Multi-service-area optical storage direct flexible system electric energy scheduling method and device

Abstract

The invention relates to the technical field of energy storage scheduling data processing, and provides a multi-service-area optical storage direct flexible system electric energy scheduling method for solving the problem of insufficient collaborative optimization of optical storage direct flexible system electric energy scheduling. LSTM is combined with a random forest model to predict photovoltaic output and load demand, and a prediction database is established; based on the clean energy consumption rate and the total operation cost, constructing a multi-objective optimization model of electric energy scheduling; designing a three-order scheduling framework based on a day-ahead plan, intra-day optimization and real-time adjustment; based on mixed integer linear programming, distributed model predictive control and an alternating direction multiplier method, solving the multi-objective optimization model by using a three-order scheduling framework to obtain an electric energy scheduling scheme; and evaluating and iteratively optimizing the electric energy dispatching scheme. According to the invention, fine regulation and control of power dispatching optimization of the optical storage direct-flexible system are realized through multi-time-scale cooperation, and the economy and stability of the multi-service-area optical storage direct-flexible system are effectively improved.

Description

Technical Field

[0001] This invention belongs to the field of energy storage dispatch data processing technology, specifically relating to a power dispatch method for a multi-service area photovoltaic-storage-DC-flexible system. Background Technology

[0002] The energy structure is gradually shifting towards low-carbon and clean energy. The grid connection of numerous new energy sources and new loads has brought higher demands for operational coordination and optimized management to traditional power systems. "Photovoltaic-storage-DC-flexible" is a key technology integration and cutting-edge development direction for building new power systems and promoting the energy revolution. It is a systematic solution to the stability problems brought about by the high proportion of clean energy connected to the grid. Against this backdrop, service area flexible DC networks, as an autonomous small-scale power system integrating clean energy, distributed power sources, energy management systems, and local loads, are gradually becoming an important component for achieving green, efficient, and safe energy use.

[0003] The service area's flexible DC grid incorporates various energy sources, such as photovoltaics, diesel generators, fuel cells, and energy storage systems. These power sources differ in cost structure and exhibit strong intermittent generation characteristics. In actual operation, the dispatch system needs to cope with complex load fluctuations and uncertainties at the source, while simultaneously ensuring safe power supply and rationally coordinating the priority use of clean energy with the economical operation of the system.

[0004] Existing technologies have proposed various scheduling strategies, including model predictive control, robust optimization, and multi-objective energy management. These methods have achieved certain results in terms of stability or convergence efficiency, but some strategies focus on a single objective (such as prioritizing clean energy or minimizing costs), failing to achieve synergistic optimization among scheduling objectives, resulting in insufficient economic or environmental performance of the system.

[0005] Therefore, there is an urgent need for a scheduling optimization method that comprehensively considers the efficient utilization and economic operation of clean energy and adapts to the multi-source heterogeneous power grid structure, so as to promote the high-quality development of photovoltaic-storage-DC-flexible systems and provide scheduling guarantees for energy structure adjustment and low-carbon goals. Summary of the Invention

[0006] To address the aforementioned technical problems, this invention proposes a power dispatching method for a multi-service-area photovoltaic-storage-DC-flexible system. It employs an LSTM neural network to predict photovoltaic output and load demand, inputting the data into a constructed multi-objective optimization model. Through a three-level dispatching framework, combined with mixed-integer linear programming, distributed model predictive control, and alternating direction multiplier method, hierarchical distributed collaborative control of power dispatching for the photovoltaic-storage-DC-flexible system is achieved. This results in improved clean energy absorption rate and optimized main grid power purchase costs.

[0007] A power dispatching method for a multi-service-area photovoltaic-storage-DC-flexible system, the method comprising: Step 110: Collect basic data of flexible DC networks in multiple service areas of the photovoltaic-storage-DC-flexible system, including at least the following parameters: distributed power generation equipment parameters, energy storage equipment parameters, historical load data, and time-of-use electricity price table of the main grid; Step 120: Using the aforementioned basic data, a long short-term memory network combined with a random forest model is used to predict the photovoltaic output and load demand of the photovoltaic-storage-direct-drive-flexible system within a set time period in the future, and the predicted data of photovoltaic output and load demand are stored to establish a prediction database. Step 130: Based on the clean energy consumption rate and total operating cost of the photovoltaic-storage-direct-drive-flexible system, construct a multi-objective optimization model for power dispatch of the photovoltaic-storage-direct-drive-flexible system. Step 140: Design a three-order scheduling framework based on daily planning, intraday optimization, and real-time adjustment; the three-order scheduling framework shall at least include a time coupling mechanism for daily planning, intraday optimization, and real-time adjustment. Step 150: Using basic data and the prediction database as input, solve the multi-objective optimization model using a third-order scheduling framework to obtain a power dispatch scheme, which includes at least: formulating a day-ahead plan, using mixed-integer linear programming to obtain the main grid power purchase plan and power dispatch scheme within the day-ahead plan period; performing intraday optimization, based on distributed model predictive control and alternating direction multiplier method, to achieve layer-based autonomous optimization and cross-layer cluster coordination of the service area flexible DC network; and performing real-time adjustment and feedback. Step 160: Evaluate and iteratively optimize the power dispatch scheme.

[0008] Compared with the prior art, the technical effects of the present invention include: (1) The power dispatching method for multi-service area photovoltaic-storage-DC-flexible systems proposed in this invention constructs a three-level dispatching framework of "day-ahead forecasting - intraday rolling - real-time adjustment," achieving refined regulation through multi-timescale collaboration. Unlike the traditional single-timescale dispatching mode, this framework can better cope with the output fluctuations of photovoltaics, formulating basic plans at the day-ahead stage, dynamically optimizing hourly during the day, and correcting deviations in real time every 15 minutes, ensuring that the clean energy consumption rate is increased to over 90%, and significantly reducing dependence on the main power grid.

[0009] (2) This invention takes the priority use of clean energy as its core objective and achieves global and local synergy through a hierarchical distributed optimization architecture. The flexible DC grid layer of the sub-service area autonomously optimizes local resources based on distributed model predictive control. The cluster control center coordinates cross-grid power mutual assistance through the alternating direction multiplier method, which not only protects the information privacy of the flexible DC grid in each service area, but also reduces the main grid's electricity purchase by more than 30% through power surplus and deficit mutual assistance within the cluster, thus taking into account both environmental protection goals and system autonomy. Attached Figure Description

[0010] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on the structures shown in these drawings without creative effort.

[0011] Figure 1 This is a schematic diagram illustrating the steps of a power dispatching method for a multi-service area photovoltaic-storage-DC-flexible system in one embodiment of the present invention; Figure 2 This is a schematic diagram of the service area flexible DC network topology of a multi-service area photovoltaic-storage-DC-flexible system in one embodiment of the present invention. Detailed Implementation

[0012] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only a part of the embodiments of the present invention, and not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those of ordinary skill in the art without creative effort are within the scope of protection of the present invention.

[0013] In one embodiment, the present invention provides a power dispatching method for a multi-service-area photovoltaic-storage-DC-flexible system, such as... Figure 1 As shown, the method includes: Step 110: Collect basic data of flexible DC networks in multiple service areas of the photovoltaic-storage-DC-flexible system, including at least the following parameters: distributed power generation equipment parameters, energy storage equipment parameters, historical load data, and time-of-use electricity price table of the main grid; Step 120: Using the aforementioned basic data, a long short-term memory network combined with a random forest model is used to predict the photovoltaic output and load demand of the photovoltaic-storage-direct-drive-flexible system within a set time period in the future, and the predicted data of photovoltaic output and load demand are stored to establish a prediction database. Step 130: Based on the clean energy consumption rate and total operating cost of the photovoltaic-storage-direct-drive-flexible system, construct a multi-objective optimization model for power dispatch of the photovoltaic-storage-direct-drive-flexible system. Step 140: Design a three-order scheduling framework based on daily planning, intraday optimization, and real-time adjustment; the three-order scheduling framework shall at least include a time coupling mechanism for daily planning, intraday optimization, and real-time adjustment. Step 150: Using basic data and the prediction database as input, solve the multi-objective optimization model using a third-order scheduling framework to obtain a power dispatch scheme, which includes at least: formulating a day-ahead plan, using mixed-integer linear programming to obtain the main grid power purchase plan and power dispatch scheme within the day-ahead plan period; performing intraday optimization, based on distributed model predictive control and alternating direction multiplier method, to achieve layer-based autonomous optimization and cross-layer cluster coordination of the service area flexible DC network; and performing real-time adjustment and feedback. Step 160: Evaluate and iteratively optimize the power dispatch scheme.

[0014] The photovoltaic-storage-DC-flexible system for power dispatching proposed in this invention comprises multiple service area flexible DC networks with identical topologies. The topology of the photovoltaic-storage-DC-flexible system composed of multiple service area flexible DC networks is as follows: Figure 2 As shown, it includes a three-layer core structure: The underlying structure is a basic unit structure composed of a flexible DC grid in the service area. Each unit is a repeating module, containing power generation / energy storage equipment (photovoltaic / energy storage system / controllable power source) and local load; such as Figure 2 As shown, all A cluster consisting of units contains corresponding Flexible DC network for each service area: ; The intermediate layer of local controllers monitors and manages the energy flow of individual units, performs layer-based autonomous optimization of the local model, including maximizing clean energy utilization and minimizing operating costs; each local controller independently manages a flexible DC grid unit. The top-level cluster coordination controller is used to achieve cross-layer cluster coordination, including collecting actual operating data of each unit, coordinating power exchange between multiple units, handling the interaction strategy between the cluster and the main network, and achieving cluster optimization goals.

[0015] Specifically, in step 110, the basic data is collected at preset time intervals. The duration of data collection Divided into multiple consecutive time periods, using An index representing a time period; The total number of time periods is given by the total number of time periods. The duration of a time period is given by the total number of time .

[0016] In one embodiment, the duration of data acquisition The total number of days divided into time periods , .

[0017] In one embodiment, the parameters of the distributed power supply device include at least the photovoltaic panel efficiency.

[0018] In one embodiment, the parameters of the distributed power supply equipment include at least: photovoltaic capacity, energy storage capacity, energy storage charging / discharging efficiency, rated power of the controllable power supply, and base load.

[0019] Furthermore, in step 120, an ensemble learning method combining a Long Short-Term Memory (LSTM) network and a Random Forest (RF) model is employed to improve the prediction accuracy of photovoltaic power output (or clean energy output) and load demand, providing reliable data support for scheduling decisions. Specifically, this includes: Step 121: Using the basic data, employ both Long Short-Term Memory (LSTM) and Random Forest models as basic predictors to predict future times. Duration of each time period The photovoltaic output and load demand within the region.

[0020] use This represents the predicted value of photovoltaic power output or load demand obtained from LSTM prediction, expressed in terms of... This represents the predicted value of photovoltaic power output or load demand obtained from the random forest model. Among them, It is a time variable, representing the current time period.

[0021] Step 122: Based on the prediction error performance of the LSTM and Random Forest models, assign weighted weights and construct an ensemble prediction model: ; in, An integrated forecast value representing photovoltaic power output or load demand; and They are and The weighted weights, and satisfying .

[0022] Step 123 introduces an error compensation mechanism based on recent predictions. When the prediction error continues to exceed a preset error value, a dynamic correction process is automatically activated to calculate the prediction compensation amount and adjust the integrated prediction values. Dynamic calibration is performed to stabilize the prediction accuracy within the error control range given by the preset error value.

[0023] The predicted compensation amount is given by the following formula: ; in, For compensation coefficient, Forgetting factor, The first time before the current time period The prediction error for each time period.

[0024] Step 124: Optimize the ensemble prediction model by dynamically adjusting the weighting weights. The dynamic adjustment formula for the weighting weights includes: ; in, It is a regulatory factor. , The mean absolute errors of the LSTM and Random Forest models are respectively used to tilt the weights toward the ensemble prediction model with smaller errors.

[0025] In one embodiment, since the prediction error of LSTM is smaller than that of the random forest model, [the following is taken]. For example, during the historical 8:00 AM time period, the photovoltaic power generation of the MG1 flexible DC grid in the service area was predicted to be 160kW, while the actual power generation was 163kW, with an error of only 2%.

[0026] In one embodiment, the trigger condition for the error compensation mechanism is set as follows: when the system detects that the prediction error exceeds [a certain threshold] for three consecutive time periods. When this occurs, the error compensation mechanism is automatically activated. The error control range is... The preset error value is .

[0027] In one embodiment, .

[0028] Furthermore, in step 130, based on the clean energy absorption rate and total operating cost of the photovoltaic-storage-DC-flexible system, a multi-objective optimization model for power dispatch of the photovoltaic-storage-DC-flexible system is constructed, including: Step 131: Using the Analytic Hierarchy Process (AHP), determine the priority weights of clean energy consumption rate and total operating cost in the objective function of the multi-objective optimization model.

[0029] Based on the analytic hierarchy process (AHP), a judgment matrix is ​​constructed for two indicators: clean energy utilization rate and total operating cost. , It is a relative importance scale obtained using the 1-9 scale method in the Analytic Hierarchy Process (AHP). ;when hour, , ; .

[0030] After normalizing the columns of the judgment matrix, take the row average to obtain the priority weight vector: ; in, It is the priority weight of clean energy consumption rate. It is the priority weight of total operating cost. Represents the transpose of a vector / matrix.

[0031] Step 132: Based on the dual system of elastic optimization and rigid constraints, construct a multi-objective optimization model: Using priority weights, the objective function of the multi-objective optimization model is designed as follows: ; in, It is the clean energy consumption rate. The total operating cost is given by the following formulas: ; ; in, , and They are Actual clean energy consumption rate during the period Clean energy available power and during the period The operating cost of the time period; under the current scheduling objective, by converting the reciprocal to unify the direction and assigning weights to force priority, the above unnormalized objective function can effectively optimize the multi-objective optimization model. Compared with the normalized objective function, it is simpler to implement and has a smaller computational load.

[0032] Step 133: Take the maximum value of the objective function and construct a multi-objective optimization model for power dispatching of the photovoltaic-storage-DC-flexible system: .

[0033] In one embodiment, take Prioritizing the utilization rate of clean energy, at this point... , At the same time, economic efficiency is also taken into account. For example, during the 8:00 period, the local clean energy consumption rate of the flexible DC grid in each service area is ≤100% (no curtailment of solar power), and the total clean energy consumption rate at the cluster level is 98% through mutual assistance (total actual consumption of the cluster is 642kW / total available capacity of the cluster is 655kW).

[0034] Step 134, the multi-objective optimization model also includes constraints: power balance constraints, equipment operation constraints, and network transmission constraints.

[0035] In one embodiment, the power balance constraint includes: Internal power balance constraints of the service area flexible DC network: ; in, yes Actual output of photovoltaic arrays within the flexible DC grid in the service area during the specified time period. yes Output of controllable power sources within the flexible DC grid in the service area during specific time periods. yes Discharge power of the time-limited energy storage system yes The mutual power received from the flexible DC network outside the current service area during the time period. yes Local load demand of flexible DC network in the service area during the time period. yes The charging power of the time-of-use energy storage system yes The mutual power supplied to the flexible DC network outside the current service area during the specified time period. The flexible DC network outside the current service area refers to all other flexible DC networks in the photovoltaic-storage-DC-flexible system besides the flexible DC network in the current service area.

[0036] Overall power balance constraints of flexible DC networks in multiple service areas: ; in, For the first The output of the photovoltaic array within the flexible DC grid of each service area. For the first The output of controllable power sources within the flexible DC grid of each service area. ; This refers to the total number of flexible DC power grids in the service area; yes Power purchased from the main grid during the specified time period It is the first Local load of the flexible DC network in each service area yes Power sold to the main grid during a given period It is the first Energy storage charging power of the flexible DC grid in each service area It is the first Energy storage discharge power of the flexible DC grid in each service area.

[0037] Table 1. Parameters of Energy Storage Equipment in the Service Area Flexible DC Grid

[0038] In one embodiment, based on the distributed power device parameter settings shown in Table 1, the device operating constraints include: Energy storage constraints: And satisfy: ; in, It is a function of energy storage capacity. yes The minimum value, yes The maximum value, It is the rated capacity of energy storage; that is, the change in capacity during a single charge and discharge cycle does not exceed [a certain value]. ; The controllable power supply must meet the ramp rate constraint: ; in, It is the upper limit of the gradient rate. yes Controllable power output during specific time periods It refers to the duration of the time period.

[0039] In one embodiment, the duration of data acquisition The total number of days divided into time periods , Future Time sky; (That is, no more than 5% of the rated power per minute). This is to prevent the equipment from starting and stopping frequently.

[0040] In one embodiment, the network transmission constraints satisfy: Main grid interaction must follow: This is used to constrain and limit the fluctuation of purchased electricity in adjacent hours to no more than 20%, in order to avoid impacting the power grid. yes Power purchased by the main power grid during the specified time period.

[0041] In one embodiment, the rated energy storage capacity is 300 kWh for MG1, and the maximum single charge / discharge capacity is 180 kWh. The controllable power supply is: MG1's 100kW micro gas turbine; Minimum continuous running time ; Fluctuations in electricity purchases by the main grid in adjacent hours Through coordination, the electricity purchase volume for the 8:00 time period was [amount missing]. No restrictions were violated.

[0042] In one embodiment, , .

[0043] Specifically, in step 140, a time-coupling mechanism for daily planning, intraday optimization, and real-time adjustment is designed, including: The current plan is to meet the modified constraints: ; in, yes Planned power for the day of the period yes Power adjusted in real time for a given time period; The above formula shows that, based on the day-ahead plan (24 hours), the intraday optimization adjusts the magnitude of the change. To ensure the stability of the plan; Real-time adjustments must meet the following constraints: ; in, yes Planned power for the time period yes Real-time power output in 15-minute intervals within a given time period; The above constraints indicate that, Real-time adjustments (every 15 minutes) are made to the hourly plan correction range. This enables precise regulation.

[0044] The following nested time-scale functions are used to achieve seamless transitions between different time periods: ; in, For correction factor, This is the planned power for the next time period. This is the planned power for the current time period. This refers to the predicted power for the next time period. The nested time-scale function reflects a decreasing adjustment intensity at different stages, adapting to the prediction accuracy, optimization objectives, and system stability requirements at different stages. For example, data shows a sudden surge in load demand during the 8:15 time period of the day. The correction is achieved through energy storage fine-tuning.

[0045] In one embodiment, the correction coefficient is used for daily planning, intraday optimization, and real-time adjustment. Take values ​​of 0.8, 0.5, and 0.2 respectively.

[0046] Furthermore, in step 150, the basic data and prediction database are used as input, and a three-order scheduling framework is used to solve the multi-objective optimization model to obtain a power dispatching scheme. Specifically, this includes: Step 151: Formulate the day-ahead plan. Use mixed-integer linear programming to obtain the main grid power purchase plan and power dispatch scheme for the day-ahead planning period, including: (1) Each service area's flexible DC network reports the next day's basic data to the cluster control center (CCC), including photovoltaic output forecast curves, load demand forecast curves, initial values ​​of energy storage capacity functions, and the availability status of controllable power sources.

[0047] (2) Global optimization objective of CCC based on multi-objective optimization model The mixed integer linear programming (MILP) method is used to allocate the clean energy consumption quota of the flexible DC grid in each service area, and a time-by-time main grid power purchase plan (prioritizing power purchase during off-peak hours) and energy storage charging and discharging scheme are formulated.

[0048] The constraints for global optimization are extended to constraints related to electricity purchase periods: ; Requires off-peak hours ( The proportion of electricity purchased should be at least 40% in order to reduce costs by taking advantage of electricity price differences.

[0049] In one embodiment, the allocation of electricity purchase periods is optimized by utilizing purchase period constraints: the proportion of electricity purchased during off-peak hours (0:00-6:00 and 22:00-24:00, electricity price 0.3 yuan / kWh) is required to be ≥40%, with a total daily purchase volume of 280kWh and off-peak purchases of 120kWh (42.9%), thereby reducing costs by leveraging the price difference. The time-of-use electricity prices of the main grid used are shown in Table 2.

[0050] Table 2. Time-of-use electricity prices on the main power grid

[0051] Set standby capacity constraints: ; in, It is the spare capacity. It is a function of energy storage capacity. yes The minimum value, yes The maximum value; ensure that the energy storage always retains 20% of its rated capacity as an emergency backup. For example, the MG3 energy storage has a capacity of 200kWh, and 40kWh needs to be reserved as a backup. Its minimum SOC of 47% (corresponding to 94kWh) meets the requirements.

[0052] In addition, through clean energy quotas Mandatory disposal, in the formula For the first Clean energy installed capacity of the flexible DC grid in each service area; To meet the total projected output of the power cluster, each service area's flexible DC grid is required to absorb at least 90% of the clean energy. For example, the total projected output for the 8:00 AM period is 630kW, with an MG1 quota of 212kW, and the actual absorption is 255kW, exceeding the quota.

[0053] (3) The feedback equipment of the flexible DC network in each service area is limited, such as the energy storage capacity (SOC) is too low to discharge. After the CCC is optimized twice, the final day-ahead plan is formed, and the upper limit of clean energy consumption, energy storage operation range and controllable power supply start-up threshold are clearly defined for each time period.

[0054] Step 152 involves intraday optimization, based on distributed model predictive control and alternating direction multiplier method, to achieve layer-based autonomous optimization and cross-layer cluster coordination of the service area flexible DC network, including: (1) Real-time data updates, in the first... Time period (e.g.) (Current time period) Collect data from the previous time period Based on actual operating data, adjust for the next two time periods. The model parameter prediction values; the actual operating data includes photovoltaic output, load demand, energy storage capacity, upper limit of effective charging and discharging power of energy storage, and operating cost; the model parameter prediction values ​​include the predicted values ​​of photovoltaic output, load demand, energy storage capacity, upper limit of effective charging and discharging power of energy storage, and operating cost; (2) Autonomous optimization of the service area flexible DC grid layer using the local controller, including: Based on distributed model predictive control (DMPC), with the objective of "maximizing the utilization rate of local clean energy", decision-making is performed. Controlled output during the period: If the actual output of clean energy (i.e., photovoltaic output) > load demand: Clean energy will be consumed first, and the remaining electricity will be charged to energy storage first. If there is still surplus, electricity will be sold to the cluster; if the actual output of clean energy is less than the load: energy will be discharged and stored first. If the power supply is insufficient, start a controllable power source (prioritizing low-cost micro gas turbines) and finally purchase power from the main grid.

[0055] In one embodiment, when the deviation between clean energy and load is ≤ ±5%, balancing is achieved through energy storage fine-tuning (e.g., MG1 load 277kW, fine-tuning threshold 13.8kW); when the deviation is > ±15% (e.g., MG3 load 168kW, threshold 25.2kW), cluster support is triggered. Simultaneously, through the formula... This guides flexible loads, potentially increasing electricity consumption by 3kW during peak hours; among which, Based on the flexible load power, For the response coefficient, The difference between the real-time electricity price and the benchmark electricity price is used to achieve price-guided load regulation.

[0056] Table 3. Examples of data for different time periods of the day

[0057] The decision-making logic for autonomous optimization of the service area's flexible DC grid layer also includes flexible load response: ; In the formula, Based on the flexible load power, For the response coefficient, The difference between the real-time electricity price and the benchmark electricity price is used to achieve price-guided load regulation. yes The total clean energy output of the flexible DC grid in the service area during the specified time period. yes The actual load power of the service area flexible DC network during the time period yes The actual adjustment power of the energy storage system during a given time period. It is the rated charging and discharging power of the energy storage system; When power support threshold When the conditions are met, i.e., the deviation between clean energy and load exceeds 15%, the cluster power support mechanism will be triggered; when the deviation is within 5%, an energy storage fine-tuning strategy will be adopted. Fine-tuning is performed with a power level not exceeding 5% of the load.

[0058] (3) Using the cluster coordination controller for cross-layer cluster coordination, including: The cross-layer cluster coordination employs the Alternating Directional Multiplier Method (ADMM) to handle power mutual assistance across service area flexible DC networks. It guides power surplus / deficit balance through correction coefficients, reducing dependence on the main grid and ensuring power conservation within the cluster. Interactive power ).

[0059] The iterative convergence mechanism for cross-layer cluster coordination employs an adaptive step size: ; In the formula, This is the penalty factor for the ADMM algorithm, initially set to 1. It is the first The penalty factor for the next iteration is that when the rate of change of power imbalance between two consecutive iterations is less than 5%, the step size is halved to accelerate convergence; its convergence condition is: ; When the cluster power imbalance is less than When convergence is reached, it is considered to have been achieved; if convergence is not achieved after more than 10 iterations, it is considered to have been terminated. and If so, the current optimal solution will be adopted and emergency coordination will be triggered.

[0060] In one embodiment, the ADMM algorithm has an initial step size of 0.1, which is halved when the rate of change of power imbalance is <5%, and the convergence condition is... For example, during the 8:00 time period, MG2 has a surplus of 26kW of power, while MG3 is short of 25kW of power. After three iterations, mutual assistance is achieved, and the imbalance of -1kW satisfies the convergence requirement.

[0061] Step 153 involves real-time adjustments and feedback, including: (1) Real-time monitoring and changing the data collection time interval Collect actual operational data (e.g., every 15 minutes), compare it with the hourly plan, and calculate the deviation value; (2) Deviation handling, if the deviation value The energy storage system rapidly charges and discharges, suppressing the energy level; if the deviation value... This could trigger emergency adjustments (e.g., briefly activating backup power or purchasing temporary power). An emergency adjustment priority strategy is established to address sudden power fluctuations and ensure system stability. This emergency adjustment priority strategy is implemented through the following power gap allocation: ; In the above formula, The flexible DC network for adjacent service areas can support power (e.g., transmission loss). ), The available power for energy storage (e.g., SOC > 30%). This is the maximum power of the backup power supply. The maximum power purchase capacity of the main power grid; Obtain the emergency response sequence representing the order of power allocation: The above formula indicates that power is first drawn from the flexible DC grid of adjacent service areas, followed by energy storage, then backup power (limited to 2 hours), and finally purchased power; at the same time, response time constraints are required. To ensure system stability.

[0062] In one embodiment, the priority strategy for emergency adjustment is as follows: (1) calling up the redundant power of the flexible DC network in the adjacent service area (transmission loss ≤5%, such as MG2 transmitting 25kW to MG3 with a loss of 1kW); (2) energy storage for rapid discharge (SOC>30%, MG1 can use 26.4kW); (3) starting the backup power supply (continuous operation ≤2h, MG3 has a backup of 60kW); (4) purchasing power from the main grid.

[0063] For example, if the response time is required to be ≤2min, and the load suddenly increases by 13kW at 8:15, the energy storage adjustment should be completed within 1 minute.

[0064] (3) Record the actual operating data for each time period, update the energy storage capacity (SOC) function value, and use it as the scheduling input for the next time period.

[0065] In one embodiment, step 160 includes: Step 161: After the daily scheduling is completed, calculate the actual clean energy consumption rate and total operating cost, and compare them with the target value; Step 162: Based on the bias analysis, adjust the prediction model parameters (such as LSTM network weights) and optimize the algorithm coefficients (such as ADMM penalty factor). Step 163: Iterate the multi-objective optimization model parameters once a week to improve scheduling accuracy and economy.

[0066] The specific process of parameter iteration for the multi-objective optimization model first calculates the comprehensive evaluation index: ; In the formula For clean energy consumption rate, For actual cost, To achieve the target cost, the principle of prioritizing clean energy utilization rate is reflected; then, the model parameters are optimized using the following particle swarm optimization (PSO) algorithm update formula: ; in, For particle velocity, For inertial weights, As a learning factor, , It is a random number. For the optimal position of an individual, To find the globally optimal position; by limiting the LSTM learning rate. and ADMM penalty factor Within a reasonable range, ensure the stability of the algorithm.

[0067] In one embodiment, , At the same time, set parameter constraints. , Limiting the LSTM learning rate and ADMM penalty factor to a reasonable range ensures algorithm stability. , Yuan, Yuan, received After iteration , .

[0068] Through the above steps, the average clean energy consumption rate reached 95.3%, the main grid electricity purchase was reduced by 32%, and the total operating cost was reduced by 28%, achieving an optimized effect that balances environmental protection, economy, and stability.

[0069] In one embodiment, the objective function is normalized to obtain an objective function of the multi-objective optimization model in the following form: ; in and The maximum / minimum historical costs are used to eliminate the influence of dimensions and enhance the model's adaptability in different scenarios.

[0070] In one embodiment, the present invention also provides an apparatus for power dispatching in a multi-service-area photovoltaic-storage-DC-flexible system, the apparatus comprising: The first module is used to collect basic data of the flexible DC network in multiple service areas of the photovoltaic-storage-DC-flexible system, including at least the following parameters: distributed power generation equipment parameters, energy storage equipment parameters, historical load data, and the time-of-use electricity price table of the main grid; The second module is used to use the aforementioned basic data, employing a long short-term memory network combined with a random forest model, to predict the photovoltaic output and load demand of the photovoltaic-storage-direct-drive-flexible system within a set time period in the future, and to store the predicted data of photovoltaic output and load demand to establish a prediction database. The third module is used to construct a multi-objective optimization model for power dispatch of the photovoltaic-storage-direct-flexible system based on the clean energy consumption rate and total operating cost of the photovoltaic-storage-direct-flexible system. The fourth module is used to design a three-order scheduling framework based on daily planning, intraday optimization, and real-time adjustment; the three-order scheduling framework includes at least the design of a time coupling mechanism for daily planning, intraday optimization, and real-time adjustment. The fifth module takes basic data and a prediction database as input, and uses a third-order scheduling framework to solve a multi-objective optimization model to obtain a power dispatch scheme. This includes at least: developing a day-ahead plan, using mixed-integer linear programming to obtain the main grid power purchase plan and power dispatch scheme for the day-ahead planning period; performing intraday optimization, based on distributed model predictive control and the alternating direction multiplier method, to achieve layer-based autonomous optimization and cross-layer cluster coordination of the service area flexible DC network; and performing real-time adjustments and feedback. The sixth module is used to evaluate and iteratively optimize the power dispatch scheme.

[0071] On the other hand, the present invention provides a computer device including a memory and a processor. The memory stores a computer program, and the processor executes the computer program to implement the steps of the power dispatching method for multi-service area optical-storage-DC-flexible systems provided in any of the above embodiments. The computer device may be a server. The computer device includes a processor, a memory, a network interface, and a database connected via a system bus. The processor of the computer device provides computing and control capabilities. The memory of the computer device includes a non-volatile storage medium and internal memory. The non-volatile storage medium stores an operating system, a computer program, and a database. The internal memory provides an environment for the operation of the operating system and computer program in the non-volatile storage medium. The database of the computer device is used to store sample data. The network interface of the computer device is used for communication with external terminals via a network connection.

[0072] On the other hand, the present invention provides a computer-readable storage medium having a computer program stored thereon, wherein when the computer program is executed by a processor, it implements the steps of the power dispatching method for multi-service area optical-storage-direct-flexible systems provided in any of the above embodiments.

[0073] Those skilled in the art will understand that all or part of the processes in the methods of the above embodiments can be implemented by a computer program instructing related hardware. The computer program can be stored in a non-volatile computer-readable storage medium, and when executed, it can include the processes of the embodiments of the above methods. Any references to memory, storage, databases, or other media used in the embodiments provided in this application can include non-volatile and / or volatile memory. Non-volatile memory can include read-only memory (ROM), programmable ROM (PROM), electrically programmable ROM (EPROM), electrically erasable programmable ROM (EEPROM), or flash memory. Volatile memory can include random access memory (RAM) or external cache memory. By way of illustration and not limitation, RAM is available in various forms, such as static RAM (SRAM), dynamic RAM (DRAM), synchronous DRAM (SDRAM), dual data rate SDRAM (DDRSDRAM), enhanced SDRAM (ESDRAM), synchronous link DRAM (SLDRAM), Rambus direct RAM (RDRAM), direct memory bus dynamic RAM (DRDRAM), and memory bus dynamic RAM (RDRAM), etc.

[0074] Matters not covered in this invention are common knowledge.

[0075] The technical features of the above embodiments can be combined in any way. For the sake of brevity, not all possible combinations of the technical features in the above embodiments are described. However, as long as there is no contradiction in the combination of these technical features, they should be considered to be within the scope of this specification.

[0076] The embodiments described above are merely illustrative of several implementation methods of this application, and while the descriptions are specific and detailed, they should not be construed as limiting the scope of the invention. It should be noted that those skilled in the art can make various modifications and improvements without departing from the concept of this application, and these modifications and improvements all fall within the protection scope of this application.

[0077] The above description is merely a preferred embodiment of the present invention and is not intended to limit the invention. Various modifications and variations can be made to the present invention by those skilled in the art. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the scope of protection of the present invention.

Claims

1. A power dispatching method for a multi-service area photovoltaic-storage-DC-flexible system, characterized in that, The method includes: Step 110: Collect basic data of flexible DC networks in multiple service areas of the photovoltaic-storage-DC-flexible system, including at least the following parameters: distributed power generation equipment parameters, energy storage equipment parameters, historical load data, and time-of-use electricity price table of the main grid; Step 120: Using the aforementioned basic data, a long short-term memory network combined with a random forest model is used to predict the photovoltaic output and load demand of the photovoltaic-storage-direct-drive-flexible system within a set time period in the future, and the predicted data of photovoltaic output and load demand are stored to establish a prediction database. Step 130: Based on the clean energy consumption rate and total operating cost of the photovoltaic-storage-direct-drive-flexible system, construct a multi-objective optimization model for power dispatch of the photovoltaic-storage-direct-drive-flexible system. Step 140: Design a three-order scheduling framework based on daily planning, intraday optimization, and real-time adjustment; the three-order scheduling framework shall at least include a time coupling mechanism for daily planning, intraday optimization, and real-time adjustment. Step 150: Using basic data and the prediction database as input, solve the multi-objective optimization model using a third-order scheduling framework to obtain a power dispatch scheme, which includes at least: formulating a day-ahead plan, using mixed-integer linear programming to obtain the main grid power purchase plan and power dispatch scheme within the day-ahead plan period; performing intraday optimization, based on distributed model predictive control and alternating direction multiplier method, to achieve layer-based autonomous optimization and cross-layer cluster coordination of the service area flexible DC network; and performing real-time adjustment and feedback. Step 160: Evaluate and iteratively optimize the power dispatch scheme.

2. The power dispatching method for a multi-service area photovoltaic-storage-DC-flexible system according to claim 1, characterized in that, Step 110, the process of collecting basic data of the flexible DC network in multiple service areas of the photovoltaic-storage-DC-flexible system, includes: dividing the duration of data collection into multiple consecutive time periods according to a preset time interval, with a total number of time periods. The duration of a time period is the preset time interval.

3. The power dispatching method for a multi-service area photovoltaic-storage-DC-flexible system according to claim 2, characterized in that, Step 120 includes: Step 121: Using the basic data, employ both Long Short-Term Memory (LSTM) networks and Random Forest models as basic predictors to predict future time. Photovoltaic output and load demand within each time period; Step 122: Based on the error performance of predictions using Long Short-Term Memory Network and Random Forest model, assign weighted weights and construct an ensemble prediction model: ； in, This represents an integrated forecast value indicating photovoltaic power output or load demand. Indicates the current time period; This represents the predicted value of photovoltaic power output or load demand obtained from the prediction of the Long Short-Term Memory network. This represents the predicted value of photovoltaic power output or load demand obtained from the random forest model. and They are and The weighted weights, and satisfying ; Step 123 introduces an error compensation mechanism based on recent predictions. When the prediction error continues to exceed a preset error value, a dynamic correction process is automatically activated to calculate the prediction compensation amount and adjust the integrated prediction values. Dynamic calibration is performed to stabilize the prediction accuracy within the error control range given by the preset error value; The predicted compensation amount is given by the following formula: ； in, For compensation coefficient, Forgetting factor, The first time before the current time period Prediction error for each time period; Step 124: Optimize the ensemble prediction model by dynamically adjusting the weighting weights. The formula for dynamically adjusting the weighting weights is given by the following equation: ； in, It is a regulatory factor. , The average absolute errors of the Long Short-Term Memory Network and the Random Forest model are respectively used to tilt the weights towards the ensemble prediction model with smaller errors.

4. The power dispatching method for a multi-service area photovoltaic-storage-DC-flexible system according to claim 3, characterized in that, Step 130 includes: Step 131: Using the analytic hierarchy process (AHP), determine the priority weights of clean energy utilization rate and total operating cost in the objective function of the multi-objective optimization model: A judgment matrix is ​​constructed for two indicators: clean energy consumption rate and total operating cost. The elements of the judgment matrix are relative importance scales obtained by the 1-9 scale method. After normalizing the columns of the judgment matrix, take the row average to obtain the priority weight vector: ； in, It is the priority weight of clean energy consumption rate. It is the priority weight of total operating cost. Represents the transpose of a vector / matrix; Step 132: Based on the dual system of elastic optimization and rigid constraints, construct a multi-objective optimization model, including: Using priority weights, design the objective function of a multi-objective optimization model: ； in, It is the clean energy consumption rate. The total operating cost is given by the following formulas: ； ； in, yes The actual clean energy consumption rate during a given period yes Clean energy available power during the period yes Operating costs for a given period; Step 133: Take the maximum value of the objective function and construct a multi-objective optimization model for power dispatching of the photovoltaic-storage-DC-flexible system: ； Step 134, the multi-objective optimization model also includes constraints: power balance constraints, equipment operation constraints, and network transmission constraints.

5. The power dispatching method for a multi-service area photovoltaic-storage-DC-flexible system according to claim 4, characterized in that, In step 134, the power balance constraint includes: Internal power balance constraints of the service area flexible DC network: ； in, yes Actual output of photovoltaic arrays within the flexible DC grid in the service area during the specified time period. yes Output of controllable power sources within the flexible DC grid in the service area during specific time periods. yes Discharge power of the time-limited energy storage system yes The mutual assistance power received from the flexible direct-to-flexible network outside the current service area during the time period. yes Local load demand of flexible DC network in the service area during the time period. yes The charging power of the time-of-use energy storage system yes The mutual assistance power transmitted to the flexible direct-to-flexible network outside the current service area during the time period; Overall power balance constraints of flexible DC networks in multiple service areas: ； in, For the first The actual output of the photovoltaic array within the flexible DC grid of each service area. For the first The output of controllable power sources within the flexible DC grid of each service area. ; This refers to the total number of flexible DC power grids in the service area; yes Power purchased from the main grid during the specified time period It is the first Local load of the flexible DC network in each service area yes Power sold to the main grid during a given period It is the first Energy storage charging power of the flexible DC grid in each service area It is the first Energy storage discharge power of the flexible DC grid in each service area; The equipment operating constraints include: Energy storage constraints: And satisfy: ； in, It is a function of energy storage capacity. yes The minimum value, yes The maximum value, It is the rated capacity of energy storage; The controllable power supply must meet the ramp rate constraint: ； in, It is the upper limit of the gradient rate. yes Controllable power output during specific time periods It refers to the duration of the time period; The network transmission constraints satisfy: Main grid interaction must follow: ;in, yes Power purchased by the main power grid during the specified time period; Minimum continuous running time ; Fluctuations in electricity purchases by the main grid in adjacent hours .

6. The power dispatching method for a multi-service area photovoltaic-storage-DC-flexible system according to claim 4, characterized in that, In step 140, a time-coupling mechanism for daily planning, intraday optimization, and real-time adjustment is designed, including: The current plan is to meet the modified constraints: ； in, yes Planned power for the day of the period yes Power adjusted in real time for a given time period; Real-time adjustments must meet the following constraints: ； in, yes Planned power for the time period yes Real-time power output in 15-minute intervals within a given time period; Nested functions based on time scales: This ensures seamless transitions between different time periods; among them, For correction factor, This is the planned power for the next time period. This is the planned power for the current time period. This is the predicted power for the next time period.

7. The power dispatching method for a multi-service area photovoltaic-storage-DC-flexible system according to claim 6, characterized in that, In step 150, the step of formulating the day-ahead plan and using mixed-integer linear programming to obtain the main grid power purchase plan and power dispatch scheme for the day-ahead plan period includes: Each service area's flexible DC network reports the next day's basic data to the cluster control center, including photovoltaic output forecast curves, load demand forecast curves, initial values ​​of energy storage capacity functions, and the availability status of controllable power sources. The cluster control center's global optimization objective based on a multi-objective optimization model The clean energy consumption quota of the flexible DC grid in each service area is allocated by mixed integer linear programming, and a time-period main grid power purchase plan and energy storage charging and discharging scheme are formulated. The constraints for global optimization are extended to constraints related to electricity purchase periods: ； In order to reduce costs by taking advantage of electricity price differences, it is required that off-peak hours be used... The proportion of electricity purchased should reach at least 40%; To ensure that energy storage always retains 20% of its rated capacity as an emergency backup, a backup capacity constraint is set: ； in, It is the spare capacity. It is a function of energy storage capacity. yes The minimum value, yes The maximum value; Calculating clean energy consumption quotas: Energy storage charging power ； in, For the first Clean energy installed capacity of the flexible DC grid in each service area; To contribute to the total predicted power of the cluster, a coefficient of 0.9 indicates that each service area's flexible DC grid is required to absorb at least 90% of the clean energy. The feedback equipment of the flexible DC grid in each service area is limited, and the final day-ahead plan is formed after secondary optimization using the cluster control center, which gives the upper limit of clean energy consumption, energy storage operation range, and controllable power supply start-up threshold for each time period.

8. The power dispatching method for a multi-service area photovoltaic-storage-DC-flexible system according to claim 7, characterized in that, In step 150, the intraday optimization, based on distributed model predictive control and alternating direction multiplier method, realizes layer-based autonomous optimization and cross-layer cluster coordination of the service area flexible DC network, including: Real-time data updates, at the Time period to collect the previous time period Actual operating data, adjusted for the next two time periods. The model parameter prediction values; the actual operating data includes photovoltaic output, load demand, energy storage capacity, upper limit of effective charging and discharging power of energy storage, and operating cost; the model parameter prediction values ​​include the predicted values ​​of photovoltaic output, load demand, energy storage capacity, upper limit of effective charging and discharging power of energy storage, and operating cost; Autonomous optimization of the service area flexible DC grid layer using a local controller includes: decision-making based on distributed model predictive control, with the goal of maximizing local clean energy utilization. Controlled output during the time period: If the actual output of clean energy is greater than the load demand: Clean energy is consumed first, and the remaining electricity is charged to energy storage first. If there is still a surplus, electricity is sold to the cluster. If the actual output of clean energy is less than the load: Energy storage is discharged first, and if it is insufficient, controllable power sources are activated. Finally, electricity is purchased from the main grid. The autonomous optimization of the service area's flexible DC grid layer adopts a flexible load response: ； in, Based on the flexible load power, The response coefficient represents the number of iterations. The difference between the real-time electricity price and the benchmark electricity price is used to achieve price-guided load regulation. When the power support threshold is met When the deviation is within 5%, the cluster power support mechanism is triggered; when the deviation is within 5%, the energy storage fine-tuning strategy is adopted. Fine-tuning is performed with a power level not exceeding 5% of the load. yes The total clean energy output of the flexible DC grid in the service area at all times. yes The actual load power of the flexible DC network in the service area at any given time. yes The actual adjustment power of the energy storage system at all times. It is the rated charging and discharging power of the energy storage system; Cross-tier cluster coordination is performed using a cluster coordination controller, including: The alternating direction multiplier method is used to handle power mutual assistance between flexible DC networks across service areas. The power surplus and deficit balance is guided by the correction coefficient, reducing dependence on the main grid and ensuring power conservation within the cluster. The iterative convergence mechanism for cross-layer cluster coordination employs an adaptive step size: ； In the formula, This is the penalty factor for the ADMM algorithm, initially set to 1. It is the first The penalty factor for the next iteration; When the rate of change of power imbalance is less than 5% in two consecutive iterations, the step size is halved to accelerate convergence. The convergence condition is: ； If the iteration count exceeds 10 and convergence is still not achieved, If so, the current optimal solution will be adopted and emergency coordination will be triggered.

9. The power dispatching method for a multi-service area photovoltaic-storage-DC-flexible system according to claim 8, characterized in that, In step 150, the intraday optimization, based on distributed model predictive control and alternating direction multiplier method, to achieve layer-based autonomous optimization and cross-layer cluster coordination of the service area flexible DC network, further includes: Real-time monitoring can be performed by changing the data collection interval. Collect actual operating data and compare it with a 1-hour time interval to calculate the deviation value; Perform deviation processing; if the deviation value The energy storage system rapidly charges and discharges, suppressing the energy level; if the deviation value... This triggered an emergency adjustment; An emergency adjustment priority strategy is established to address sudden power fluctuations and ensure system stability. This emergency adjustment priority strategy is implemented through the following power gap allocation: ； in, The flexible DC network in adjacent service areas can support power. For energy storage available power, This is the maximum power of the backup power supply. The maximum power purchase capacity of the main power grid; Obtain the emergency response sequence representing the order of power allocation: ； Record the actual operating data for each time period, update the energy storage capacity function value, and use it as the scheduling input for the next time period.

10. The power dispatching method for a multi-service area photovoltaic-storage-DC-flexible system according to claim 9, characterized in that, Step 160 includes: Step 161: After the daily power dispatch is completed, calculate the actual clean energy consumption rate and total operating cost, and compare them with the target value. Step 162: Based on bias analysis, adjust the prediction model parameters and optimize the algorithm coefficients of the Long Short-Term Memory Network; Step 163: Iterate the multi-objective optimization model parameters once a week, including: Calculate the comprehensive evaluation index: ； in, For clean energy consumption rate, For actual cost, For target cost; The model parameters are optimized using the following update formula from the particle swarm optimization algorithm: ； in, For particle velocity, For inertial weights, As a learning factor, , It is a random number. For the optimal position of an individual, To find the globally optimal position; by limiting the LSTM learning rate. and ADMM penalty factor Within a reasonable range, ensure the stability of the algorithm.

Images