefficient. The continuous change characteristics of the judgment coefficient are used to identify the phase, and a graded judgment is performed according to the interval in which the coefficient is located, distinguishing between the early warning window period and different levels of pre-disaster correction window period. S3. When it is determined that the warning window period has been entered, the optimization objective is to minimize the weighted sum of scheduling operation cost and power imbalance risk, where the weight of scheduling operation cost is higher than the weight of power imbalance risk. By constructing an optimization model of a two-layer master-slave game and solving for the Nash equilibrium, the basic pre-control strategy is obtained. S4. When it is determined that the pre-disaster correction window period has been entered, the optimization objective will be switched to prioritizing the safe operation of the power grid. At the same time, the rolling correction cycle will be dynamically adjusted according to the forecast entropy change rate. The cycle will be automatically shortened when the change is drastic, and the baseline cycle will be maintained otherwise. The latest real-time meteorological data and real-time power grid operation data will be introduced in each rolling cycle to resolve the optimization model and dynamically correct the pre-control strategy. S5. After each strategy generation or correction, verify the hard constraints of power grid operation. If they are not met, automatically identify and mark the dominant deviation factors that cause the constraints to fail to meet the standards. Feed back the marking results to S2 to dynamically correct the weight coefficients corresponding to each deviation dimension of the forecast entropy. At the same time, backtrack and adjust the allocation of control resources until all hard constraints meet the standards.

2. The method for generating a power grid power imbalance risk defense and control scheme based on extreme weather as described in claim 1, characterized in that: in S2: the multi-dimensional deviation includes the time deviation, spatial vector deviation, and intensity deviation of extreme weather forecasts; in addition to the basic three dimensions, it also includes extended dimensions such as load fluctuation and abnormal equipment failure rate added according to the characteristics of the power grid; The continuous monotonic mapping function adopts the Sigmoid function. When the expected arrival time of the disaster is greater than or equal to a preset threshold, it is determined to enter the early warning window period. Otherwise, the degree of matching of the pre-disaster correction window period is determined according to the operational stage judgment coefficient.

3. A method for generating a power grid power imbalance risk defense and control scheme based on extreme weather as described in claim 1, characterized in that: in S3: the structure of the optimization model of the two-layer master-slave game is as follows: the upper layer is decided by the dispatch center on the output of the generating units, the charging and discharging power of energy storage and the load adjustment amount, and the lower layer sets up a virtual adversary. Within the power disturbance range defined by the forecast deviation, the fluctuation mode that causes the most severe power grid power imbalance is selected. When the upper and lower layers reach Nash equilibrium, the basic pre-control strategy is obtained; the power disturbance range adopts asymmetric multi-cell constraint, and its expansion degree is proportional to the forecast entropy.

4. The method for generating a power grid power imbalance risk prevention and control scheme based on extreme weather as described in claim 1, characterized in that: in S3: the execution actions of the basic pre-control strategy include: Maintain the baseline output of conventional generating units, reserve an adjustable capacity margin within a preset percentage range for energy storage equipment, lock in flexible load resources in the region and keep them on standby, control new energy equipment to operate within a conservative output range, and do not perform large-scale load shedding operations; during the early warning window period, the weight of the operating cost item is greater than the weight of the power risk item.

5. The method for generating a power grid power imbalance risk prevention and control scheme based on extreme weather as described in claim 1, characterized in that: The adjustment method for the rolling correction cycle in S4 is as follows: when the absolute value of the forecast entropy change rate exceeds the set threshold, the rolling cycle is shortened to the preset short cycle, and the baseline rolling cycle is maintained for other operating conditions.

6. The method for generating a power grid power imbalance risk prevention and control scheme based on extreme weather as described in claim 1, characterized in that: The revised pre-control strategy in S4 includes: adding standby units as needed, using energy storage devices to dynamically charge and discharge levels to suppress power fluctuations, implementing tiered flexible load control in high-risk areas based on the risk potential field, adjusting power flow distribution to avoid risks in weak branches, and automatically releasing redundant standby resources when the real-time risk is lower than the predicted value.

7. The method for generating a power grid power imbalance risk prevention and control scheme based on extreme weather as described in claim 1, characterized in that: S5 The hard constraints for power grid operation include: controlling the active power balance error of the entire region and each zone within a first preset error range, and controlling the power grid frequency fluctuation range within a second preset error range; the dominant deviation factors marked by the attribution backtracking include at least one of time deviation, spatial deviation, intensity deviation, load fluctuation, and abnormal equipment failure rate; after identifying the dominant deviation factors, the weight of the corresponding deviation dimension increases by a preset step size, and the sum of the weights of each dimension is always 1. After the weight of any dimension increases, the weights of the other dimensions decrease proportionally, and an upper limit for the weight coefficient is set.

8. A method for generating a power grid power imbalance risk prevention and control scheme based on extreme weather as described in claim 1, characterized in that: in S1: the method for constructing the dynamic influence factor matrix is: to perform a dual mapping of geographical distance and electrical distance between meteorological grid data and the power grid node impedance matrix, and the matrix is ​​reshaped in real time or updated according to a preset cycle as the extreme weather moves; when the rate of change of forecast entropy exceeds a preset threshold, the matrix is ​​triggered to be updated immediately.

9. A method for generating a power grid power imbalance risk defense and control scheme based on extreme weather as described in claim 1, characterized in that: in S2: the power grid vulnerability index is calculated and synthesized by weighting at least two types of sub-indicators among the critical load ratio, grid structure weakness, reserve capacity adequacy, and historical extreme weather failure rate, and each sub-indicator has been normalized to the [0,1] interval before weighting; in S4: the power grid equipment failure probability is updated in real time according to the real-time meteorological intensity, equipment type, and equipment operating conditions, and differentiated failure probability calculation rules are configured for different types of extreme weather.

10. A device for generating a power grid power imbalance risk prevention and control scheme based on extreme weather, characterized in that: include: A processor and a memory, the memory storing a computer program executable by the processor, wherein the processor, when executing the computer program, implements the method of any one of claims 1-9.

Images