A cloud-edge-device collaborative management and control system based on the power Internet of Things

Abstract

This invention relates to the field of power distribution technology, and more particularly to a cloud-edge-device collaborative management and control system based on the power Internet of Things (IoT). It includes: acquiring real-time load fluctuation and new energy generation uncertainty data from a global data source in the cloud, generating an initial power parameter combination, and selecting the optimal combination as the global optimization parameter set; adjusting the global optimization parameter set based on local device characteristics to determine refined power setpoints; if the deviation exceeds a preset threshold, simulating a time-series sequence under a multi-device collaborative scenario, generating a time-series coordination scheme and reconfiguring parameters to obtain a coordinated and consistent device response sequence; optimizing unstable points in the response sequence to determine a stable power allocation path; judging path adaptability and generating correction instructions to be transmitted to terminal devices; terminal devices performing synchronous adjustments and tracking grid frequency and voltage changes in real time, ultimately achieving a smooth mode switching. This invention effectively improves the stability and energy utilization efficiency of the power grid in complex environments.

Description

Technical Field

[0001] This invention relates to the field of power management and control technology, and in particular to a cloud-edge-device collaborative management and control system based on the power Internet of Things. Background Technology

[0002] Against the backdrop of the current energy transition and the rapid development of smart grids, the proportion of renewable energy generation is continuously increasing, and the share of clean energy such as wind power and photovoltaics in the power system is rising steadily. However, these renewable energy sources have inherent characteristics of high volatility and unpredictability, causing the power system to face problems such as frequent frequency and voltage fluctuations and increased difficulty in load balancing during operation. At the same time, with the increasing diversification and uncertainty of end-user electricity demand, the traditional grid operation mode relying on centralized dispatch and single control strategies is no longer able to meet the requirements of real-time performance and stability.

[0003] Most existing power management solutions rely on centralized cloud processing. While this can achieve some global optimization at the macro level, it often neglects the response characteristics of local devices. This frequently leads to discrepancies between dispatch commands and device capabilities during actual execution, resulting in uneven power distribution, response lags, and decreased system stability. Especially in scenarios with high penetration of renewable energy, power systems need to simultaneously resolve the contradiction between global optimization and local adaptability to avoid significant disturbances during mode switching, or even grid instability or power outages.

[0004] Therefore, there is an urgent need to propose a collaborative management and control system for the power Internet of Things that can combine global optimization in the cloud with local fine-grained control at the edge. This system should not only utilize the powerful data processing capabilities of the cloud to perform global modeling and optimization of new energy power generation and load fluctuations, but also introduce real-time acquisition and adjustment of equipment characteristic data at the edge to achieve localized parameter correction. In scenarios of mode switching and uncertain disturbances, intelligent algorithms should be used to dynamically coordinate the response sequence and power distribution among devices to ensure grid frequency and voltage stability and improve the overall reliability of operation and the efficiency of new energy utilization. Summary of the Invention

[0005] To address the aforementioned issues, this invention provides a cloud-edge-device collaborative management and control system based on the power Internet of Things (IoT) to improve the stability and energy utilization efficiency of the power system. In particular, it addresses the challenges posed by load fluctuations and generation uncertainties in the context of the rapid development of new energy sources, achieving an effective combination of global optimization and local fine-tuning to ensure the stable operation of the power grid in complex environments.

[0006] This invention discloses a cloud-edge-device collaborative management and control system based on the power Internet of Things, which mainly includes:

[0007] The data acquisition unit is used to acquire real-time load fluctuation and uncertainty data of new energy power generation from global data sources in the cloud, and process them through particle swarm optimization algorithm to generate multiple initial power parameter combinations, and select the optimal combination as the global optimization parameter set;

[0008] The local adjustment unit is used to obtain the characteristic data of the local device at the edge end according to the global optimization parameter set, and to perform localized adjustment of the global optimization parameter set based on the characteristics of the local device to determine the refined power setting value.

[0009] The timing coordination unit is used to generate a timing coordination scheme by simulating the timing sequence under the multi-device collaborative scenario through reinforcement learning if the deviation between the refined power setting value and the current state of the terminal device exceeds a preset threshold; extract the synchronization gain coefficient and response time constant according to the timing coordination scheme; reconfigure parameters for sudden changes that occur during mode switching; and obtain a coordinated and consistent device response sequence.

[0010] The global verification unit is used to perform global verification through the device response sequence, and uses the particle swarm optimization algorithm to iteratively optimize the unstable points in the response sequence to determine a stable power allocation path.

[0011] The instruction correction unit is used to determine the adaptability of the stable power distribution path based on real-time monitoring data feedback. If there are local fluctuations, it generates a correction instruction and transmits the correction instruction to the terminal device.

[0012] The terminal control unit is used to perform synchronous adjustment according to the correction command, track the changes in grid frequency and voltage in real time, and finally obtain a smooth mode switching result.

[0013] As a preferred embodiment of the present invention, the data acquisition unit is configured as follows:

[0014] Real-time load fluctuation data and new energy power generation uncertainty data are obtained from a global data source via the cloud. A particle swarm optimization algorithm is used to process the real-time load fluctuation data and the new energy power generation uncertainty data to generate multiple initial power parameter combinations. For each of the multiple initial power parameter combinations, a global optimization objective function value is calculated, where the global optimization objective function value is based on load balance and power generation efficiency. Based on the global optimization objective function value, the optimal combination is selected from the multiple initial power parameter combinations to obtain a global optimization parameter set.

[0015] As a preferred embodiment of the present invention, the local adjustment unit is configured as follows:

[0016] Acquire local device characteristic data at the edge, wherein the local device characteristic data includes response time constant and gain coefficient; adjust the global optimization parameter set locally based on the local device characteristic data to generate an adjusted parameter set; calculate the power allocation adaptability value for each parameter in the adjusted parameter set, wherein the power allocation adaptability value is based on device response speed and power stability; determine a refined power setting value based on the power allocation adaptability value.

[0017] As a preferred embodiment of the present invention, a timing coordination unit is used to generate a timing coordination scheme, including:

[0018] The system acquires the current state data of the terminal device and calculates the deviation between the refined power setting value and the current state data of the terminal device. If the deviation exceeds a preset threshold, it initializes a reinforcement learning model, which includes a state space, an action space, and a reward function. The refined power setting value and the current state data of the terminal device are used as inputs to the state space. Through the reinforcement learning model, an action sequence in a multi-device collaborative scenario is generated, where the action sequence represents the power adjustment actions of each device. Based on the reward function, the action sequence is optimized to generate a time-series sequence.

[0019] Based on the timing sequence, a timing coordination scheme is generated, wherein the timing coordination scheme includes the power adjustment time point and adjustment range of each device.

[0020] As a preferred embodiment of the present invention, a timing coordination unit is used to reconfigure parameters in response to sudden changes that occur during mode switching, thereby obtaining a coordinated and consistent device response sequence, including:

[0021] The synchronization gain coefficient and response time constant of each device are extracted from the timing coordination scheme; the parameter reconfiguration requirement value of each device is calculated for sudden changes during mode switching; the synchronization gain coefficient and response time constant are adjusted according to the parameter reconfiguration requirement value to generate a reconfigured parameter set; and a coordinated device response sequence is generated according to the reconfigured parameter set, wherein the device response sequence includes the power output timing of each device.

[0022] As a preferred embodiment of the present invention, a global verification unit is used to determine a stable power allocation path, including:

[0023] The device response sequence is globally verified in the cloud to obtain frequency simulation results and voltage simulation results; the particle swarm optimization algorithm is used to iteratively optimize the unstable points in the frequency simulation results and voltage simulation results; based on the results of the iterative optimization, a stable power allocation path is generated, wherein the stable power allocation path includes the power allocation ratio and timing arrangement of each device.

[0024] As a preferred embodiment of the present invention, the instruction correction unit is used to generate correction instructions, including:

[0025] Real-time monitoring data is acquired at the edge, including equipment operating status and power grid fluctuation data; based on the real-time monitoring data, the local fluctuation value of the stable power distribution path is calculated; if the local fluctuation value exceeds a preset fluctuation threshold, a path adaptability judgment result is generated; based on the path adaptability judgment result, a correction instruction is generated, wherein the correction instruction includes adjustment parameters of the power distribution path.

[0026] As a preferred embodiment of the present invention, the terminal control unit is configured as follows:

[0027] Obtain the correction instruction; adjust the power output parameters of each device according to the correction instruction; track the frequency response data of each device in real time in response to changes in grid frequency; track the voltage response data of each device in real time in response to changes in grid voltage; generate a stable mode switching result based on the frequency response data and the voltage response data, wherein the stable mode switching result includes the stable operating state of the grid.

[0028] As a preferred embodiment of the present invention, the data acquisition unit is used to generate multiple initial power parameter combinations, including:

[0029] Initialize the particle swarm, where each particle represents an initial power parameter combination; for each particle, calculate its fitness value, where the fitness value is based on load fluctuation smoothness and renewable energy generation utilization rate;

[0030] Based on the fitness value, the position and velocity of the particles are updated to generate new power parameter combinations; the particle swarm is iteratively updated until a preset convergence condition is met to obtain multiple initial power parameter combinations.

[0031] This invention also provides a cloud-edge-device collaborative management and control method based on the power Internet of Things. Based on the above system implementation, the method includes:

[0032] Step S1: Obtain real-time load fluctuation and new energy power generation uncertainty data from the global data source in the cloud, process them through the particle swarm algorithm to generate multiple initial power parameter combinations, and select the optimal combination as the global optimization parameter set;

[0033] Step S2: Obtain the characteristic data of the local device at the edge based on the global optimization parameter set, and make localized adjustments to the global optimization parameter set based on the characteristics of the local device to determine the refined power setting value;

[0034] Step S3: If the deviation between the refined power setting value and the current state of the terminal device exceeds a preset threshold, a timing sequence is generated by simulating the timing sequence under the multi-device collaborative scenario through reinforcement learning; the synchronization gain coefficient and response time constant are extracted according to the timing coordination scheme, and parameters are reconfigured for sudden changes that occur during mode switching to obtain a coordinated and consistent device response sequence.

[0035] Step S4: Perform global verification through the device response sequence, use particle swarm optimization algorithm to iteratively optimize unstable points in the response sequence, and determine a stable power allocation path; based on real-time monitoring data feedback, determine the adaptability of the stable power allocation path, and if there are local fluctuations, generate a correction instruction and transmit the correction instruction to the terminal device;

[0036] Step S5: Perform synchronous adjustment according to the correction instruction, track the changes in grid frequency and voltage in real time, and finally obtain a smooth mode switching result.

[0037] The technical solutions provided by the embodiments of the present invention may include the following beneficial effects:

[0038] This invention acquires real-time load fluctuation and renewable energy generation uncertainty data from the cloud, and generates multiple initial power parameter combinations using a particle swarm optimization algorithm. The optimal combination is selected as the global optimization parameter set, providing foundational data for subsequent local adjustments and ensuring the effectiveness of the optimization scheme under the actual conditions of grid load demand and renewable energy generation. Based on the characteristic data of local equipment, the global optimization parameter set is locally adjusted to refine the power setpoints and adapt them to the specific response capabilities of the equipment, such as the equipment's response time constant and gain coefficient. This local adjustment ensures the timeliness and stability of equipment responses and reduces scheduling errors caused by differences in local characteristics. During grid operation, if the refined power setpoint deviates from the current state of the terminal equipment by more than a preset threshold, a reinforcement learning algorithm is used to simulate a multi-device collaborative scenario and generate a timing coordination scheme. This not only achieves power coordination between devices through reinforcement learning, but also... The optimization also considers potential sudden changes during grid mode switching. By adjusting the synchronization gain coefficient and response time constant, it ensures that all equipment can respond in a coordinated manner to sudden load or renewable energy fluctuations. The optimized equipment response sequence is processed through global verification, and the particle swarm optimization algorithm is used to iteratively optimize unstable points, ultimately forming a stable power distribution path to ensure system stability under different operating modes. When real-time monitoring data feedback shows local fluctuations in the power distribution path, correction instructions are generated based on this feedback and transmitted to the terminal equipment. The terminal equipment performs synchronization adjustments and tracks grid frequency and voltage changes in real time. Through precise data tracking and adjustment, it ensures that the grid can maintain stable operation under various disturbances. Through the cooperation of the above technical solutions, the grid can achieve smooth mode switching in the face of sudden fluctuations and uncertainties, significantly improving the grid's robustness and energy utilization efficiency. Attached Figure Description

[0039] Figure 1 This is a structural diagram of a cloud-edge-device collaborative management and control system based on the power Internet of Things in an embodiment of the present invention;

[0040] Figure 2 This is a diagram illustrating the effect of the correction instruction in an embodiment of the present invention;

[0041] Figure 3 This is a flowchart of a cloud-edge-device collaborative management and control method based on the power Internet of Things in an embodiment of the present invention. Detailed Implementation

[0042] To make the objectives, technical solutions, and advantages of this invention clearer, the invention will be described in detail below with reference to the accompanying drawings and specific embodiments.

[0043] With the large-scale grid connection of new energy sources such as wind power and photovoltaics, the operation of the power system exhibits significant characteristics of high volatility and high uncertainty. Random changes in wind speed and solar illumination conditions directly lead to strong uncertainty in the output of new energy sources. Traditional grid load forecasting and power allocation methods mostly rely on historical data and static models, lacking the ability to adapt to real-time dynamics and effectively cope with instantaneous fluctuations. As a result, the power system is prone to problems such as uneven power distribution, frequency and voltage instability in actual operation. Secondly, under the traditional dispatch framework, the dispatch center often uses global optimization methods to generate power allocation instructions. However, these instructions fail to fully consider the differences in response characteristics between different devices. For example, the response time constants and gain coefficients of thermal power units, energy storage units and new energy devices are significantly different. When global dispatch instructions are issued to the terminal, deviations often occur because local devices cannot keep up in time, causing the actual system execution effect to deviate from the target plan. This contradiction between global optimization and local execution has become a major bottleneck restricting the stable operation of the power grid. Thirdly, the complexity of multi-device collaborative control exacerbates the problem. In large-scale renewable energy integration scenarios, multiple devices need to adjust their power in a predetermined order and magnitude within a short period to ensure grid stability. However, existing methods often rely solely on empirical rules when handling multi-device collaboration, lacking deep learning and adaptive capabilities to handle dynamic uncertainties. This can easily lead to transient shocks and system oscillations during mode switching. Therefore, methods such as... Figure 1 This embodiment proposes a cloud-edge-device collaborative management and control system based on the power Internet of Things, specifically including:

[0044] The data acquisition unit is used to acquire real-time load fluctuation and uncertainty data of new energy power generation from a global data source in the cloud, process the data using a particle swarm optimization algorithm, generate multiple initial power parameter combinations, and select the optimal combination as the global optimization parameter set; specifically including:

[0045] Real-time load fluctuation data and new energy power generation uncertainty data are obtained from a global data source via the cloud. A particle swarm optimization algorithm is used to process the real-time load fluctuation data and the new energy power generation uncertainty data to generate multiple initial power parameter combinations. For each of the multiple initial power parameter combinations, a global optimization objective function value is calculated, where the global optimization objective function value is based on load balance and power generation efficiency. Based on the global optimization objective function value, the optimal combination is selected from the multiple initial power parameter combinations to obtain a global optimization parameter set.

[0046] Specifically, real-time load fluctuation data and uncertainties in renewable energy generation are acquired from a global data source via the cloud, and a global optimization parameter set is generated based on this. The cloud server accesses a distributed database and a multi-source sensor network to read in real-time load fluctuation data reflecting the dynamics of grid demand, as well as uncertainties such as wind speed and solar radiation characterizing the randomness of renewable energy generation. The cloud processing unit iteratively processes this real-time data using a particle swarm optimization algorithm. In the initialization phase, the load fluctuation data and renewable energy uncertainty data are mapped into multiple particle vectors. Each particle vector contains two parameter dimensions: power output level and reserve capacity. The reserve capacity refers to the amount of electricity available in the grid. During the operation of the power system, a portion of dispatchable power capacity is reserved to cope with load fluctuations, uncertainties in new energy power generation, and sudden faults. In the iterative update phase, each particle continuously adjusts its position and velocity based on its current fitness value and historical and global optimal solutions, thereby gradually approaching the optimal power allocation solution set. The particle swarm optimization algorithm achieves parallel search of multi-dimensional parameters by simulating swarm behavior, ensuring that several convergent solutions can be obtained simultaneously under different load scenarios. These convergent solutions are combinations of multiple initial power parameters. Furthermore, for these multiple initial combinations, a global optimization objective function is constructed, and its function value is calculated one by one. This objective function consists of a load balance term and a power generation efficiency term. The load balancing term is calculated by determining the mean square error between the actual load and the predicted load, resulting in a balance score. The predicted load is obtained through a prediction model, which will be detailed below. This balance score measures the degree to which power output matches demand. The generation efficiency term is calculated by determining the ratio of actual renewable energy output to total power demand, resulting in an efficiency score that reflects the renewable energy absorption capacity. During function value calculation, the two indicators are weighted and summed using preset weights. For example, the weights of the balance and efficiency terms can be set to 0.6 and 0.4 respectively to ensure system stability while maximizing renewable energy utilization, especially in high-uncertainty scenarios. An uncertainty penalty function is introduced into the objective function. This penalty function is quantified by the variance of the input data and multiplied by a penalty factor to enhance the robustness of the generated power allocation path. Through the above calculations, each initial power parameter combination corresponds to a unique objective function value, thus forming a logical mapping relationship between data input, parameter combinations, and objective function values. The initial combination with the smallest objective function value is selected as the global optimization parameter set to ensure that this combination has optimal performance in terms of both load balance and renewable energy utilization. The above technical solution realizes efficient optimization and reasonable selection of global parameters in the power Internet of Things environment, laying a data and parameter foundation for subsequent local fine-tuning at the edge.

[0047] Furthermore, the data acquisition unit is used to generate multiple combinations of initial power parameters, including:

[0048] Initialize the particle swarm, where each particle represents an initial power parameter combination; for each particle, calculate its fitness value, where the fitness value is based on load fluctuation smoothness and renewable energy generation utilization rate;

[0049] Based on the fitness value, the position and velocity of the particles are updated to generate new power parameter combinations; the particle swarm is iteratively updated until a preset convergence condition is met to obtain multiple initial power parameter combinations.

[0050] Specifically, in one embodiment, a particle swarm is initialized in the cloud, where each particle corresponds to a power parameter combination. This power parameter combination includes key parameters such as power output level and reserve capacity to comprehensively characterize the potential power allocation state of the power system under different operating conditions. A fitness value is calculated for each particle. This fitness value is not a single indicator but consists of two parts: load fluctuation smoothness and renewable energy generation utilization rate. The load fluctuation smoothness is obtained by normalizing the standard deviation of real-time load fluctuation data. The standard deviation reflects the magnitude of changes in grid demand, and the normalization process eliminates the influence of different dimensions, making the results comparable. The renewable energy generation utilization rate is based on... The ratio between actual and predicted power generation from renewable energy sources is calculated. The predicted power generation is obtained from the prediction model, ensuring that the utilization rate reflects both the real-time and predictive nature of renewable energy output. During the training phase, the prediction model uses historical data accumulated over a long period, including historical meteorological data such as wind speed curves, sunshine curves, weather records such as cloud cover, wind speed, humidity, air pressure, and power generation duration, along with corresponding historical power generation and output of the generating units. During the operation phase, the model inputs recent meteorological forecast data and outputs predicted power generation values ​​for a future period. The load fluctuation smoothness and renewable energy power generation utilization rate are weighted and summed according to preset weights to form a unique... Corresponding to the fitness value of the particle, the aforementioned weight settings combine the dual requirements of grid stability and energy utilization efficiency. For example, the smoothness weight is increased in high-fluctuation scenarios, while the utilization rate weight is increased when new energy sources are abundant, thereby ensuring that the fitness value can dynamically reflect the comprehensive performance under the operating environment. Through the calculation of the above fitness value, a mapping relationship from real-time data input to fitness output is established, so that each power parameter combination has a quantitative evaluation standard. The particle swarm optimization algorithm also iteratively updates the velocity and position of the particles based on the fitness value. The particle position represents the specific value of the power parameter combination in multi-dimensional space, and the velocity represents its dynamic trend of evolution towards a better direction. In each iteration, the particle... Guided by both its historical best and global best solutions, the particle swarm continuously adjusts its parameters to gradually approach the optimal power allocation path. During the iteration process, real-time load fluctuation data and new energy uncertainty data are continuously updated to ensure that the particle swarm evolution results closely match the actual grid operation status. As the iteration progresses, when the preset convergence condition is met, multiple particle positions in the particle swarm converge, resulting in multiple initial power parameter combinations. These parameter combinations cover the power allocation requirements under different new energy access scenarios for wind power and photovoltaics. For example, in a wind-dominated regional power grid, the fitness value corresponding to the combination of low load fluctuation smoothness and high utilization rate is relatively high, thereby guiding the particles to evolve in a direction that is conducive to reducing the impact of intermittent wind power.In scenarios primarily powered by photovoltaic (PV) power generation, the fitness calculation incorporates adjustments based on solar irradiance uncertainty data. The resulting parameter combination maintains power output stability and continuity even under significant differences in grid electricity demand between peak and off-peak periods. This technical solution ensures that the generated initial power parameter combination meets both grid stability requirements and the efficient utilization of renewable energy, thus providing a high-quality global parameter foundation for subsequent localized adjustments at the edge.

[0051] The local adjustment unit is used to acquire characteristic data of local devices at the edge based on the global optimization parameter set, and to perform localized adjustments to the global optimization parameter set based on the characteristics of the local devices to determine a refined power setting value; specifically, it includes:

[0052] Acquire local device characteristic data at the edge, wherein the local device characteristic data includes response time constant and gain coefficient; adjust the global optimization parameter set locally based on the local device characteristic data to generate an adjusted parameter set; calculate the power allocation adaptability value for each parameter in the adjusted parameter set, wherein the power allocation adaptability value is based on device response speed and power stability; determine a refined power setting value based on the power allocation adaptability value.

[0053] Specifically, in one embodiment, to obtain a more refined power setting value, characteristic data of the local device is first collected at the edge via sensors. This local device characteristic data includes a response time constant and a gain coefficient. The response time constant characterizes the time interval required for the device to reach a steady-state output power after receiving a control command, and the gain coefficient characterizes the amplification factor of the input power command signal. The initial power parameters corresponding to the global optimization parameter set are then processed with the response time constant to obtain a preliminary adjustment value, thereby ensuring that the global parameters are compatible with the dynamic response speed of the local device. Then, based on… The aforementioned gain coefficients are used to weight and correct the initial adjustment values, thereby generating an adjusted parameter set that reflects the true response characteristics of the local device. To ensure the applicability of the adjusted parameter set in multi-device scenarios, power allocation adaptability calculations are required for each parameter. This calculation process includes obtaining the device's response speed by taking the reciprocal of the response time constant to quantify the device's power variation capability per unit time. Then, by performing variance calculations on the sequence of the adjusted parameters within a preset time window, the power stability index is obtained. This index measures the stability of the device's output over a continuous period. Subsequently, the response speed and the power stability index are compared. Multiplying these values ​​yields a power allocation adaptability value, which mathematically reflects a comprehensive trade-off between rapid response and stable output. A larger value indicates that the parameter better matches the operating characteristics of the target equipment; conversely, a smaller value indicates that the parameter is less compatible with the operating characteristics of the target equipment. In wind farm mode switching scenarios, even under conditions of significant wind speed variations and rapid fluctuations in load demand, wind turbines can achieve a stable response based on adjusted parameters, ensuring continuous and stable grid power. On the other hand, in uncertain scenarios of photovoltaic power generation, the response time constant of the equipment is also collected and combined with a continuous power output sequence to calculate... The degree of fluctuation of the parameters within a certain time window forms a power stability index, which is combined with the response speed obtained from the response time constant to calculate the power allocation adaptability value. This adaptability value logically reflects the ability of photovoltaic equipment to adapt to voltage changes under uncertain light intensity conditions. When this value is high, it indicates that the selected parameters can maintain good stability under voltage disturbance conditions, thereby improving the robustness of power allocation. Through the above applications in wind power and photovoltaic scenarios, it can be seen that this invention can not only achieve dynamic matching between global parameters and local characteristics, but also maintain the stable operation of the power system in scenarios with significant fluctuations in new energy sources.

[0054] The above technical solution enables the collection of characteristic data from wind power and photovoltaic equipment to form corresponding adjusted parameter sets in multi-source coupled new energy applications. Then, the power distribution adaptability value of each is calculated, and the parameter with the highest adaptability value is selected as the refined power setting value. This allows different types of equipment to achieve matched operation in multi-source uncertain environments, thereby ensuring stable and controllable power distribution of the power grid under cloud-edge-device collaboration.

[0055] The timing coordination unit is used to generate a timing coordination scheme by simulating the timing sequence under the multi-device collaborative scenario through reinforcement learning if the deviation between the refined power setting value and the current state of the terminal device exceeds a preset threshold; extract the synchronization gain coefficient and response time constant according to the timing coordination scheme; reconfigure parameters for possible sudden changes during mode switching; and obtain a coordinated and consistent device response sequence.

[0056] The components used to generate the timing coordination scheme include:

[0057] The process involves: acquiring current state data of the terminal device; calculating the deviation between the refined power setting value and the current state data of the terminal device; if the deviation exceeds a preset threshold, initializing a reinforcement learning model, wherein the reinforcement learning model includes a state space, an action space, and a reward function; using the refined power setting value and the current state data of the terminal device as inputs to the state space; generating an action sequence for a multi-device collaborative scenario using the reinforcement learning model, wherein the action sequence represents the power adjustment actions of each device; optimizing the action sequence according to the reward function to generate a time-series sequence; and generating a time-series coordination scheme based on the time-series sequence, wherein the time-series coordination scheme includes the power adjustment time points and adjustment magnitudes of each device.

[0058] Specifically, in one embodiment, when generating a timing coordination scheme, real-time operating status data of terminal devices is first acquired through edge devices. This status data includes parameters such as power output level and response characteristics. The data acquisition process is conducted through a distributed sensor network, ensuring the timeliness and accuracy of the acquired information. The real-time status data is then compared with a previously determined refined power setting value, and the deviation between the two is calculated. If the deviation exceeds a preset threshold, the generation of the timing coordination scheme is triggered. During this process, a reinforcement learning algorithm is introduced to simulate the operating sequence in a multi-device collaborative scenario. This reinforcement learning algorithm constructs a state space, action space, and reward function based on a Markov decision process. The state space describes the operating status of different devices under the current load and power output. The action space corresponds to the power adjustment strategies that each device may execute, and the reward function is designed to improve the overall stability of the system while reducing deviation. During the algorithm execution, the agent generates candidate action sequences through iterative exploration and utilization. The initial sequence may contain the power adjustment behaviors of multiple devices at different times. The sequence is optimized based on the reward feedback until it converges to a high-reward path that can reduce the overall deviation and ensure frequency and voltage stability. The resulting action sequence not only considers the response characteristics of a single device but also incorporates the cooperative relationship between devices, thus effectively handling the complexity brought about by multi-source fluctuations under the condition of new energy access. Furthermore, based on the optimized action sequence, the power adjustment amplitude that each device should execute at a specific time point is extracted to form a coordination scheme with temporal characteristics.

[0059] The above technical solution, through a timing coordination scheme, enables synchronous response among multiple devices, ensuring that the overall power distribution path remains consistent and stable in the face of sudden fluctuations, thereby guaranteeing the smooth switching and efficient operation of the power grid in complex operating scenarios.

[0060] Furthermore, the timing coordination unit is used to reconfigure parameters in response to sudden changes that may occur during mode switching, resulting in a coordinated and consistent device response sequence, including:

[0061] The synchronization gain coefficient and response time constant of each device are extracted from the timing coordination scheme; the parameter reconfiguration requirement value of each device is calculated for sudden changes during mode switching; the synchronization gain coefficient and response time constant are adjusted according to the parameter reconfiguration requirement value to generate a reconfigured parameter set; and a coordinated device response sequence is generated according to the reconfigured parameter set, wherein the device response sequence includes the power output timing of each device.

[0062] Specifically, in one embodiment, when reconfiguring parameters for sudden changes that may occur during mode switching, it is necessary to parse and obtain the synchronization gain coefficient and response time constant of each device from the aforementioned timing coordination scheme. The timing coordination scheme is generated by a reinforcement learning algorithm simulating a multi-device collaborative scenario, and contains coordination information of different devices under a specific timing sequence. By decomposing the scheme into data blocks corresponding to each device and reading the corresponding synchronization gain coefficient and response time constant from them, accurate initial input can be provided for subsequent parameter reconfiguration. If sudden situations such as real-time load fluctuations or unstable output of new energy occur during mode switching, operating data such as voltage and frequency are collected at the edge and such sudden data is input into the particle swarm optimization algorithm for processing. The aforementioned particle swarm optimization algorithm generates corresponding parameter reconfiguration requirement values ​​for each device through swarm intelligence iterative search in multiple rounds of updates. The aforementioned requirement values ​​essentially quantify the magnitude and priority of adjustments required for each device, ensuring that the system can provide differentiated correction suggestions for the operating status of different devices.

[0063] After obtaining the reconfiguration demand value, it is further mapped to an adjustment factor, and the extracted synchronization gain coefficient and response time constant are dynamically corrected to form a new reconfiguration parameter set. The above correction process enables the parameters to adapt to external fluctuations in real time and maintain the synchronization and response consistency between devices. Based on the reconfigured parameter set, a time-series framework is constructed to generate a coordinated device response sequence. The device response sequence shows the collaborative response relationship of multiple devices in mode switching in the form of power output changing over time. Simulation and verification ensure that the power output of each device can remain consistent at the set time points. Through the above technical solution, the generated device response sequence can not only cover the power adjustment demand under sudden conditions, but also realize the linkage optimization of multi-source data and multi-parameters in high-fluctuation scenarios such as photovoltaic and wind power, thereby effectively improving the stability and robustness of power grid mode switching.

[0064] A global verification unit is used to perform global verification based on the device response sequence, employing a particle swarm optimization algorithm to iteratively optimize unstable points in the response sequence and determine a stable power allocation path; specifically, it includes:

[0065] The device response sequence is globally verified in the cloud to obtain frequency simulation results and voltage simulation results; the particle swarm optimization algorithm is used to iteratively optimize the unstable points in the frequency simulation results and voltage simulation results; based on the results of the iterative optimization, a stable power allocation path is generated, wherein the stable power allocation path includes the power allocation ratio and timing arrangement of each device.

[0066] Specifically, in one embodiment, global verification is performed using the aforementioned device response sequence. A particle swarm optimization (PSO) algorithm is used to iteratively optimize unstable points in the response sequence to determine a stable power allocation path. Specifically, the cloud receives the device response sequence fed back from the edge device. This sequence contains power setpoints and coordination schemes for multiple devices at a specific time sequence. The cloud establishes numerical simulation models of frequency and voltage, performs global verification calculations on the sequence, and obtains frequency and voltage simulation results, thereby revealing the dynamic fluctuation characteristics that the power grid may exhibit under the aforementioned response conditions. Furthermore, points exceeding a preset stability threshold in the simulation results are identified and input as unstable points into the PSO algorithm for processing. The PSO algorithm uses iterative search based on swarm intelligence as its core mechanism; each particle corresponds to a candidate parameter adjustment scheme. The solution improves the performance of unstable points by fine-tuning the power allocation ratio or timing. During operation, the algorithm continuously updates the position and velocity of particles and dynamically corrects the search direction by combining individual optimality and swarm optimality feedback. By constructing a fitness function to evaluate the stability effect of each candidate solution, the system can gradually converge to the global optimal solution that minimizes fluctuations during iteration. The above optimization process is not only applicable to the handling of local voltage or frequency anomalies, but can also be extended to power grid scenarios of different scales, realizing adaptive adjustment from microgrids to regional power grids, thereby improving the robustness of the entire system under multi-device collaborative conditions. After completing the iterative optimization, the cloud generates a stable power allocation path based on the optimization results. This path clearly defines the proportional relationship of each device in the overall power allocation and arranges the timing of synchronous adjustment of each device.

[0067] Through the above technical solutions, the resulting power distribution path can maintain logical connection with the previous equipment response sequence, ensuring a smooth transition of the power grid mode switching under complex fluctuations, thereby significantly improving the stability of power grid operation and energy utilization efficiency in the scenario of new energy access.

[0068] The instruction correction unit is used to determine the adaptability of the stable power distribution path based on real-time monitoring data feedback. If local fluctuations exist, it generates a correction instruction and transmits the correction instruction to the terminal device. Specifically, it includes:

[0069] Real-time monitoring data is acquired at the edge, including equipment operating status and power grid fluctuation data; based on the real-time monitoring data, the local fluctuation value of the stable power distribution path is calculated; if the local fluctuation value exceeds a preset fluctuation threshold, a path adaptability judgment result is generated; based on the path adaptability judgment result, a correction instruction is generated, wherein the correction instruction includes adjustment parameters of the power distribution path.

[0070] Specifically, in one embodiment, the adaptability of a stable power distribution path is judged by real-time monitoring data feedback. When local fluctuations are detected, a correction command is generated and sent to the terminal device to ensure smooth switching of the power grid under complex operating scenarios. The edge device first acquires real-time monitoring data through distributed sensors. This real-time monitoring data includes not only the operating status parameters of various terminal devices but also overall power grid fluctuation information, such as frequency offset and voltage fluctuations. After being transmitted to the processing unit, the monitoring data is parsed in real time, and the local fluctuation value corresponding to the stable power distribution path is calculated. This fluctuation value is obtained through statistical analysis of the monitoring data sequence and is used to quantify the degree of deviation of the power path during actual operation. When the calculation result shows that the local fluctuation value exceeds a preset threshold, a path adaptability judgment result is generated, and the local location exceeding the threshold is marked as an unstable point, thereby identifying the part of the power distribution path that needs correction. A correction command is generated based on the judgment result. The instructions contain power adjustment parameters for different nodes or devices, and their content has been optimized by algorithms to ensure that the adjusted path can regain dynamic balance. After the above correction instructions are sent to the terminal devices via the cloud or edge, the terminal devices parse the power setpoints and perform parameter synchronous adjustments in the local control system. To further ensure the correction effect, frequency response data and voltage response data are tracked in real time during the execution of the correction instructions. The real-time tracked frequency response data is filtered to remove noise to obtain an accurate frequency sequence, and the voltage response data is corrected by fusing the response time constant of the device with the data collected by the voltage sensor to obtain a voltage curve that better matches the characteristics of the device. The above two types of data are integrated into a fused dataset and input into the particle swarm optimization algorithm for iterative calculation. The algorithm evaluates the overall deviation and gives a stability index by continuously updating the search path. If the index meets the stability condition, a new mode switching result is output, indicating that the power grid has maintained a stable operating state after the execution of the correction instructions.

[0071] The above-mentioned technical solution, through real-time monitoring, dynamic correction and cloud-edge-device collaborative optimization, can not only quickly respond to disturbances caused by fluctuations in new energy output or sudden load increases, but also achieve disturbance-free mode switching through the complementarity of multi-source data and the iterative convergence of algorithms, thereby significantly improving the reliability and energy utilization efficiency of the power grid in a highly uncertain environment.

[0072] The terminal equipment unit is used to perform synchronous adjustments according to the correction instructions, track changes in grid frequency and voltage in real time, and ultimately obtain a smooth mode switching result; specifically, it includes:

[0073] The terminal device acquires the correction instruction; adjusts the power output parameters of each device according to the correction instruction; tracks the frequency response data of each device in real time in response to changes in grid frequency; tracks the voltage response data of each device in real time in response to changes in grid voltage; and generates a stable mode switching result based on the frequency response data and the voltage response data, wherein the stable mode switching result includes the stable operating state of the grid.

[0074] Specifically, in one embodiment, a smooth mode switching result is ultimately obtained by performing synchronous adjustments according to the aforementioned correction instructions and combining them with real-time monitoring data for dynamic tracking. The terminal device first receives the correction instructions from the cloud or edge, which are generated by the optimized power allocation path and include the power setting values ​​that each device needs to adjust. After receiving the instructions, the terminal device parses the parameters and inputs the parsing results into the terminal control unit, thereby achieving synchronous adjustment of the power output of each device. During the adjustment process, the grid frequency changes are continuously monitored. Frequency data is collected by sensors deployed at each device, and the collected data is filtered to remove noise and smooth it, forming a continuous and accurate frequency response sequence. This process effectively eliminates high-frequency disturbances caused by the uncertainty of new energy output, thus more realistically reflecting the dynamic changes of the grid frequency. Simultaneously, the grid voltage is tracked in real time, with voltage sensors at the device end. After collecting fluctuation data, the system performs corrections based on the device's own response time constant and gain coefficient to generate a voltage response curve consistent with actual operating characteristics. In scenarios involving multiple devices working together, the voltage data is further corrected using time-series simulations to improve tracking accuracy and ensure coordination between different devices. The processed frequency and voltage response data are then merged into a unified dataset and input into a particle swarm optimization algorithm for iterative calculations. This algorithm identifies and optimizes unstable points in the merged data by simulating swarm intelligence to search for the optimal solution, thereby obtaining indicators characterizing the overall stability of the power grid. If these indicators meet preset stability conditions, a stable mode switching result is output. This switching result not only reflects a stable operating state where both the grid frequency and voltage remain within safe ranges but also demonstrates that in complex scenarios with high new energy penetration and rapid load fluctuations, the system can achieve seamless mode switching through a cloud-edge-device collaborative mechanism. Figure 2As shown, after the aforementioned terminal equipment corrects the power distribution path through correction commands, the local fluctuations are effectively suppressed, and the stability of the power grid is significantly improved, thereby ensuring the smooth operation of the system under different load and new energy fluctuation conditions. The above technical solution forms a closed-loop dynamic control process through synchronous adjustment driven by correction commands, dual real-time tracking of frequency and voltage, and iterative optimization of fused data, thereby significantly improving the stability of the power grid and energy utilization efficiency in complex environments.

[0075] This invention also provides a cloud-edge-device collaborative management and control method based on the power Internet of Things, which is based on the above-described system implementation, such as... Figure 3 As shown, the method includes:

[0076] Step S1: Obtain real-time load fluctuation and new energy power generation uncertainty data from the global data source in the cloud, process them through the particle swarm algorithm to generate multiple initial power parameter combinations, and select the optimal combination as the global optimization parameter set;

[0077] Step S2: Obtain the characteristic data of the local device at the edge based on the global optimization parameter set, and make localized adjustments to the global optimization parameter set based on the characteristics of the local device to determine the refined power setting value;

[0078] Step S3: If the deviation between the refined power setting value and the current state of the terminal device exceeds a preset threshold, a timing sequence is generated by simulating the timing sequence under the multi-device collaborative scenario through reinforcement learning; the synchronization gain coefficient and response time constant are extracted according to the timing coordination scheme, and parameters are reconfigured for possible sudden changes during mode switching to obtain a coordinated and consistent device response sequence.

[0079] Step S4: Perform global verification through the device response sequence, use particle swarm optimization algorithm to iteratively optimize unstable points in the response sequence, and determine a stable power allocation path; based on real-time monitoring data feedback, determine the adaptability of the stable power allocation path, and if there are local fluctuations, generate a correction instruction and transmit the correction instruction to the terminal device;

[0080] Step S5: The terminal device performs synchronous adjustment based on the correction command, tracks the changes in grid frequency and voltage in real time, and finally obtains a smooth mode switching result.

[0081] In summary, this invention acquires real-time load fluctuation and uncertainty data of new energy power generation from the cloud, generates multiple initial power parameter combinations using a particle swarm optimization algorithm, selects the optimal combination as the global optimization parameter set, provides basic data for subsequent local adjustments, and ensures the effectiveness of the optimization scheme under the actual conditions of grid load demand and new energy power generation. Based on the characteristic data of local equipment, the global optimization parameter set is locally adjusted to refine the power set value and adapt to the specific response capabilities of the equipment, such as the equipment's response time constant and gain coefficient. This local adjustment ensures the timeliness and stability of equipment response, reducing scheduling errors caused by local characteristic differences. During grid operation, if the deviation between the refined power setpoint and the current state of the terminal equipment exceeds a preset threshold, a reinforcement learning algorithm is used to simulate multi-device collaborative scenarios and generate a timing coordination scheme. This not only optimizes power adjustment between devices but also considers sudden changes that may occur during grid mode switching. By adjusting the synchronization gain coefficient and response time constant, it ensures that each device can respond consistently to sudden load or new energy fluctuations. The optimized device response sequence is processed through global verification, and the particle swarm optimization algorithm is used to iteratively optimize unstable points, ultimately forming a stable power allocation path to ensure system stability under different operating modes. When real-time monitoring data feedback shows local fluctuations in the power allocation path, correction instructions are generated based on this feedback and transmitted to the terminal equipment. The terminal equipment performs synchronization adjustment and tracks grid frequency and voltage changes in real time. Through precise data tracking and adjustment, it ensures that the grid can maintain stable operation under various disturbances. Through the cooperation of the above technical solutions, the grid can achieve smooth mode switching in the face of sudden fluctuations and uncertainties, significantly improving the robustness and energy utilization efficiency of the grid.

[0082] Although the present invention has been described in detail above with general descriptions and specific embodiments, modifications or improvements can be made to it, which will be obvious to those skilled in the art. Therefore, all such modifications or improvements made without departing from the spirit of the present invention fall within the scope of protection claimed by the present invention.

Claims

1. A cloud-edge-device collaborative management and control system based on the power Internet of Things, characterized in that, include: The data acquisition unit is used to acquire real-time load fluctuation and uncertainty data of new energy power generation from global data sources in the cloud, and process them through particle swarm optimization algorithm to generate multiple initial power parameter combinations, and select the optimal combination as the global optimization parameter set; The local adjustment unit is used to obtain the characteristic data of the local device at the edge end according to the global optimization parameter set, and to perform localized adjustment of the global optimization parameter set based on the characteristics of the local device to determine the refined power setting value. The timing coordination unit is used to generate a timing coordination scheme by simulating the timing sequence in a multi-device collaborative scenario through reinforcement learning if the deviation between the refined power setting value and the current state of the terminal device exceeds a preset threshold; extract the synchronization gain coefficient and response time constant according to the timing coordination scheme; reconfigure parameters for sudden changes that occur during mode switching; and obtain a coordinated and consistent device response sequence. The global verification unit is used to perform global verification through the device response sequence, and uses the particle swarm optimization algorithm to iteratively optimize the unstable points in the response sequence to determine a stable power allocation path. The instruction correction unit is used to determine the adaptability of the stable power distribution path based on real-time monitoring data feedback. If there are local fluctuations, it generates a correction instruction and transmits the correction instruction to the terminal device. The terminal control unit is used to perform synchronous adjustment according to the correction command, track the changes in grid frequency and voltage in real time, and finally obtain a smooth mode switching result.

2. The system as described in claim 1, characterized in that, The data acquisition unit is configured as follows: Real-time load fluctuation data and new energy power generation uncertainty data are obtained from a global data source via the cloud. A particle swarm optimization algorithm is used to process the real-time load fluctuation data and the new energy power generation uncertainty data to generate multiple initial power parameter combinations. For each of the multiple initial power parameter combinations, a global optimization objective function value is calculated, where the global optimization objective function value is based on load balance and power generation efficiency. Based on the global optimization objective function value, the optimal combination is selected from the multiple initial power parameter combinations to obtain a global optimization parameter set.

3. The system as described in claim 1, characterized in that, The local adjustment unit is configured as follows: Acquire local device characteristic data at the edge, wherein the local device characteristic data includes response time constant and gain coefficient; adjust the global optimization parameter set locally based on the local device characteristic data to generate an adjusted parameter set; calculate the power allocation adaptability value for each parameter in the adjusted parameter set, wherein the power allocation adaptability value is based on device response speed and power stability; determine a refined power setting value based on the power allocation adaptability value.

4. The system as described in claim 1, characterized in that, The timing coordination unit, used to generate timing coordination schemes, includes: The system acquires the current state data of the terminal device and calculates the deviation between the refined power setting value and the current state data of the terminal device. If the deviation exceeds a preset threshold, it initializes a reinforcement learning model, which includes a state space, an action space, and a reward function. The refined power setting value and the current state data of the terminal device are used as inputs to the state space. Through the reinforcement learning model, an action sequence in a multi-device collaborative scenario is generated, where the action sequence represents the power adjustment actions of each device. Based on the reward function, the action sequence is optimized to generate a time-series sequence. Based on the timing sequence, a timing coordination scheme is generated, wherein the timing coordination scheme includes the power adjustment time point and adjustment range of each device.

5. The system as described in claim 4, characterized in that, The timing coordination unit is used to reconfigure parameters in response to sudden changes that occur during mode switching, resulting in a coordinated and consistent device response sequence, including: The synchronization gain coefficient and response time constant of each device are extracted from the timing coordination scheme; the parameter reconfiguration requirement value of each device is calculated for sudden changes during mode switching; the synchronization gain coefficient and response time constant are adjusted according to the parameter reconfiguration requirement value to generate a reconfigured parameter set; and a coordinated device response sequence is generated according to the reconfigured parameter set, wherein the device response sequence includes the power output timing of each device.

6. The system as described in claim 1, characterized in that, A global verification unit, used to determine a stable power allocation path, includes: The device response sequence is globally verified in the cloud to obtain frequency simulation results and voltage simulation results; the particle swarm optimization algorithm is used to iteratively optimize the unstable points in the frequency simulation results and voltage simulation results; based on the results of the iterative optimization, a stable power allocation path is generated, wherein the stable power allocation path includes the power allocation ratio and timing arrangement of each device.

7. The system as described in claim 1, characterized in that, The instruction correction unit, used to generate correction instructions, includes: Real-time monitoring data is acquired at the edge, including equipment operating status and power grid fluctuation data; based on the real-time monitoring data, the local fluctuation value of the stable power distribution path is calculated; if the local fluctuation value exceeds a preset fluctuation threshold, a path adaptability judgment result is generated; based on the path adaptability judgment result, a correction instruction is generated, wherein the correction instruction includes adjustment parameters of the power distribution path.

8. The system as described in claim 1, characterized in that, The terminal control unit is configured as follows: Obtain the correction instruction; adjust the power output parameters of each device according to the correction instruction; track the frequency response data of each device in real time in response to changes in grid frequency; track the voltage response data of each device in real time in response to changes in grid voltage. Based on the frequency response data and the voltage response data, a stable mode switching result is generated, wherein the stable mode switching result includes the stable operating state of the power grid.

9. The system as described in claim 2, characterized in that, The data acquisition unit is used to generate multiple combinations of initial power parameters, including: Initialize the particle swarm, where each particle represents an initial power parameter combination; for each particle, calculate its fitness value, where the fitness value is based on load fluctuation smoothness and renewable energy generation utilization rate; Based on the fitness value, the position and velocity of the particles are updated to generate new power parameter combinations; the particle swarm is iteratively updated until a preset convergence condition is met to obtain multiple initial power parameter combinations.

10. A cloud-edge-device collaborative management and control method based on the power Internet of Things, implemented based on the system as described in any one of claims 1-9, characterized in that, The method includes: Step S1: Obtain real-time load fluctuation and new energy power generation uncertainty data from the global data source in the cloud, process them through the particle swarm algorithm to generate multiple initial power parameter combinations, and select the optimal combination as the global optimization parameter set; Step S2: Obtain the characteristic data of the local device at the edge based on the global optimization parameter set, and make localized adjustments to the global optimization parameter set based on the characteristics of the local device to determine the refined power setting value; Step S3: If the deviation between the refined power setting value and the current state of the terminal device exceeds a preset threshold, a timing coordination scheme is generated by simulating the timing sequence under the multi-device collaborative scenario through reinforcement learning; the synchronization gain coefficient and response time constant are extracted according to the timing coordination scheme, and parameters are reconfigured for sudden changes that occur during mode switching to obtain a coordinated and consistent device response sequence. Step S4: Perform global verification through the device response sequence, use particle swarm optimization algorithm to iteratively optimize unstable points in the response sequence, and determine a stable power allocation path; based on real-time monitoring data feedback, determine the adaptability of the stable power allocation path, and if there are local fluctuations, generate a correction instruction and transmit the correction instruction to the terminal device; Step S5: Perform synchronous adjustment according to the correction instruction, track the changes in grid frequency and voltage in real time, and finally obtain a smooth mode switching result.

Images