Flexible load multi-time scale coordination control method for novel power system

Abstract

The invention discloses a flexible load multi-time scale coordination control method for a novel power system, and the method comprises the specific steps: collecting millisecond, minute and hour level flexible loads and environment data, and constructing a multi-time scale load response model; a millisecond-level model and a second-level model are utilized, a real-time operation state is combined, a second-level adjusting instruction is generated, and an execution result is fed back; based on minute-to-hour level prediction data, integrating adjustable load, distributed energy and energy storage constraints, optimizing a load operation plan in a rolling manner, and coordinating source-network load-storage combined operation; and making a long-term load operation plan in combination with day-level and above prediction data, and executing deviation self-adaptive correction according to short and medium periods. According to the method, hierarchical control and collaborative optimization of the adjustable load under different time scales from millisecond level to day level are realized, and the operation stability, the adjustment flexibility and the auxiliary service capability of the novel power system under the high-proportion new energy access condition are remarkably improved.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of power system operation and control, and particularly relates to a flexible load multi-time scale coordinated control method for a new type of power system. BACKGROUND

[0002] In recent years, with the deepening of global energy structure transformation, the installed capacity and proportion of new energy are rapidly increasing, and the penetration rate of renewable energy such as wind power and photovoltaic in the power system is continuously increasing. The new type of power system has gradually become the main direction of power grid development. The new type of power system takes high proportion of new energy as the main power supply, has the characteristics of clean and low carbon, distributed access and multi-energy complementation, but its intermittency, volatility and uncertainty also significantly increase the complexity and instability of system operation. Especially under the condition of high proportion of new energy grid connection, the stability margin of system frequency and voltage is reduced, the demand for peak regulation, frequency modulation and standby capacity is significantly increased, and higher requirements are put forward for the flexible regulation capacity of the power system.

[0003] Under this background, the demand side flexible load (also known as adjustable load or interruptible load) has gradually become an important part of the flexible resource of the power system. Flexible load refers to the load resource that can actively adjust its power consumption under the premise of not significantly affecting the production and life of users according to the operation state and dispatching instruction of the power grid, including industrial interruptible load, commercial building air conditioning load, residential electric water heater, electric vehicle charging and swapping load, etc. These loads can play a role similar to regulating power supply in power system operation through peak clipping, valley filling, emergency response, etc., and significantly improve the regulation capacity and operation economy of the system.

[0004] However, the existing flexible load control technology is mostly focused on a single time scale, such as only playing a role in second-level regulation in real-time frequency modulation, or only participating in load optimization scheduling in day-ahead and intraday markets, lacking dynamic connection and collaborative optimization between multiple time scales. This fragmented regulation mode leads to the fact that the flexibility potential of flexible load is not fully released, and the short-period fast control and long-period economic scheduling are often unrelated, even in some cases there is mutual interference, thereby affecting the coordination and stability of the overall operation of the power system.

[0005] In addition, in the new type of power system with high proportion of new energy access, the uncertainty of power grid operation increases, and the rapid fluctuation of new energy output in short period and the seasonal change in long period are superimposed, making it difficult for traditional single time scale load control strategy to balance rapidity and economy. For example, in the case of frequency drop, if there is no millisecond to second level load fast response, the system frequency may drop below the safety threshold; and in the long-term operation, if the annual prediction of load and new energy output is not combined for optimization scheduling, it will lead to low utilization rate of standby resources and rising operation cost. Summary of the Invention

[0006] The purpose of this invention is to provide a flexible load multi-timescale coordinated control method for new power systems. This method not only realizes hierarchical control and coordinated optimization of adjustable loads at different timescales, from milliseconds to days, but also significantly improves the operational stability, regulation flexibility, and ancillary service capabilities of new power systems under conditions of high proportion of renewable energy access.

[0007] The technical solution of the present invention to solve the above-mentioned technical problems is as follows:

[0008] This invention provides a flexible load multi-timescale coordinated control method for novel power systems, comprising the following steps:

[0009] S1: Collect millisecond, minute, and hourly flexible load and environmental data to construct a multi-timescale load response model, providing a parameter basis for subsequent control and rolling scheduling;

[0010] S2: Utilizing millisecond-level and second-level models, combined with real-time operating status, second-level adjustment commands are generated to achieve immediate response to frequency and voltage fluctuations and to provide feedback on the execution results;

[0011] S3: Based on minute-to-hour forecast data, comprehensively considers flexible load, distributed energy resources and energy storage constraints, continuously optimizes load operation plans, and coordinates the joint operation of power generation, grid, load and storage;

[0012] S4: Combine daily and higher-level forecast data to formulate long-term load operation plans and adaptively correct deviations based on short- and medium-term cycle execution.

[0013] Optionally, in step S1, a multi-source synchronous acquisition device is deployed to simultaneously acquire power operation parameters and environmental parameters on the flexible load side; the power operation parameters include three-phase voltage U a U b U c Three-phase current I a ,I b ,I c Active power P(t), reactive power Q(t), and power factor Environmental parameters include temperature T(t), humidity H(t), light intensity L(t), and wind speed V. w (t).

[0014] Optionally, in step S1, the multi-source synchronous acquisition device includes a high-speed power parameter acquisition unit, an environmental sensing unit, and a meteorological data interface module. The high-speed power parameter acquisition unit has a sampling frequency range of 10kHz to 50kHz and is used to capture transient voltage and current waveforms. The environmental sensing unit records external factors such as temperature, humidity, wind speed, and light intensity in real time. The meteorological data interface module interacts with the meteorological forecasting platform via API to obtain the weather change trend for the next few hours to days. After the multi-source synchronously acquired data is preprocessed by edge computing nodes, hierarchical regression is used to quickly fit the load response curve under the known pattern, and neural networks are used to learn nonlinear and time-varying characteristics. Finally, a high-precision load dynamic response model is generated on three time scales: milliseconds, minutes, and hours.

[0015] Alternatively, in step S1, a time-scale mapping function is used. Synchronize the raw data to time series at various scales:

[0016]

[0017] Where X raw X is the original sampling sequence. Δt For sequences at the target time resolution;

[0018] Based on this, a multi-timescale load response model is constructed; using a hierarchical modeling method, the millisecond, minute, and hourly features are represented as follows:

[0019] F ms =f ms (U(t),I(t),T(t),H(t)) (2)

[0020] F min =f min (P(t),Q(t),L(t),V w (t)) (3)

[0021] F h =f h (P avg (h),T avg (h),W forecast (h)) (4) Among them, f ms ,f min ,f h These are feature extraction functions for milliseconds, minutes, and hours, respectively. forecast (h) represents hourly meteorological forecast data;

[0022] To integrate characteristics from different time scales, a multi-time-scale load response prediction function is established:

[0023]

[0024] Φ(·) is a cross-scale feature fusion model that combines hierarchical regression and time attention mechanism, which can simultaneously capture the response characteristics of load to voltage disturbances and environmental changes at different time scales.

[0025] Optionally, in step S2, based on the multi-time-scale load response model constructed in step S1, a second-level fast control algorithm is introduced. The second-level fast control algorithm includes: disturbance detection, second-level predictive control, constraint solving, and command issuance and feedback.

[0026] The disturbance detection is performed by real-time monitoring of the bus voltage U(t) and the system frequency f(t) to calculate the instantaneous deviation.

[0027] ΔU(t)=U(t)-U ref Δf(t)=f(t)-f ref (6)

[0028] U ref and f ref Given the rated voltage and rated frequency, when |Δf(t)|>δ U Or |Δf(t)|>δ f When this happens, the quick adjustment mode is triggered;

[0029] The aforementioned second-level predictive control calls millisecond-level and second-level response models f. ms with f s In the prediction time domain T p Internal estimation of the power response curve of flexible load:

[0030]

[0031] Where k∈[1,T] p ], where ω is the environmental factor weight coefficient and E(t+k) is the predicted environmental state vector;

[0032] The constraint solution involves constructing a second-level optimization objective function based on the prediction results:

[0033]

[0034] Where α, β, and γ are the weights of frequency, voltage deviation, and regulation cost, respectively, and C adj The load adjustment cost function;

[0035] The constraints include:

[0036] P min ≤P load (t+k)+u(t+k)≤P max (9)

[0037] |u(t+k)-u(t+k-1)|≤R max (10)

[0038] The solution is obtained using a fast iterative method based on quadratic programming (QP), which guarantees that the optimal solution is found within 100ms.

[0039] The aforementioned instruction issuance and feedback optimizes the result u. * (t) is sent to the load controller at the execution end via a low-latency communication link as a second-level power adjustment command. The controller adjusts the operating status according to the command, and the execution end feeds back the current power change ΔP every second. exec and execution success rate η exec The data is sent to the control center for rolling scheduling correction in subsequent step S3.

[0040] Alternatively, in step S3, a rolling prediction-optimization-execution closed-loop mechanism is introduced on a time scale of minutes to hours. This mechanism includes rolling prediction, rolling optimization, and coordinated execution.

[0041] The rolling forecast utilizes the minute-level and hour-level load response models from step S1, combined with new energy output forecasts. and the energy storage state of charge (SOC(t)) to generate the future T f Hourly multi-energy power prediction sequence:

[0042]

[0043] in This indicates the predicted charging and discharging power of the energy storage system;

[0044] The aforementioned rolling optimization uses ΔT = 30 min as the default optimization period and dynamically adjusts the future scheduling scheme based on real-time operational data. The optimization objective function is:

[0045]

[0046] Among them, u load Load adjustment command vector; u ESS C is the energy storage charge / discharge control vector; load C is the load adjustment cost function; ESS This comes at the cost of energy storage lifespan loss and energy consumption.

[0047] The constraints include:

[0048] (1) Adjustable load range:

[0049]

[0050] (2) Energy storage constraints:

[0051] SOC min， ≤SOC(t+k)≤SOC max (14)

[0052] (3) Constraints on energy storage charge and discharge rates:

[0053] The solution employs mixed integer quadratic programming (MIQP), which supports hybrid optimization of discrete load switching control and continuous energy storage power control, ensuring that the solution time is less than 5 seconds within the optimization cycle.

[0054] The coordinated execution and rolling optimization results generate two types of control instructions:

[0055] Short-term instructions: directly transmitted to the second-level control step S2 to achieve dynamic adjustment of load and energy storage within minutes;

[0056] Long-term instructions: as future T f The hourly operation reference is provided for the long-term plan correction in step S4.

[0057] Optionally, in step S4, an annual operation planning and adaptive correction mechanism is introduced at the long-term level. This step includes long-term forecasting and plan generation steps.

[0058] The aforementioned long-term forecasting steps are based on historical annual operational data and seasonal meteorological data W. year (d) Electricity demand trend D trend (d) and the annual adjustment factor for new energy output κ RES (d) Construct a long-term load forecasting model:

[0059]

[0060] Where d is the date, f long A time-series forecasting model with decomposed seasonal trends (STLF-SD) is used to separate annual trends from short-term fluctuations;

[0061] The aforementioned plan generation step involves developing an annual flexible load operation plan based on the forecast results. The goal is to minimize operating costs and improve system stability.

[0062]

[0063] Among them, C op (d) represents operating costs; For carbon emission costs; S flex (d) represents the flexibility score for flexible loads; α, β, and γ are weighting coefficients.

[0064] Optionally, step S4 may further include a deviation detection step.

[0065] During operation, the plan deviation rate is calculated by combining the execution feedback from steps S2 and S3:

[0066]

[0067] When |δ plan (t)|>θ dev When this occurs, adaptive correction is triggered.

[0068] Optionally, step S4 may further include an adaptive correction step.

[0069] The correction strategy is based on the idea of ​​rolling re-optimization, mapping the deviation value to the adjustment amount:

[0070]

[0071] Where λ δ η is the deviation compensation coefficient. cost Cost sensitivity coefficient The current cost gradient is used; the revised plan will be synchronously updated to the mid-cycle scheduling in step S3 to achieve cross-cycle linkage adjustment.

[0072] The present invention has the following beneficial effects:

[0073] This invention proposes a multi-timescale coordinated control method for adjustable loads in novel power systems. By constructing a hierarchical load response model covering millisecond, minute, hourly, and daily timescales, and introducing a cross-timescale coordinated optimization mechanism, it achieves full-cycle closed-loop control from millisecond-level rapid disturbance response to annual operation plan optimization. This method can dynamically adjust the operating state of adjustable loads under conditions of fluctuating renewable energy output, changing weather conditions, and uncertain load demand, ensuring frequency and voltage stability while improving the system's economy and decarbonization level. In the short cycle, this invention utilizes a prediction-optimization-execution closed loop to achieve second-level load regulation capabilities to cope with transient disturbances in the power grid. In the medium cycle, it introduces rolling optimization scheduling of source-grid-load-storage to coordinate load regulation with distributed energy and energy storage operation. In the long cycle, it formulates annual operation strategies based on weather forecasts, electricity price trends, and load change patterns, and incorporates short- and medium-cycle feedback for adaptive correction, ensuring long-term operational stability and economy. The method of this invention not only improves the regulation potential and utilization efficiency of adjustable loads in new power systems, but also significantly reduces the reserve capacity requirement, and enhances the safety, flexibility and sustainable operation of the power system under conditions of high proportion of new energy access. Attached Figure Description

[0074] Figure 1 A flowchart of the adjustable load multi-timescale coordinated control method for novel power systems provided by the present invention; Detailed Implementation

[0075] The principles and features of the present invention are described below with reference to the accompanying drawings. The examples given are only for explaining the present invention and are not intended to limit the scope of the present invention.

[0076] Example

[0077] The technical solution of the present invention to solve the above-mentioned technical problems is as follows:

[0078] This invention provides a method for coordinated control of adjustable loads across multiple time scales in novel power systems, with reference to... Figure 1 As shown, the method includes:

[0079] S1: Collect adjustable load and environmental data at the millisecond, minute, and hour levels to build a multi-timescale load response model, providing a precise parameter basis for subsequent rapid control and rolling scheduling;

[0080] This involves deploying multi-source synchronous acquisition devices to simultaneously acquire power operation parameters and environmental parameters on the adjustable load side. Power operation parameters include three-phase voltage U... a U b U c Three-phase current I a ,I b ,I c Active power P(t), reactive power Q(t), and power factor Environmental parameters include temperature T(t), humidity H(t), light intensity L(t), and wind speed V. w (t). The sampling frequency ranges from 5kHz to 20kHz in the millisecond range, 1Hz in the minute range, and 1 / 60Hz in the hour range to ensure the integrity of features at different time scales.

[0081] The multi-source synchronous acquisition device includes a high-speed power parameter acquisition unit, an environmental sensing unit, and a meteorological data interface module. The high-speed power parameter acquisition unit has a sampling frequency of no less than 10kHz and is used to capture transient voltage and current waveforms. The environmental sensing unit records external factors such as temperature, humidity, wind speed, and light intensity in real time. The meteorological data interface module interacts with the meteorological forecasting platform via API to obtain weather change trends for the next few hours to days. After the acquired data is preprocessed (denoising, normalization, and feature extraction) by edge computing nodes, hierarchical regression is used to quickly fit the load response curve under known patterns. Neural networks are used to learn nonlinear and time-varying characteristics, ultimately generating a high-precision load dynamic response model at three time scales: milliseconds, minutes, and hours.

[0082] To ensure the consistency of data across multiple time scales, this invention employs a time scale mapping function. Synchronize the raw data to time series at various scales:

[0083]

[0084] Where X raw X is the original sampling sequence. Δt The sequence is at the target time resolution.

[0085] Based on this, a multi-time-scale load response model is constructed. This invention uses a hierarchical modeling method to represent millisecond-level, minute-level, and hour-level features as follows:

[0086] F ms =f ms (U(t),I(t),T(t),H(t)) (2)

[0087] F min =f min (P(t),Q(t),L(t),V w (t)) (3)

[0088] F h =f h (P avg (h),T avg (h),W forecast (h)) (4) Among them, f ms ,f min ,f h These are feature extraction functions for milliseconds, minutes, and hours, respectively. forecast (h) represents hourly meteorological forecast data.

[0089] To integrate characteristics from different time scales, a multi-time-scale load response prediction function is established:

[0090]

[0091] Φ(·) is a cross-scale feature fusion model that combines hierarchical regression and time attention mechanism, which can simultaneously capture the response characteristics of load to voltage disturbances and environmental changes at different time scales.

[0092] The model's output can be used not only for millisecond-level fast adjustment instruction generation (S2), but also for minute-level rolling optimization (S3) and hour-level operation plan correction (S4), achieving parameter consistency across the entire chain.

[0093] S2: Utilizing millisecond-level and second-level models, combined with real-time operating status, second-level adjustment commands are generated to achieve immediate response to frequency and voltage fluctuations and to provide feedback on the execution results;

[0094] In this process, based on the multi-timescale load response model constructed in step S1, a second-level fast control algorithm is introduced. The second-level fast control algorithm includes: disturbance detection, second-level predictive control, constraint solving, and command issuance and feedback.

[0095] The disturbance detection is performed by real-time monitoring of the bus voltage U(t) and the system frequency f(t) to calculate the instantaneous deviation.

[0096] ΔU(t)=U(t)-U ref Δf(t)=f(t)-f ref (6)

[0097] U ref and f ref These represent the rated voltage and rated frequency, respectively. When |Δf(t)|>δ U Or |Δf(t)|>δ f When this happens, the quick adjustment mode is triggered.

[0098] The aforementioned second-level predictive control involves the predictive control unit calling millisecond-level and second-level response models f. ms with f s In the prediction time domain T p Internal estimation of the power response curve of adjustable load:

[0099]

[0100] Where k∈[1,T] p ], where ω is the environmental factor weight coefficient and E(t+k) is the predicted environmental state vector.

[0101] The constraint solution involves constructing a second-level optimization objective function based on the prediction results:

[0102]

[0103] Where α, β, and γ are the weights of frequency, voltage deviation, and regulation cost, respectively, and C adj This is the cost function for load adjustment.

[0104] The constraints include:

[0105] P min ≤P load (t+k)+u(t+k)≤P max (9)

[0106] |u(t+k)-u(t+k-1)|≤Rmax (10)

[0107] The solution is obtained using a fast iterative method based on quadratic programming (QP), which guarantees that the optimal solution is found within 100ms.

[0108] The aforementioned instruction issuance and feedback optimizes the result u. * (t) is sent as a second-level power adjustment command to the load controller at the execution end via a low-latency communication link (latency <20ms). The controller adjusts its operating status according to the command. The execution end feeds back the current power change ΔP every second. exec and execution success rate η exec The data is sent to the control center for rolling scheduling correction in subsequent step S3.

[0109] In step S2, the fast optimization algorithm based on model predictive control (MPC) uses the multi-timescale load model from step S1 to predict the system state for the next 0.5 to 2 seconds in real time; the objective function is resolved in each control cycle, and the weights of the objective function are dynamically adjusted according to the grid frequency deviation, voltage deviation and load regulation cost; at the same time, the upper limit of equipment power regulation, operating safety margin and user comfort limit are considered, and the second-level power regulation command is sent to the execution load through a low-latency communication link (latency less than 20ms) to realize closed-loop control and real-time response.

[0110] Through the above method, the present invention achieves a second-level closed-loop rapid adjustment capability. When frequency and voltage disturbances caused by fluctuations in the output of new energy sources occur, it can make power adjustments within 1 second, which significantly improves the dynamic stability of the new power system.

[0111] S3: Based on minute-to-hour forecast data, comprehensively consider adjustable load, distributed energy resources and energy storage constraints, continuously optimize load operation plans, and coordinate the joint operation of power generation, grid, load and storage;

[0112] Specifically, a rolling forecasting-optimization-execution closed-loop mechanism is introduced on a timescale ranging from minutes to hours to achieve coordinated scheduling of adjustable loads with distributed energy resources and energy storage systems. This mechanism includes rolling forecasting, rolling optimization, and coordinated execution.

[0113] The rolling forecast utilizes the minute-level and hour-level load response models from step S1, combined with new energy output forecasts. and the energy storage state of charge (SOC(t)) to generate the future T f Hourly multi-energy power prediction sequence:

[0114]

[0115] in This indicates the predicted charging and discharging power of the energy storage system.

[0116] The aforementioned rolling optimization uses ΔT = 30 minutes as the default optimization period and dynamically adjusts the future scheduling scheme based on real-time operational data. The optimization objective function is:

[0117]

[0118] Among them, u load Load adjustment command vector; u ESS C is the energy storage charge / discharge control vector; load C is the load adjustment cost function; ESS This comes at the cost of energy storage lifespan loss and energy consumption.

[0119] The constraints include:

[0120] (1) Adjustable load range:

[0121]

[0122] (2) Energy storage constraints:

[0123] SOC min， ≤SOC(t+k)≤SOC max (14)

[0124] (3) Constraints on energy storage charge and discharge rates:

[0125] The solution employs mixed integer quadratic programming (MIQP), which supports hybrid optimization of discrete load switching control and continuous energy storage power control, ensuring a solution time of less than 5 seconds within the optimization cycle.

[0126] The coordinated execution and rolling optimization results generate two types of control instructions:

[0127] Short-term commands: directly transmitted to the S2 second-level control module to achieve dynamic adjustment of load and energy storage within minutes;

[0128] Long-term instructions: as future T f The hourly operating reference is provided to the S4 long-cycle plan correction module.

[0129] In step S3, the medium-cycle rolling optimization scheduling uses a basic cycle of 30 minutes and also has a dynamic cycle adjustment function. When it is predicted that the output of new energy sources or the load will fluctuate drastically, the cycle is automatically shortened to 10 minutes. During the optimization scheduling, the power probability distribution prediction of distributed photovoltaic and wind power is introduced. Combined with the state of charge (SOC) of energy storage, charge and discharge efficiency, and cycle life constraints, a multi-objective mixed integer linear programming (MILP) model is adopted to optimize the power allocation strategy of source-grid-load-storage under the premise of ensuring grid safety margin and user comfort, and to generate a backup scheduling plan for the next two scheduling cycles to improve scheduling robustness.

[0130] In this way, step S3 can ensure optimal coordination of power flow among source (new energy output), grid (grid dispatch constraints), load (adjustable load), and storage (energy storage device) on a time scale of minutes to hours, thereby improving the economy and stability of system operation.

[0131] S4: Combine daily and higher-level forecast data to formulate long-term load operation plans and adaptively correct them based on short- and medium-term cycle execution deviations to ensure economic efficiency and stability throughout the year.

[0132] This invention introduces an annual operation planning and adaptive correction mechanism at the long-term level to achieve seasonal optimization and dynamic adjustment of the adjustable load strategy. This process includes four core steps: long-term forecasting, plan generation, deviation detection, and correction execution.

[0133] 1. Long-term forecast

[0134] Based on historical annual operational data and seasonal meteorological data W year (d) Electricity demand trend D trend (d) and the annual adjustment factor for new energy output κ RES (d) Construct a long-term load forecasting model:

[0135]

[0136] Where d is the date, f long A time-series forecasting model with decomposed seasonal trends (STLF-SD) is used to separate annual trends from short-term fluctuations.

[0137] 2. Plan Generation

[0138] Develop an annual adjustable load operation plan based on forecast results. The goal is to minimize operating costs and improve system stability.

[0139]

[0140] Among them, C op (d) represents operating costs; For carbon emission costs; S flex (d) represents the adjustable load flexibility score; α, β, and γ are weighting coefficients.

[0141] 3. Deviation detection

[0142] During operation, the plan deviation rate is calculated by combining the execution feedback from S2 and S3:

[0143]

[0144] When |δ plan (t)|>θ dev When the threshold of 3% is reached, adaptive correction is triggered.

[0145] 4. Adaptive correction

[0146] The correction strategy is based on the idea of ​​rolling re-optimization, mapping the deviation value to the adjustment amount:

[0147]

[0148] Where λ δ η is the deviation compensation coefficient. cost Cost sensitivity coefficient This represents the current cost gradient. The revised plan will be synchronously updated to the S3 mid-cycle scheduling module, enabling cross-cycle coordinated adjustments.

[0149] In step S4, the long-term operation plan incorporates the annual electricity price forecast curve, the seasonal decomposition model of electricity demand, and the meteorological correction factor for renewable energy output based on the seasonal dispatch model to achieve annual optimization of the load strategy. During the execution of the plan, deviation accumulation analysis is performed using short- and medium-term execution data. When the cumulative deviation exceeds 3% or the operating economic indicators fall below the set threshold, the adaptive optimization process is automatically triggered, and the rescheduling module is invoked to adjust subsequent plans, ensuring that the long-term operation goals are consistent with the actual execution and improving the full-cycle control capability of adjustable loads.

[0150] Through the above steps, S4 can ensure that the long-term operation strategy of adjustable load remains optimal in response to seasonal climate changes, fluctuations in new energy annual conditions, and changes in load trends, while maintaining closed-loop consistency with the short and medium-term execution results, thus achieving dual guarantees of economy and stability.

[0151] This invention also provides a storage medium storing a computer program. When executed by a processor, the computer program implements some or all of the steps in various embodiments of the flexible load multi-timescale coordinated control method for novel power systems provided by this invention. The storage medium may be a magnetic disk, optical disk, read-only memory (ROM), or random access memory (RAM), etc.

[0152] Those skilled in the art will clearly understand that the techniques in the embodiments of the present invention can be implemented using software plus necessary general-purpose hardware platforms. Based on this understanding, the technical solutions in the embodiments of the present invention, or the parts that contribute to the prior art, can be embodied in the form of a software product. This computer software product can be stored in a storage medium, such as ROM / RAM, magnetic disk, optical disk, etc., and includes several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute the methods described in various embodiments or certain parts of the embodiments of the present invention.

[0153] This invention proposes a multi-timescale coordinated control method for adjustable loads in novel power systems. It aims to construct a hierarchical modeling and collaborative optimization system for adjustable loads covering millisecond, minute, hourly, and daily timescales. Through the organic integration of short-cycle rapid response, medium-cycle rolling scheduling, and long-cycle plan correction, it achieves a synergistic improvement in the operational stability, economy, and flexibility of novel power systems under conditions of high-proportion renewable energy integration. This method can cope with rapid fluctuations in renewable energy power while also considering long-term global optimization, enabling adjustable loads to play a highly efficient role in various scenarios such as frequency regulation, voltage control, peak shaving and valley filling, and ancillary services. This effectively alleviates system regulation pressure, reduces reserve capacity requirements, and enhances the sustainable operation capability of the power system.

[0154] The above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the protection scope of the present invention.

Claims

1. A flexible load multi-timescale coordinated control method for novel power systems, characterized in that, Includes the following steps: S1: Collect millisecond, minute, and hourly flexible load and environmental data to construct a multi-timescale load response model, providing a parameter basis for subsequent control and rolling scheduling; S2: Utilizing millisecond-level and second-level models, combined with real-time operating status, second-level adjustment commands are generated to achieve immediate response to frequency and voltage fluctuations and to provide feedback on the execution results; S3: Based on minute-to-hour forecast data, comprehensively considers flexible load, distributed energy resources and energy storage constraints, continuously optimizes load operation plans, and coordinates the joint operation of power generation, grid, load and storage; S4: Combine daily and higher-level forecast data to formulate long-term load operation plans and adaptively correct deviations based on short- and medium-term cycle execution.

2. The flexible load multi-timescale coordinated control method for novel power systems according to claim 1, characterized in that, In step S1, a multi-source synchronous acquisition device is deployed to simultaneously acquire power operation parameters and environmental parameters on the flexible load side; the power operation parameters include three-phase voltage U. a U b U c Three-phase current I a ,I b ,I c Active power P(t), reactive power Q(t), and power factor Environmental parameters include temperature T(t), humidity H(t), light intensity L(t), and wind speed V. w (t).

3. The flexible load multi-timescale coordinated control method for novel power systems according to claim 2, characterized in that, In step S1, the multi-source synchronous acquisition device includes a high-speed power parameter acquisition unit, an environmental sensing unit, and a meteorological data interface module. The high-speed power parameter acquisition unit has a sampling frequency range of 10kHz to 50kHz and is used to capture transient voltage and current waveforms. The environmental sensing unit records external factors such as temperature, humidity, wind speed, and light intensity in real time. The meteorological data interface module interacts with the meteorological forecasting platform via API to obtain the weather change trend for the next few hours to days. After the multi-source synchronously acquired data is preprocessed by edge computing nodes, hierarchical regression is used to quickly fit the load response curve under the known pattern. The neural network is used to learn nonlinear and time-varying characteristics, and finally a high-precision load dynamic response model is generated on three time scales: millisecond, minute, and hour.

4. The flexible load multi-timescale coordinated control method for novel power systems according to claim 1, characterized in that, In step S1, a time-scale mapping function is used. Synchronize the raw data to time series at various scales: Where X raw X is the original sampling sequence. Δt For sequences at the target time resolution; Based on this, a multi-timescale load response model is constructed; using a hierarchical modeling method, the millisecond, minute, and hourly features are represented as follows: F ms =F ms (U(t),I(t),T(t),H(t)) (2) F min =f min (P(t),Q(t),L(t),V w (t)) (3) F h =f h (p avg (h),T avg (h),W forecast (h)) (4) Among them, f ms ,f min ,f h These are feature extraction functions for milliseconds, minutes, and hours, respectively. forecast (h) represents hourly meteorological forecast data; To integrate characteristics from different time scales, a multi-time-scale load response prediction function is established: Φ(·) is a cross-scale feature fusion model that combines hierarchical regression and time attention mechanism, which can simultaneously capture the response characteristics of load to voltage disturbances and environmental changes at different time scales.

5. The flexible load multi-timescale coordinated control method for novel power systems according to claim 1, characterized in that, In step S2, based on the multi-time-scale load response model constructed in step S1, a second-level fast control algorithm is introduced. The second-level fast control algorithm includes: disturbance detection, second-level predictive control, constraint solving, and command issuance and feedback. The disturbance detection is performed by real-time monitoring of the bus voltage U(t) and the system frequency f(t) to calculate the instantaneous deviation. ΔU(t)=U(t)-U ref ,Δf(t)=f(t)-f ref (6) U ref and f ref Given the rated voltage and rated frequency, when |Δf(t)|>δ U Or |Δf(t)|>δ f When this happens, the quick adjustment mode is triggered; The aforementioned second-level predictive control calls millisecond-level and second-level response models f. ms with f s In the prediction time domain T p Internal estimation of the power response curve of flexible load: Where k∈[1,T] p ], where ω is the environmental factor weight coefficient and E(t+k) is the predicted environmental state vector; The constraint solution involves constructing a second-level optimization objective function based on the prediction results: Where α, β, and γ are the weights of frequency, voltage deviation, and regulation cost, respectively, and C adj The load adjustment cost function; The constraints include: P min ≤P load (t+k)+u(t+k)≤P max (9) |u(t+k)-u(t+k-1)|≤R max (10) The solution is obtained using a fast iterative method based on quadratic programming (QP), which guarantees that the optimal solution is found within 100ms. The aforementioned instruction issuance and feedback optimizes the result u. * (t) is sent to the load controller at the execution end via a low-latency communication link as a second-level power adjustment command. The controller adjusts the operating status according to the command, and the execution end feeds back the current power change ΔP every second. exec and execution success rate η exec The data is sent to the control center for rolling scheduling correction in subsequent step S3.

6. The flexible load multi-timescale coordinated control method for novel power systems according to claim 1, characterized in that, In step S3, a rolling prediction-optimization-execution closed-loop mechanism is introduced on a time scale from minutes to hours. This mechanism includes rolling prediction, rolling optimization, and coordinated execution. The rolling forecast utilizes the minute-level and hour-level load response models from step S1, combined with new energy output forecasts. and the energy storage state of charge (SOC(t)) to generate the future T f Hourly multi-energy power prediction sequence: in This indicates the predicted charging and discharging power of the energy storage system; The aforementioned rolling optimization uses ΔT = 30 min as the default optimization period and dynamically adjusts the future scheduling scheme based on real-time operational data. The optimization objective function is: Among them, u load Load adjustment command vector; u ESS C is the energy storage charge / discharge control vector; load C is the load adjustment cost function; ESS This comes at the cost of energy storage lifespan loss and energy consumption. The constraints include: (1) Adjustable load range: (2) Energy storage constraints: SOC min， ≤SOC(t+k)≤SOC max (14) (3) Constraints on energy storage charge and discharge rates: The solution employs mixed integer quadratic programming (MIQP), which supports hybrid optimization of discrete load switching control and continuous energy storage power control, ensuring that the solution time is less than 5 seconds within the optimization cycle. The coordinated execution and rolling optimization results generate two types of control instructions: Short-term instructions: directly transmitted to the second-level control step S2 to achieve dynamic adjustment of load and energy storage within minutes; Long-term instructions: as future T f The hourly operation reference is provided for the long-term plan correction in step S4.

7. The flexible load multi-timescale coordinated control method for novel power systems according to claim 1, characterized in that, In step S4, an annual operation planning and adaptive correction mechanism is introduced at the long-term level. This step includes long-term forecasting and plan generation steps. The aforementioned long-term forecasting steps are based on historical annual operational data and seasonal meteorological data W. year (d) Electricity demand trend D trend (d) and the annual adjustment factor for new energy output k RES (d) Construct a long-term load forecasting model: Where d is the date, f long A time-series forecasting model with decomposed seasonal trends (STLF-SD) is used to separate annual trends from short-term fluctuations; The aforementioned plan generation step involves developing an annual flexible load operation plan based on the forecast results. The goal is to minimize operating costs and improve system stability. Among them, C op (d) represents operating costs; For carbon emission costs; S flex (d) represents the flexibility score of the flexible load; α, β, γ are the weighting coefficients.

8. The flexible load multi-timescale coordinated control method for novel power systems according to claim 7, characterized in that, Step S4 also includes a deviation detection step. During operation, the plan deviation rate is calculated by combining the execution feedback from steps S2 and S3: When |δ plan (t)|>θ dev When this occurs, adaptive correction is triggered.

9. A flexible load multi-timescale coordinated control method for novel power systems according to claim 7, characterized in that, Step S4 also includes an adaptive correction step. The correction strategy is based on the idea of ​​rolling re-optimization, mapping the deviation value to the adjustment amount: Where λ δ η is the deviation compensation coefficient. cost Cost sensitivity coefficient This represents the current cost gradient; The revised plan will be updated synchronously to the mid-cycle scheduling in step S3, realizing cross-cycle linkage adjustment.

Images