A load dispatching method and system for prefabricated substations

By constructing an objective function that includes generation costs and consistency penalties, and combining historical load and renewable energy data, the load dispatching of prefabricated substations is optimized, solving the problem of dispatching schemes deviating from actual operating needs and improving the safety and economy of the dispatching schemes.

CN123121597APending Publication Date: 2026-11-17HENAN TIANLI ELECTRIC EQUIP
View PDF 0 Cites 0 Cited by

Patent Information

Application Number
CN202611497606.X
Authority / Receiving Office
CN · China
Patent Type
Applications(China)
Current Assignee / Owner
Filing Date
2026-09-24
Publication Date
2026-11-17

AI Technical Summary

Technical Problem

Existing load dispatching methods for prefabricated substations are difficult to balance renewable energy fluctuations, historical load consistency, and transformer thermal safety constraints, which makes dispatching schemes prone to deviating from actual operating needs, increasing the risk of operational imbalance and transformer overheating.

Method used

An objective function containing generation cost and consistency penalty terms is constructed. By combining historical load data and renewable energy output forecast data, the allowable fluctuation range is determined. The load dispatch scheme is adjusted through a piecewise penalty function, the transformer winding hot spot temperature is predicted, and the optimization solution is performed by combining power flow and capacity constraints.

Benefits of technology

It improves the safety and adaptability of load dispatching schemes, reduces the risk of transformer overheating damage, and enhances the economy and consistency of dispatching schemes.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN123121597A_ABST
    Figure CN123121597A_ABST
Patent Text Reader

Abstract

This invention provides a load dispatching method and system for prefabricated substations. The method includes collecting historical load data, transformer status data, and renewable energy output forecast data for the dispatching period of the prefabricated substation; constructing an objective function including a generation cost term and a consistency penalty term; calculating the historical load projection benchmark and weighted dispersion for the dispatching time based on load data from multiple historical moments, and determining the allowable load fluctuation range centered on the historical load projection benchmark by combining the width of the renewable energy output forecast interval; when the planned load is within this range, the consistency penalty value is zero; when the planned load exceeds this range, a segmented penalty value is calculated based on the difference between it and the nearest interval endpoint; further, the winding hot spot temperature for the next dispatching period is predicted based on transformer status data, and the segmented penalty function is adjusted asymmetrically accordingly; and the optimal load dispatching scheme is obtained by combining power flow constraints and capacity constraints.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] This application belongs to the field of dispatching, and in particular relates to a load dispatching method and system for prefabricated substations. Background Technology

[0002] In localized power distribution networks that incorporate renewable energy, load dispatching is a crucial element in ensuring stable system operation and controlling operating costs. Conventional load dispatching methods for prefabricated substations typically aim to reduce generation costs or overall operating costs. They combine grid power flow constraints and transformer rated capacity constraints, utilizing historical load data and renewable energy output forecasts to formulate future dispatching plans. These methods can, to some extent, balance supply and demand and address power fluctuations caused by the integration of distributed energy resources.

[0003] However, in actual operation, renewable energy output exhibits randomness and volatility, making it difficult for existing dispatch strategies to effectively correlate the reasonable fluctuation range of planned load with historical load distribution characteristics. When planned load deviates significantly from historical conventional load patterns, the lack of corresponding penalty or constraint mechanisms can easily lead to dispatch schemes deviating from conventional electricity consumption characteristics, increasing the risk of local operational imbalances. Simultaneously, existing methods do not fully consider the impact of load changes on the internal thermal processes of transformers. The hot spot temperature of transformer windings directly relates to equipment insulation aging and operational safety. A single static capacity constraint is insufficient to characterize the heat accumulation effect and thermal tolerance state within a specific dispatch period, potentially leading to overly conservative dispatch results or overheating risks under continuous heavy load conditions. Therefore, a load dispatch mechanism for prefabricated substations is urgently needed that, while considering operational economy, can define reasonable boundaries for load fluctuations and form a load dispatch scheme that balances equipment thermal safety and historical load consistency. Summary of the Invention

[0004] To address the problem that existing load dispatching technologies struggle to balance renewable energy fluctuations, historical load consistency, and transformer thermal safety constraints, leading to dispatching schemes that easily deviate from actual operational needs, this invention provides a load dispatching method and system for prefabricated substations.

[0005] In a first aspect, the present invention provides the following technical solution: a load dispatching method for a prefabricated substation, comprising: Acquire historical load data, transformer status data, and renewable energy output forecast data for the waiting period of the prefabricated substation, and construct an objective function that includes a power generation cost term and a consistency penalty term. Based on load data from multiple historical moments, the historical load projection benchmark and weighted dispersion for the current time to be dispatched are calculated. According to the interval width of the renewable energy output forecast data and the weighted dispersion, an allowable fluctuation interval centered on the historical load projection benchmark is determined. When the planned load value at the time to be dispatched exceeds the allowable fluctuation interval, the difference between the planned load value and the nearest endpoint of the allowable fluctuation interval is taken as the independent variable, and a consistency penalty value is calculated based on the piecewise penalty function. When the planned load value is within the allowable fluctuation interval, the consistency penalty value is zero. Based on the transformer status data, the winding hot spot temperature for the next scheduling period is predicted. Based on the winding hot spot temperature, the piecewise penalty function is asymmetrically adjusted. Based on the objective function, and combined with the power flow constraints and capacity constraints of the prefabricated substation, the optimal load scheduling scheme is obtained.

[0006] Secondly, the present invention provides the following technical solution: a prefabricated substation load dispatching system, comprising the following modules: A module is built to acquire historical load data, transformer status data, and renewable energy output forecast data for the period to be dispatched from the prefabricated substation, and to build an objective function that includes a power generation cost term and a consistency penalty term. The calculation module is used to calculate the historical load projection benchmark and weighted dispersion for the current time to be dispatched based on load data from multiple historical moments. Based on the interval width of the renewable energy output forecast data and the weighted dispersion, an allowable fluctuation interval centered on the historical load projection benchmark is jointly determined. When the planned load value for the time to be dispatched exceeds the allowable fluctuation interval, the difference between the planned load value and the nearest endpoint of the allowable fluctuation interval is taken as the independent variable, and a consistency penalty value is calculated based on a piecewise penalty function. When the planned load value is within the allowable fluctuation interval, the consistency penalty value is zero. The scheduling module is used to predict the winding hot spot temperature of the next scheduling period based on the transformer status data, perform asymmetric adjustment on the segmented penalty function based on the winding hot spot temperature, and solve the optimal load scheduling scheme based on the objective function and in combination with the power flow constraints and capacity constraints of the box-type substation.

[0007] This invention constructs an objective function that includes a generation cost term and a consistency penalty term, and optimizes it by combining power flow constraints and capacity constraints to balance dispatch economy and operational safety. By using the weighted dispersion of historical load and the width of the renewable energy output prediction interval, an allowable fluctuation range centered on the historical load projection benchmark is determined. Segmented penalties are applied only to planned loads exceeding this range, reducing dispatch results that deviate abnormally from historical load characteristics. Furthermore, the predicted transformer winding hot spot temperature is incorporated into the penalty function adjustment process. When overheating risk exists, the upper limit penalty intensity is increased, thereby suppressing dispatch schemes that continue to increase load, reducing the risk of transformer overheating damage, and improving the safety and adaptability of the load dispatch scheme. Attached Figure Description

[0008] Figure 1 This is a topology diagram of a microgrid system for a prefabricated substation. Figure 2 A schematic diagram illustrating the calculation of the weighted benchmark for historical load samples; Figure 3 This is a schematic diagram of the asymmetric consistency penalty function for incorporating transformer overheating risk. Detailed Implementation

[0009] To make the objectives, technical solutions, and advantages of this invention clearer, the invention will be further described in detail below with reference to the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are merely illustrative and not intended to limit the invention.

[0010] This invention provides a load dispatching method for prefabricated substations, comprising: S1: Obtain historical load data, transformer status data, and renewable energy output forecast data for the waiting period of the prefabricated substation, and construct an objective function that includes a power generation cost term and a consistency penalty term.

[0011] The data acquisition and monitoring control system installed within the prefabricated substation collects historical active load data every 15 minutes over the past 30 days. Simultaneously, temperature sensors and instrument transformers are used to collect data on the transformer's top oil temperature, ambient temperature, and three-phase current, serving as transformer status data.

[0012] The system acquires meteorological data on light intensity, wind speed, and temperature released by the meteorological center, and reads the rated capacity of photovoltaic modules, the rated capacity of wind turbines, and historical actual output data of wind and photovoltaic power. This data is then input into a pre-trained Long Short-Term Memory (LSTM) network model, which outputs predicted output data for photovoltaic and wind renewable energy sources for each scheduling cycle within the scheduled period, along with the prediction error range. The prediction error range includes the upper and lower limits of the prediction range corresponding to the same scheduled time.

[0013] An objective function is established to minimize the total operating cost. This objective function is expressed as the algebraic sum of the generation cost term and the consistency penalty term. The generation cost term includes the cost of purchasing electricity from the main grid, the depreciation cost of charging and discharging energy storage devices deployed within the prefabricated substation, and the fuel operating cost of the diesel generator and micro gas turbine. The cost of purchasing electricity from the main grid is calculated based on the grid power, electricity price, and scheduling step duration for each scheduling step. The depreciation cost of charging and discharging energy storage devices is calculated based on the charging power, discharging power, scheduling step duration, and unit charging / discharging depreciation coefficient of the energy storage devices. The fuel operating cost of the diesel generator and micro gas turbine is calculated based on the planned output of the corresponding equipment, scheduling step duration, and unit generation fuel cost. The consistency penalty term is used to constrain the degree to which the scheduling plan deviates from historical normal load characteristics. The microgrid system topology of the prefabricated substation is as follows: Figure 1 As shown, it consists of the main grid, transformers, AC busbars, distributed power sources, loads, and dispatch control system. Solid lines represent electrical connections, and dashed lines represent data or control connections.

[0014] As one possible implementation, acquiring historical load data, transformer status data, and renewable energy output forecast data for the period to be dispatched from the prefabricated substation includes: Read the actual active power values ​​at the same time and the adjacent previous time within the past 30 days from the database of the data acquisition and monitoring control system as historical load data; Sensors installed in the prefabricated substation are used to collect real-time data on the transformer top oil temperature, ambient temperature, and three-phase load current as transformer status data. Extract the predicted output power of wind power generation, the predicted output power of photovoltaic power generation, and their corresponding prediction intervals for each scheduling time within the scheduling period. Determine the predicted output power of renewable energy based on the sum of the predicted output power of wind power generation and the predicted output power of photovoltaic power generation. Determine the predicted output interval of renewable energy based on the sum of the upper and lower limits of the prediction intervals of the two.

[0015] Connecting to the SCADA data acquisition and monitoring control system's backend database via a communication network interface, and using the upcoming scheduling time (assuming 14:00 the next day) as a baseline, the system retrieves the actual active power meter readings for each of the past 30 days at 14:00 and the preceding scheduling interval, assuming a step size of 15 minutes, i.e., 13:45. For example, it extracts the load as 450kW at 14:00 on day 1 and 445kW at 13:45, and repeats this process until 30 power pair matrices consisting of the same time and the preceding time are collected, thus generating the required historical load data.

[0016] Meanwhile, at the edge of the site, data from the sensor array installed in the transformer substation is read via RS485 bus. This allows for the acquisition of the transformer top oil temperature (e.g., 55.2℃) returned by the PT100 oil surface temperature controller, the ambient temperature (e.g., 28.5℃) returned by the external meteorological temperature measuring instrument, and the effective values ​​of the three-phase load currents (e.g., 120A for phase A, 122A for phase B, and 119A for phase C) obtained through real-time sampling by the current transformer (CT). These parameters are then packaged and encapsulated as the current transformer status data.

[0017] The system acquires forecast data on solar irradiance, wind speed, and ambient temperature for the scheduled period released by the meteorological center, and reads the rated capacity of photovoltaic modules, rated capacity of wind turbines, and historical actual output data of wind and photovoltaic power recorded by the new energy monitoring device in the prefabricated substation. The solar irradiance, wind speed, ambient temperature, rated capacity, and historical actual output data are then used to form a model input sequence according to the scheduling step size. This sequence is input into a pre-trained long short-term memory network model, which outputs the predicted wind power output, predicted photovoltaic power output, and corresponding prediction error range for each scheduling cycle within the scheduled period.

[0018] The Long Short-Term Memory (LSTM) network model includes an input layer, at least one LSTM network layer, and a fully connected output layer. The input layer receives meteorological forecast features, equipment capacity features, and historical output features. The LSTM network layer extracts the temporal correlation features of wind power and photovoltaic power output as they change with meteorological conditions and historical output. The fully connected output layer outputs the predicted values ​​of wind power output, photovoltaic power output, and their prediction error ranges.

[0019] During the model training phase, historical meteorological data, historical equipment capacity data, and historical wind power and photovoltaic actual output data are used as training sample inputs. The actual output values ​​of wind power and photovoltaic power corresponding to the future scheduling step are used as supervision labels to train the long short-term memory network model. After training, the prediction error interval is determined based on the historical residual distribution between the model prediction value and the actual output value. The prediction error interval includes the upper limit and lower limit of the prediction interval corresponding to the same scheduling time.

[0020] For example, if the scheduling period is the next 4 hours and the scheduling step size is 15 minutes, then the prediction results are generated for 16 scheduling step sizes. Assuming that for 14:00 tomorrow, the model outputs a wind power prediction of 120.5kW and a photovoltaic power prediction of 300.2kW, then adding these two together gives the total renewable energy output prediction of 420.7kW at that time. Simultaneously, adding the upper limit of the wind power prediction interval to the upper limit of the photovoltaic power prediction interval gives the upper limit of the total renewable energy output prediction interval, and adding the lower limit of the wind power prediction interval to the lower limit of the photovoltaic power prediction interval gives the lower limit of the total renewable energy output prediction interval. This yields a time series prediction result that includes wind power, photovoltaic power, and total renewable energy output.

[0021] S2, based on load data from multiple historical moments, calculate the historical load projection benchmark and weighted dispersion for the current time to be dispatched. Based on the interval width of the renewable energy output forecast data and the weighted dispersion, jointly determine an allowable fluctuation range centered on the historical load projection benchmark. When the planned load value at the time to be dispatched exceeds the allowable fluctuation range, take the difference between the planned load value and the nearest endpoint of the allowable fluctuation range as the independent variable, and calculate the consistency penalty value based on the piecewise penalty function. When the planned load value is within the allowable fluctuation range, the consistency penalty value is zero.

[0022] During the optimization process, the distributed power generation plan, energy storage device charging and discharging plan, and adjustable load response are used as scheduling variables. The optimization solver generates candidate scheduling schemes under the conditions of satisfying power flow constraints and capacity constraints. The planned load value at the corresponding scheduling time is calculated based on each candidate scheduling scheme. The planned load value is the equivalent net active power load value on the AC bus side of the prefabricated substation calculated based on the candidate scheduling scheme of the current iteration.

[0023] Historical load data from the past few days that are at the same time point as the current scheduling time are extracted. The weight of each historical sample is determined according to the load change magnitude between the current scheduling time and the adjacent previous time. The historical sample with the smaller load change magnitude corresponds to a higher weight. The historical load projection benchmark is calculated based on the load value of each historical sample and its corresponding weight, and the weighted dispersion is further calculated.

[0024] The difference between the upper and lower limits of the forecast interval for the same dispatchable time in the renewable energy output forecast data in the aforementioned steps is extracted as the interval width. Using the first weight parameter and the second weight parameter, the product of the first weight parameter and the interval width plus the product of the second weight parameter and the weighted dispersion is used as the interval offset. The lower limit of the allowable fluctuation interval is obtained by subtracting the interval offset from the historical load projection benchmark, and the upper limit of the allowable fluctuation interval is obtained by adding the interval offset to the historical load projection benchmark.

[0025] When calculating the consistency penalty term, the planned load value at the time to be scheduled is compared with the allowable fluctuation range in real time. If the planned load value is greater than the upper limit of the allowable fluctuation range, the independent variable is the difference between the planned load value and the upper limit. If the planned load value is less than the lower limit of the allowable fluctuation range, the independent variable is the difference between the lower limit and the planned load value.

[0026] Set a segmented penalty function. When the planned load value is higher than the upper limit of the allowable fluctuation range, calculate the consistency penalty value according to the sum of the first penalty term and the second penalty term on the upper limit side. When the planned load value is lower than the lower limit of the allowable fluctuation range, calculate the consistency penalty value according to the sum of the first penalty term and the second penalty term on the lower limit side. When the planned load value is within the allowable fluctuation range, set the consistency penalty value to zero.

[0027] As one possible implementation, the step of calculating the historical load projection benchmark and weighted dispersion for the current time to be scheduled, based on load data from multiple historical moments, includes: Calculate the absolute value of the difference between the load value of each historical sample at the current time to be scheduled and the load value at the adjacent previous time, and obtain the first parameter representing the load change amplitude of the historical sample. Add the first parameter to the preset minimum normal number and take the reciprocal to obtain the initial weight of the corresponding historical sample. Divide the initial weight by the sum of the initial weights of all historical samples to obtain the normalized weight corresponding to the load value of each historical sample. The historical load projection benchmark for the current time to be scheduled is obtained by weighted summation based on the load values ​​of each historical sample and their corresponding normalized weights. The squared deviation between each historical sample load value and the historical load projection benchmark is calculated, and the squared deviation is multiplied by the corresponding normalized weight to obtain the discrete term. The weighted dispersion, which represents the stability of the historical load pattern, is obtained by summing all the discrete terms and taking the square root.

[0028] For the extracted 30 days of historical data, for the record on day i, extract the load value at the same time. and the load value of the adjacent previous time moment Using formulas Perform difference and absolute value operations to obtain the first parameter. For example, if the load is 500kW at 14:00 on a certain day and 480kW at 13:45, then The calculation is 20kW, reflecting the load variation during that time period on that day.

[0029] Using pre-calibrated minimal positive numbers For example, setting it to 0.001kW will... and Add and perform reciprocal operation For example, 0.04999 yields the initial weight value. This process is repeated to obtain the initial weights for all 30 days. Then, the initial weight for a specific day is divided by the sum of the initial weights for all 30 days, thus obtaining the dimensionless weight coefficients for each historical moment distributed between 0 and 1. Higher weights will be assigned to historical reference days where load changes are more stable.

[0030] When determining the baseline value, the load values ​​at the same time in each historical moment within 30 days are used. Each of them and their corresponding normalized weights Multiply them, then sum the 30 multipliers. Assume the calculated... The value is 495.5 kW, which serves as the historical load projection reference for the current time to be scheduled.

[0031] Extract load values ​​for each historical day sequentially. minus For example, if we obtain 4.5kW, we square it to get the squared deviation, and then multiply it by its respective weight. Generate discrete terms corresponding to each historical sample. Summate all discrete terms from the 30 time points, and take the square root of the overall sum. For example, if the sum is calculated to be 150.06kW... 2 The weighted dispersion D, obtained by taking the square root, is approximately 12.25 kW. Historical load samples and load differences are also relevant. The principle of weight calculation is as follows Figure 2 As shown, hollow circles represent the load at the adjacent previous time point, and solid circles represent the load at the same time point. Sample weights are assigned based on the stability of load changes, and the weighted sum is used to obtain the historical load projection benchmark. .

[0032] As one possible implementation, determining an allowable fluctuation range centered on the historical load projection benchmark based on the interval width of the renewable energy output forecast data and the weighted dispersion includes: The interval width is obtained by subtracting the lower limit of the prediction interval from the upper limit of the prediction interval at the same time to be dispatched in the renewable energy output prediction data. Multiply the interval width by the first weighting parameter to obtain the first width expansion amount; The weighted dispersion is multiplied by the second weighting parameter to obtain the second broadening amount; The first width expansion amount and the second width expansion amount are added together to calculate the range expansion radius of the allowable fluctuation range; The upper limit of the allowable fluctuation range is obtained by adding the interval expansion radius value to the historical load projection benchmark. The lower limit of the allowable fluctuation range is obtained by subtracting the interval expansion radius value from the historical load projection benchmark.

[0033] For the time to be dispatched, extract the upper limit value of the predicted power from the predicted range of renewable energy output corresponding to the same time to be dispatched. For example, given a predicted upper limit of 350kW for combined wind and solar power, and a predicted lower limit value, such as a predicted lower limit of 300kW, the range width is calculated to be 350-300=50kW.

[0034] The first weight parameter, pre-defined (e.g., 0.4), is multiplied by the interval width to obtain the first width expansion, E1 = 20kW, representing the source-end fluctuation characteristics. Simultaneously, the weighted dispersion D generated in the previous step (e.g., 12.25kW) is used, and the second weight parameter, presumably 1.5, is extracted. Multiplying these two values ​​yields the second width expansion, E2 = 18.375kW. The first and second width expansions are then added together to obtain the interval expansion radius, B = 38.375kW.

[0035] Obtain historical load projection reference The value is 495.5kW, which is used as the central anchor point. Addition and subtraction are performed to determine the boundary: through... Calculate the upper limit of the allowable fluctuation range, which is 533.875kW; through The lower limit of the allowable fluctuation range is calculated, which is 457.125kW. This step transforms the source and load characteristic parameters into a range threshold band [457.125kW, 533.875kW], so that the scheduling optimization constraint boundary can change according to the real-time operating environment.

[0036] As one possible implementation, when the planned load value at the time to be scheduled exceeds the allowable fluctuation range, the difference between the planned load value and the nearest endpoint of the allowable fluctuation range is taken as the independent variable, and a consistency penalty value is calculated based on the piecewise penalty function, including: When the planned load value is greater than the upper limit of the allowable fluctuation range, first calculate the difference between the planned load value and the upper limit of the allowable fluctuation range, and use this difference as the upper limit over-limit difference; then, according to the preset upper limit side secondary penalty coefficient and upper limit side primary penalty coefficient, perform secondary term penalty calculation and primary term penalty calculation on the upper limit over-limit difference respectively, and add the two to obtain the consistency penalty value when it is higher than the upper limit of the allowable fluctuation range. When the planned load value is less than the lower limit of the allowable fluctuation range, first calculate the difference between the lower limit of the allowable fluctuation range and the planned load value, and use this difference as the lower limit over-limit difference; then, according to the preset lower limit side secondary penalty coefficient and lower limit side primary penalty coefficient, perform secondary term penalty calculation and primary term penalty calculation on the lower limit over-limit difference respectively, and add the two to obtain the consistency penalty value when it is lower than the lower limit of the allowable fluctuation range.

[0037] During the solution exploration process of the optimization solver, the basic electricity demand forecast value for the time to be scheduled is first determined based on historical electricity demand data. This historical electricity demand data is read from user-side metering devices or data acquisition and monitoring control systems, and is used to characterize the actual electricity demand of the load supplied by the prefabricated substation at the corresponding historical time. The basic electricity demand forecast value represents the electricity demand before deducting the current distributed power generation output plan, energy storage discharge plan, and adjustable load reduction. Specifically, M historical electricity demand values ​​that are the same as the time to be scheduled within the historical period are selected and sorted from most recent to oldest according to the collection time to obtain a historical electricity demand sequence. The initial weight corresponding to the i-th historical electricity demand value in the historical electricity demand sequence is set to 1 / i, and each initial weight is divided by the sum of all initial weights to obtain a normalized weight. Each historical electricity demand value is multiplied by its corresponding normalized weight and then summed to obtain the basic electricity demand forecast value.

[0038] The optimization solver uses distributed power generation output plans, energy storage device charging and discharging plans, and adjustable load response as scheduling variables, and generates current candidate scheduling schemes under the conditions of satisfying power flow constraints, capacity constraints, and equipment output constraints. Among them, wind power output plans and photovoltaic power output plans are determined based on the predicted wind power and photovoltaic power output values ​​and their allowable output ranges at the corresponding scheduling time, energy storage device charging and discharging plans are determined based on the upper limit of energy storage device power and state of charge constraints, and adjustable load response is determined based on the preset responsive load range.

[0039] Based on the current candidate scheduling scheme, the planned load value is calculated according to the active power balance relationship. Specifically, the planned load value The equivalent net active load value of the AC bus side of the prefabricated substation under the current candidate dispatch scheme is calculated as follows: subtract the planned output value of the distributed power source, the discharge power of the energy storage device, and the adjustable load reduction from the basic electricity demand forecast value at the time of dispatch, and add the charging power of the energy storage device; when the adjustable load response indicates an increase in electricity consumption, it is included in the direction of increasing the equivalent net active load.

[0040] For example, for the scheduled time of 14:00 tomorrow, weighted calculations are performed using historical electricity demand data corresponding to 14:00 within the past 30 days, resulting in a basic electricity demand forecast of 800kW. The candidate scheduling scheme generated by the optimization solver in the current iteration is: planned wind power output of 120.5kW, planned photovoltaic power output of 300.2kW, energy storage device discharge power of 50kW, adjustable load reduction of 20kW, and energy storage device charging power of 0kW. The planned load value under this candidate scheduling scheme is... =800−120.5−300.2−50−20+0=309.3kW. If the charging power of the energy storage device in another candidate scheduling scheme is 30kW, then this 30kW is included in the direction of increasing the equivalent net active load.

[0041] Taking another candidate scheduling scheme as an example, the planned load value obtained from this candidate scheme is compared with the calculated range of [457.125kW, 533.875kW]. If the planned load value is greater than the upper limit threshold, for example... If the power is 550kW, then the system enters the upper limit penalty path. Subtracting the upper limit of 533.875kW from 550kW gives the over-limit difference x = 16.125kW.

[0042] Read the upper limit side second penalty coefficient Set at 0.5 yuan / (kW) 2 ), and read the penalty coefficient on the upper limit side. The price is set at 2.0 yuan / kW. For cases where the planned load exceeds the upper limit of the allowable fluctuation range, the price will be calculated according to... Calculate the consistency penalty value, where x is the upper bound deviation. (As mentioned above...) Taking a value of 550kW and an upper limit of 533.875kW as an example, x = 16.125kW. ≈162.25 yuan.

[0043] If the generated planned load value is less than the lower limit of the allowable fluctuation range, for example... If the load is 440kW, then the system switches to the lower limit penalty path. Subtracting the planned load of 440kW from the lower limit value of 457.125kW yields the lower limit overrun difference y = 17.125kW. The secondary penalty coefficient on the lower limit side is then read. Set at 0.3 yuan / (kW) 2 ), and read the lower limit side first penalty coefficient. The price is set at 1.5 yuan / kW. For planned load values ​​below the lower limit of the allowable fluctuation range, the price will be calculated according to... Calculate the consistency penalty value, then ≈113.67 yuan. when When the value is within the allowable fluctuation range, for example, when it equals 500kW, the consistency penalty term C is set to zero.

[0044] S3. Based on the transformer status data, predict the winding hot spot temperature for the next scheduling period. Based on the winding hot spot temperature, perform asymmetric adjustment on the piecewise penalty function. Based on the objective function, and in combination with the power flow constraints and capacity constraints of the prefabricated substation, solve the problem to obtain the optimal load scheduling scheme.

[0045] Based on the transformer hot spot temperature calculation model, the collected transformer top oil temperature, ambient temperature, and three-phase current are used as inputs. Combined with the transformer's rated loss, preset rated top oil temperature rise parameters, preset rated hot spot temperature rise parameters, and thermal time constant, the predicted value of the transformer winding hot spot temperature for the next scheduling period is calculated.

[0046] A hotspot temperature safety threshold is set. When the predicted winding hotspot temperature exceeds this safety threshold, it indicates a risk of transformer overheating and aging. In this case, the piecewise penalty function is adjusted, increasing the primary and secondary penalty coefficients when the temperature exceeds the upper limit of the allowable fluctuation range, while keeping the penalty coefficient unchanged when the temperature is below the lower limit of the allowable fluctuation range. This ensures that the dispatching system prioritizes suppressing further load increases under overheating risk. Conversely, when the predicted winding hotspot temperature is below the safety threshold, the original penalty coefficients are maintained to preserve dispatching space for the transformer within the safe temperature range. The consistency penalty function change pattern before and after transformer overheating is as follows: Figure 3 As shown.

[0047] A set of constraints is constructed, including power flow constraints such as the active and reactive power balance equations at nodes and upper and lower limits for node voltage amplitude. Capacity constraints include constraints on transformer apparent power not being overloaded, maximum transmission power of lines, and upper and lower limits for the charging and discharging power and state of charge of energy storage devices.

[0048] The constructed objective function and all constraints are transformed into a constraint optimization model. The constraint optimization solver is called to perform iterative optimization calculations on the model, and outputs the output plans of each distributed power source, the energy storage charging and discharging plan, and the control timing of the adjustable load that minimize the objective function value. This is then used as the optimal load scheduling scheme and sent to the controller of the prefabricated substation for execution.

[0049] As one possible implementation, predicting the winding hot spot temperature for the next scheduling period based on the transformer state data includes: The current load rate is determined by the ratio of the effective value of the three-phase load current to the corresponding rated current value at the current moment. Based on the current load rate, the preset rated hot spot temperature rise parameter and the preset hot spot temperature rise index, calculate the current winding hot spot temperature difference and add it to the current transformer top oil temperature value to obtain the initial winding hot spot temperature. The steady-state limit temperature of the winding is calculated based on the ambient temperature value, the preset loss ratio constant, the preset rated top oil temperature rise parameter, the preset rated hot spot temperature rise parameter, and the load rate used for steady-state temperature rise prediction. The load rate used for steady-state temperature rise prediction is the current load rate during real-time state prediction and the planned load rate corresponding to the candidate scheduling scheme during the optimization solution process. The thermal decay factor is calculated based on the scheduling delay step and the thermal time constant, and the winding hot spot temperature value at the next scheduling moment is obtained based on the initial winding hot spot temperature, the winding steady-state limit temperature and the thermal decay factor.

[0050] Obtain the top oil temperature returned by the sensor at this time, assuming the measured value is 55.2℃. Extract the effective values ​​of the currently collected three-phase AC current, assuming phase A is 120A, phase B is 122A, and phase C is 119A, and divide them by the corresponding phase rated current value, assuming the corresponding phase rated current value is 200A, to obtain the load ratio of each phase; average the load ratio of each phase to obtain the current load rate of approximately 0.602.

[0051] The current load rate is raised to the power of a preset hot spot temperature rise index, and then multiplied by a preset rated hot spot temperature rise parameter to obtain the current winding hot spot temperature difference. The preset rated hot spot temperature rise parameter represents the benchmark for the winding hot spot temperature rise of the transformer under rated load, and the preset hot spot temperature rise index represents the degree of influence of load rate changes on the winding hot spot temperature rise. For example, when the current load rate is approximately 0.602, raising the current load rate to the power of the preset hot spot temperature rise index and multiplying it by the preset rated hot spot temperature rise parameter yields a current winding hot spot temperature difference of 15.0℃. The preset hot spot temperature rise index is preferably set to 1.6, and the preset rated hot spot temperature rise parameter is preferably set to 33.5℃. This current winding hot spot temperature difference is then added to the top oil temperature to obtain the initial winding hot spot temperature at the current moment. =70.2℃.

[0052] The current ambient temperature is obtained, assumed to be 28.5℃; the copper-iron loss ratio constant R is read, assumed to be R=5, indicating that the ratio of copper loss to iron loss under rated operating conditions is 5:1; simultaneously, the preset rated top-layer oil temperature rise parameter, the preset rated hot spot temperature rise parameter, the top-layer oil temperature rise index, and the hot spot temperature rise index are read. In real-time state prediction, the load loss correction coefficient is calculated using the current load rate; in the optimization solution process, the load loss correction coefficient is calculated using the planned load rate corresponding to the candidate scheduling scheme. Specifically, the load rate used for steady-state temperature rise prediction is first squared, multiplied by the copper-iron loss ratio constant R, and then 1 is added. The result is divided by the sum of R and 1 to obtain the load loss correction coefficient; then, this load loss correction coefficient is exponentially calculated according to the top-layer oil temperature rise index and multiplied by the preset rated top-layer oil temperature rise parameter to obtain the predicted top-layer oil temperature rise value. Finally, the load rate used for steady-state temperature rise prediction is exponentially calculated according to the hot spot temperature rise index and multiplied by the preset rated hot spot temperature rise parameter to obtain the predicted winding hot spot temperature difference value. Finally, the predicted values ​​of ambient temperature, top oil temperature rise, and winding hot spot temperature difference are added together to obtain the predicted steady-state limit temperature of the transformer. For example, if the preset rated top oil temperature rise parameter is 64.4℃ and the top oil temperature rise index is 0.8, and the preset rated hot spot temperature rise parameter is 33.5℃ and the hot spot temperature rise index is 1.6, then by substituting these parameters, the calculation is obtained. =78.5℃.

[0053] Then, the pre-set scheduling delay step size variable Δt and the transformer thermal time constant τ are read, assuming Δt is 15 minutes and τ is 120 minutes. Δt is divided by τ to obtain 0.125, and its negative is taken as the exponent term, i.e., -0.125. The exponential function value corresponding to this exponent term is then calculated with the natural constant e as the base, yielding the thermal decay factor 0.8825. This thermal decay factor is used to represent the temperature of the current winding hot spot towards the steady-state limit temperature within one scheduling step. The degree of proximity.

[0054] The initial winding hot spot temperature was 70.2℃. Multiplying by the attenuation factor, we get the first value of 61.95℃; using the steady-state limiting temperature of 78.5℃... Multiplying by one and subtracting the attenuation factor, we get the second value: 9.22℃. Adding the first value of 61.95℃ to the second value of 9.22℃ gives a total of 71.17℃, which will be used as the winding hot spot temperature for the next scheduling time. .

[0055] During the optimization process, when a candidate scheduling scheme changes the transformer load level for the next scheduling period, the planned load factor for the next scheduling period is calculated based on the planned active power, planned reactive power, and transformer rated capacity corresponding to the candidate scheduling scheme. This planned load factor is then used as the load factor for steady-state temperature rise prediction, and the winding steady-state limit temperature is recalculated. and the winding hot spot temperature at the next scheduling time This allows the predicted winding hot spot temperature to be updated as the candidate scheduling scheme changes.

[0056] As one possible implementation, the asymmetric adjustment of the piecewise penalty function based on the winding hot spot temperature includes: When the predicted winding hot spot temperature value for the next scheduling period exceeds the transformer safety temperature parameter constant, the difference between the winding hot spot temperature value and the transformer safety temperature parameter constant is calculated, and this difference is used as the overrun temperature difference value. Based on the over-temperature difference value and the preset mapping gain constant, the penalty increment multiplier is calculated using an exponential mapping method, so that the penalty increment multiplier increases as the over-temperature difference value increases; The upper limit-side secondary penalty coefficient and upper limit-side primary penalty coefficient used when the planned load value exceeds the upper limit of the allowable fluctuation range are multiplied by the penalty increment multiplier, respectively, to increase the consistency penalty intensity when the planned load value exceeds the upper limit of the allowable fluctuation range. When the planned load value is lower than the lower limit of the allowable fluctuation range, the secondary penalty coefficient and the primary penalty coefficient on the lower limit side will not be scaled or adjusted, and will remain unchanged.

[0057] In another example, the predicted winding hotspot temperature for the next scheduling period The temperature is 105.0℃, and the transformer safety temperature parameter constant is 98.0℃. Therefore, 105.0℃ is greater than 98.0℃. The subtraction operation is performed to obtain the over-temperature difference value ΔT = 7.0℃.

[0058] Find the mapping gain constant k, assuming a configuration value of 0.2 / ℃. Multiply the mapping gain constant k by the excess temperature difference ΔT, i.e., 0.2 × 7.0, to obtain the exponent term 1.4. Calculate the exponent value M≈4.055 with the natural constant e as the base, and use it as the penalty increment multiplier M.

[0059] After obtaining the incremental multiplier M, retrieve the original upper limit side secondary penalty coefficient, such as 0.5, and the upper limit side primary penalty coefficient, such as 2.0, and multiply them by the penalty incremental multiplier 4.055 to obtain the adjusted upper limit side penalty coefficient.

[0060] After the operation is completed, the upper limit side secondary penalty coefficient is updated to 2.0275, and the upper limit side primary penalty coefficient is updated to 8.11. For the lower limit side secondary penalty coefficient of 0.3 and the lower limit side primary penalty coefficient of 1.5, which are lower than the lower limit, no scaling update is performed to keep their values ​​unchanged, so as to avoid additional suppression in the load reduction direction.

[0061] As one possible implementation, the optimal load dispatching scheme is obtained by solving the objective function in conjunction with the power flow constraints and capacity constraints of the prefabricated substation, including: Based on the safe operating range of voltage at each node of the system, the maximum transmission capacity of each substation branch, and the power balance relationship, power flow constraints are established. Based on the rated capacity indicated on the transformer nameplate, the planned apparent power output of the transformer during dispatching and operation shall not exceed its maximum allowable output, thus establishing a capacity constraint condition. The objective function, power flow constraints, and capacity constraints are combined to establish an augmented Lagrangian function that includes an optimization variable vector, equality constraint multipliers, and inequality constraint multipliers. During the iterative solution process, the optimization variable vector is updated successively based on the first-order gradient information and second-order partial derivative information of the augmented Lagrange function, and the corresponding constraint multipliers are updated synchronously. After each iteration update, the L2 norm of the current gradient vector is calculated and compared with the preset convergence tolerance. When the L2 norm is less than the preset convergence tolerance, the iteration is stopped and the current solution is determined to be the optimal solution after convergence. The optimal solution is converted into a sequence of control commands corresponding to distributed power sources, energy storage devices and adjustable loads within the prefabricated substation, and the sequence of control commands is issued as the optimal load scheduling scheme.

[0062] In the calculation phase, node and branch parameters are imported based on the microgrid topology of the prefabricated substation, and the safe operating voltage range for each node is set, for example, from 0.95 pu to 1.05 pu. The wind power output plan, photovoltaic power output plan, diesel generator output plan, micro gas turbine output plan, energy storage device charging power, energy storage device discharging power, adjustable load response, node voltage variables, and branch power flow variables corresponding to each scheduling step within the scheduling period are arranged in a preset order to form an optimization variable vector X.

[0063] Based on the active power balance and reactive power balance relationships at the nodes, an equal power flow constraint is established, denoted as . =0; Based on the node voltage safety range, branch maximum transmission capacity, transformer capacity limit, energy storage charging and discharging power limit, and distributed power output upper and lower limits, establish inequality operating constraints, denoted as ≤0.

[0064] Retrieve the maximum rated apparent capacity of the transformer from the equipment register. For example, 630kVA. For the current candidate scheduling scheme, read the planned active power flowing through the transformer branch from the optimization variable vector X. and obtain the planned reactive power. ;Will and The sum of squares and the square root are combined to obtain the planned apparent power. and establish Capacity constraints. For example, when When the power is 630kVA, the constraint is that the planned apparent power does not exceed 630kVA.

[0065] Among them, the planned active power The planned active power flowing through the transformer under the current candidate scheduling scheme; the planned reactive power The reactive-active power ratio is calculated through conversion. Specifically, historical active power and reactive power measurements are read from the data acquisition and monitoring control system for the same or adjacent operating time as the time to be scheduled. After excluding samples with zero historical active power measurements, the historical reactive power measurements are divided by the historical active power measurements to obtain the reactive-active power ratio for the corresponding time period. When multiple historical samples exist, the average or weighted average of the multiple reactive-active power ratios is taken to obtain the predicted reactive-active power ratio. Then, the planned active power under the current candidate scheduling scheme is calculated. Multiplying the predicted reactive power-to-active power ratio by the predicted reactive power-to-active power ratio yields the planned reactive power. For example, if the active power measurement value of the historical time period adjacent to the time to be dispatched is 500kW and the reactive power measurement value is 240kvar, then the reactive power-to-active power ratio is 0.48; if the planned active power flowing through the transformer under the current candidate dispatch scheme... The planned reactive power is 309.3kW. It is the product of 309.3kW and 0.48, which is approximately 148.46kvar.

[0066] Establish a comprehensive objective function f(X), combine equality power flow constraints and inequality operation constraints, and construct an augmented Lagrangian function L(X,λ,μ) using the equality constraint multiplier column vector λ and the inequality constraint multiplier column vector μ.

[0067] For the k-th iteration, calculate the first derivative of L(X,λ,μ) with respect to the optimization variable vector X to obtain the gradient column vector ∇L, and calculate the corresponding Hessian matrix H, then substitute it into the descent equation. Parameters are updated in the middle, and synchronously according to the process. Update the equality constraint multipliers, and in accordance with Update the inequality constraint multipliers, where ρ is the preset penalty update coefficient.

[0068] After each update, the square root of the sum of squares of all elements in the current gradient column vector ∇L is calculated to obtain the L2 Euclidean norm 2. For example, if the gradient norm 2 calculated in the current step is 0.0051, it is compared with the convergence tolerance ε, which is set to, for example, 1.0 × 10⁻⁶. -4 When the gradient norm 2 decreases to 0.00008 and is less than ε during iteration, the convergence condition is met, the iteration stops, and the optimal solution vector is output. .

[0069] From the optimal solution vector The corresponding variable values ​​are directly extracted to form a load scheduling decision sequence. For example, from... The active power output of the distributed power unit is set to 210.5kW, the discharge power of the energy storage device is set to 110.2kW, and the adjustable load enters the preset response state. Then, the load scheduling decision sequence is converted into a control command message sequence and sent to the corresponding execution terminal to complete the load scheduling scheme release.

[0070] This invention also proposes a load dispatching system for prefabricated substations, comprising the following modules: A module is built to acquire historical load data, transformer status data, and renewable energy output forecast data for the period to be dispatched from the prefabricated substation, and to build an objective function that includes a power generation cost term and a consistency penalty term. The calculation module is used to calculate the historical load projection benchmark and weighted dispersion for the current time to be dispatched based on load data from multiple historical moments. Based on the interval width of the renewable energy output forecast data and the weighted dispersion, an allowable fluctuation interval centered on the historical load projection benchmark is jointly determined. When the planned load value for the time to be dispatched exceeds the allowable fluctuation interval, the difference between the planned load value and the nearest endpoint of the allowable fluctuation interval is taken as the independent variable, and a consistency penalty value is calculated based on a piecewise penalty function. When the planned load value is within the allowable fluctuation interval, the consistency penalty value is zero. The scheduling module is used to predict the winding hot spot temperature of the next scheduling period based on the transformer status data, perform asymmetric adjustment on the segmented penalty function based on the winding hot spot temperature, and solve the optimal load scheduling scheme based on the objective function and in combination with the power flow constraints and capacity constraints of the box-type substation.

[0071] As one possible implementation, acquiring historical load data, transformer status data, and renewable energy output forecast data for the period to be dispatched from the prefabricated substation includes: Read the actual active power values ​​at the same time and the adjacent previous time within the past 30 days from the database of the data acquisition and monitoring control system as historical load data; Sensors installed in the prefabricated substation are used to collect real-time data on the transformer top oil temperature, ambient temperature, and three-phase load current as transformer status data. Extract the predicted output power of wind power generation, the predicted output power of photovoltaic power generation, and their corresponding prediction intervals for each scheduling time within the scheduling period. Determine the predicted output power of renewable energy based on the sum of the predicted output power of wind power generation and the predicted output power of photovoltaic power generation. Determine the predicted output interval of renewable energy based on the sum of the upper and lower limits of the prediction intervals of the two.

[0072] As one possible implementation, the step of calculating the historical load projection benchmark and weighted dispersion for the current time to be scheduled, based on load data from multiple historical moments, includes: Calculate the absolute value of the difference between the load value of each historical sample at the current time to be scheduled and the load value at the adjacent previous time, and obtain the first parameter representing the load change amplitude of the historical sample. Add the first parameter to the preset minimum normal number and take the reciprocal to obtain the initial weight of the corresponding historical sample. Divide the initial weight by the sum of the initial weights of all historical samples to obtain the normalized weight corresponding to the load value of each historical sample. The historical load projection benchmark for the current time to be scheduled is obtained by weighted summation based on the load values ​​of each historical sample and their corresponding normalized weights. The squared deviation between each historical sample load value and the historical load projection benchmark is calculated, and the squared deviation is multiplied by the corresponding normalized weight to obtain the discrete term. The weighted dispersion, which represents the stability of the historical load pattern, is obtained by summing all the discrete terms and taking the square root.

[0073] As one possible implementation, determining an allowable fluctuation range centered on the historical load projection benchmark based on the interval width of the renewable energy output forecast data and the weighted dispersion includes: The interval width is obtained by subtracting the lower limit of the prediction interval from the upper limit of the prediction interval at the same time to be dispatched in the renewable energy output prediction data. Multiply the interval width by the first weighting parameter to obtain the first width expansion amount; The weighted dispersion is multiplied by the second weighting parameter to obtain the second broadening amount; The first width expansion amount and the second width expansion amount are added together to calculate the range expansion radius of the allowable fluctuation range; The upper limit of the allowable fluctuation range is obtained by adding the interval expansion radius value to the historical load projection benchmark. The lower limit of the allowable fluctuation range is obtained by subtracting the interval expansion radius value from the historical load projection benchmark.

[0074] As one possible implementation, when the planned load value at the time to be scheduled exceeds the allowable fluctuation range, the difference between the planned load value and the nearest endpoint of the allowable fluctuation range is taken as the independent variable, and a consistency penalty value is calculated based on the piecewise penalty function, including: When the planned load value is greater than the upper limit of the allowable fluctuation range, first calculate the difference between the planned load value and the upper limit of the allowable fluctuation range, and use this difference as the upper limit over-limit difference; then, according to the preset upper limit side secondary penalty coefficient and upper limit side primary penalty coefficient, perform secondary term penalty calculation and primary term penalty calculation on the upper limit over-limit difference respectively, and add the two to obtain the consistency penalty value when it is higher than the upper limit of the allowable fluctuation range. When the planned load value is less than the lower limit of the allowable fluctuation range, first calculate the difference between the lower limit of the allowable fluctuation range and the planned load value, and use this difference as the lower limit over-limit difference; then, according to the preset lower limit side secondary penalty coefficient and lower limit side primary penalty coefficient, perform secondary term penalty calculation and primary term penalty calculation on the lower limit over-limit difference respectively, and add the two to obtain the consistency penalty value when it is lower than the lower limit of the allowable fluctuation range.

[0075] As one possible implementation, predicting the winding hot spot temperature for the next scheduling period based on the transformer state data includes: The current load rate is determined by the ratio of the effective value of the three-phase load current to the corresponding rated current value at the current moment. Based on the current load rate, the preset rated hot spot temperature rise parameter and the preset hot spot temperature rise index, calculate the current winding hot spot temperature difference and add it to the current transformer top oil temperature value to obtain the initial winding hot spot temperature. The steady-state limit temperature of the winding is calculated based on the ambient temperature value, the preset loss ratio constant, the preset rated top oil temperature rise parameter, the preset rated hot spot temperature rise parameter, and the load rate used for steady-state temperature rise prediction. The load rate used for steady-state temperature rise prediction is the current load rate during real-time state prediction and the planned load rate corresponding to the candidate scheduling scheme during the optimization solution process. The thermal decay factor is calculated based on the scheduling delay step and the thermal time constant, and the winding hot spot temperature value at the next scheduling moment is obtained based on the initial winding hot spot temperature, the winding steady-state limit temperature and the thermal decay factor.

[0076] As one possible implementation, the asymmetric adjustment of the piecewise penalty function based on the winding hot spot temperature includes: When the predicted winding hot spot temperature value for the next scheduling period exceeds the transformer safety temperature parameter constant, the difference between the winding hot spot temperature value and the transformer safety temperature parameter constant is calculated, and this difference is used as the overrun temperature difference value. Based on the over-temperature difference value and the preset mapping gain constant, the penalty increment multiplier is calculated using an exponential mapping method, so that the penalty increment multiplier increases as the over-temperature difference value increases; The upper limit-side secondary penalty coefficient and upper limit-side primary penalty coefficient used when the planned load value exceeds the upper limit of the allowable fluctuation range are multiplied by the penalty increment multiplier, respectively, to increase the consistency penalty intensity when the planned load value exceeds the upper limit of the allowable fluctuation range. When the planned load value is lower than the lower limit of the allowable fluctuation range, the secondary penalty coefficient and the primary penalty coefficient on the lower limit side will not be scaled or adjusted, and will remain unchanged.

[0077] As one possible implementation, the optimal load dispatching scheme is obtained by solving the objective function in conjunction with the power flow constraints and capacity constraints of the prefabricated substation, including: Based on the safe operating range of voltage at each node of the system, the maximum transmission capacity of each substation branch, and the power balance relationship, power flow constraints are established. Based on the rated capacity indicated on the transformer nameplate, the planned apparent power output of the transformer during dispatching and operation shall not exceed its maximum allowable output, thus establishing a capacity constraint condition. The objective function, power flow constraints, and capacity constraints are combined to establish an augmented Lagrangian function that includes an optimization variable vector, equality constraint multipliers, and inequality constraint multipliers. During the iterative solution process, the optimization variable vector is updated successively based on the first-order gradient information and second-order partial derivative information of the augmented Lagrange function, and the corresponding constraint multipliers are updated synchronously. After each iteration update, the L2 norm of the current gradient vector is calculated and compared with the preset convergence tolerance. When the L2 norm is less than the preset convergence tolerance, the iteration is stopped and the current solution is determined to be the optimal solution after convergence. The optimal solution is converted into a sequence of control commands corresponding to distributed power sources, energy storage devices and adjustable loads within the prefabricated substation, and the sequence of control commands is issued as the optimal load scheduling scheme.

[0078] The above are merely preferred embodiments of the present invention and are not intended to limit the present invention in any other way. Any person skilled in the art may make changes or modifications to the above-disclosed technical content to create equivalent embodiments that can be applied to other fields. However, any simple modifications, equivalent changes, and modifications made to the above embodiments based on the technical essence of the present invention without departing from the scope of the present invention shall still fall within the protection scope of the present invention.

Claims

1. A load dispatching method for a prefabricated substation, characterized in that, Includes the following steps: Acquire historical load data, transformer status data, and renewable energy output forecast data for the waiting period of the prefabricated substation, and construct an objective function that includes a power generation cost term and a consistency penalty term. Based on load data from multiple historical moments, the historical load projection benchmark and weighted dispersion for the current time to be dispatched are calculated. According to the interval width of the renewable energy output forecast data and the weighted dispersion, an allowable fluctuation interval centered on the historical load projection benchmark is determined. When the planned load value at the time to be dispatched exceeds the allowable fluctuation interval, the difference between the planned load value and the nearest endpoint of the allowable fluctuation interval is taken as the independent variable, and a consistency penalty value is calculated based on the piecewise penalty function. When the planned load value is within the allowable fluctuation interval, the consistency penalty value is zero. Based on the transformer status data, the winding hot spot temperature for the next scheduling period is predicted. Based on the winding hot spot temperature, the piecewise penalty function is asymmetrically adjusted. Based on the objective function, and combined with the power flow constraints and capacity constraints of the prefabricated substation, the optimal load scheduling scheme is obtained.

2. The method according to claim 1, characterized in that, The acquisition of historical load data, transformer status data, and renewable energy output forecast data for the period to be dispatched from the prefabricated substation includes: Read the actual active power values ​​at the same time and the adjacent previous time within the past 30 days from the database of the data acquisition and monitoring control system as historical load data; Sensors installed in the prefabricated substation are used to collect real-time data on the transformer top oil temperature, ambient temperature, and three-phase load current as transformer status data. Extract the predicted output power of wind power generation, the predicted output power of photovoltaic power generation, and their corresponding prediction intervals for each scheduling time within the scheduling period. Determine the predicted output power of renewable energy based on the sum of the predicted output power of wind power generation and the predicted output power of photovoltaic power generation. Determine the predicted output interval of renewable energy based on the sum of the upper and lower limits of the prediction intervals of the two.

3. The method according to claim 1, characterized in that, The calculation of the historical load projection benchmark and weighted dispersion at the current time to be scheduled, based on load data from multiple historical moments, includes: Calculate the absolute value of the difference between the load value of each historical sample at the current time to be scheduled and the load value at the adjacent previous time, and obtain the first parameter representing the load change amplitude of the historical sample. Add the first parameter to the preset minimum normal number and take the reciprocal to obtain the initial weight of the corresponding historical sample. Divide the initial weight by the sum of the initial weights of all historical samples to obtain the normalized weight corresponding to the load value of each historical sample. The historical load projection benchmark for the current time to be scheduled is obtained by weighted summation based on the load values ​​of each historical sample and their corresponding normalized weights. The squared deviation between each historical sample load value and the historical load projection benchmark is calculated, and the squared deviation is multiplied by the corresponding normalized weight to obtain the discrete term. The weighted dispersion, which represents the stability of the historical load pattern, is obtained by summing all the discrete terms and taking the square root.

4. The method according to claim 3, characterized in that, The step of determining an allowable fluctuation range centered on the historical load projection benchmark by combining the interval width of the renewable energy output forecast data and the weighted dispersion includes: The interval width is obtained by subtracting the lower limit of the prediction interval from the upper limit of the prediction interval at the same time to be dispatched in the renewable energy output prediction data. Multiply the interval width by the first weighting parameter to obtain the first width expansion amount; The weighted dispersion is multiplied by the second weighting parameter to obtain the second broadening amount; The first width expansion amount and the second width expansion amount are added together to calculate the range expansion radius of the allowable fluctuation range; The upper limit of the allowable fluctuation range is obtained by adding the interval expansion radius value to the historical load projection benchmark. The lower limit of the allowable fluctuation range is obtained by subtracting the interval expansion radius value from the historical load projection benchmark.

5. The method according to claim 1, characterized in that, When the planned load value at the time to be scheduled exceeds the allowable fluctuation range, the difference between the planned load value and the nearest endpoint of the allowable fluctuation range is taken as the independent variable, and a consistency penalty value is calculated based on the piecewise penalty function, including: When the planned load value is greater than the upper limit of the allowable fluctuation range, first calculate the difference between the planned load value and the upper limit of the allowable fluctuation range, and use this difference as the upper limit over-limit difference; then, according to the preset upper limit side secondary penalty coefficient and upper limit side primary penalty coefficient, perform secondary term penalty calculation and primary term penalty calculation on the upper limit over-limit difference respectively, and add the two to obtain the consistency penalty value when it is higher than the upper limit of the allowable fluctuation range. When the planned load value is less than the lower limit of the allowable fluctuation range, first calculate the difference between the lower limit of the allowable fluctuation range and the planned load value, and use this difference as the lower limit over-limit difference; then, according to the preset lower limit side secondary penalty coefficient and lower limit side primary penalty coefficient, perform secondary term penalty calculation and primary term penalty calculation on the lower limit over-limit difference respectively, and add the two to obtain the consistency penalty value when it is lower than the lower limit of the allowable fluctuation range.

6. The method according to claim 1, characterized in that, The step of predicting the winding hot spot temperature for the next scheduling period based on the transformer status data includes: The current load rate is determined by the ratio of the effective value of the three-phase load current to the corresponding rated current value at the current moment. Based on the current load rate, the preset rated hot spot temperature rise parameter and the preset hot spot temperature rise index, calculate the current winding hot spot temperature difference and add it to the current transformer top oil temperature value to obtain the initial winding hot spot temperature. The steady-state limit temperature of the winding is calculated based on the ambient temperature value, the preset loss ratio constant, the preset rated top oil temperature rise parameter, the preset rated hot spot temperature rise parameter, and the load rate used for steady-state temperature rise prediction. The load rate used for steady-state temperature rise prediction is the current load rate during real-time state prediction and the planned load rate corresponding to the candidate scheduling scheme during the optimization solution process. The thermal decay factor is calculated based on the scheduling delay step and the thermal time constant, and the winding hot spot temperature value at the next scheduling moment is obtained based on the initial winding hot spot temperature, the winding steady-state limit temperature and the thermal decay factor.

7. The method according to claim 1, characterized in that, The asymmetric adjustment of the piecewise penalty function based on the winding hot spot temperature includes: When the predicted winding hot spot temperature value for the next scheduling period exceeds the transformer safety temperature parameter constant, the difference between the winding hot spot temperature value and the transformer safety temperature parameter constant is calculated, and this difference is used as the overrun temperature difference value. Based on the over-temperature difference value and the preset mapping gain constant, the penalty increment multiplier is calculated using an exponential mapping method, so that the penalty increment multiplier increases as the over-temperature difference value increases; The upper limit-side secondary penalty coefficient and upper limit-side primary penalty coefficient used when the planned load value exceeds the upper limit of the allowable fluctuation range are multiplied by the penalty increment multiplier, respectively, to increase the consistency penalty intensity when the planned load value exceeds the upper limit of the allowable fluctuation range. When the planned load value is lower than the lower limit of the allowable fluctuation range, the secondary penalty coefficient and the primary penalty coefficient on the lower limit side will not be scaled or adjusted, and will remain unchanged.

8. The method according to claim 1, characterized in that, The optimal load dispatching scheme is obtained by solving the objective function in conjunction with the power flow constraints and capacity constraints of the prefabricated substation, including: Based on the safe operating range of voltage at each node of the system, the maximum transmission capacity of each substation branch, and the power balance relationship, power flow constraints are established. Based on the rated capacity indicated on the transformer nameplate, the planned apparent power output of the transformer during dispatching and operation shall not exceed its maximum allowable output, thus establishing a capacity constraint condition. The objective function, power flow constraints, and capacity constraints are combined to establish an augmented Lagrangian function that includes an optimization variable vector, equality constraint multipliers, and inequality constraint multipliers. During the iterative solution process, the optimization variable vector is updated successively based on the first-order gradient information and second-order partial derivative information of the augmented Lagrange function, and the corresponding constraint multipliers are updated synchronously. After each iteration update, the L2 norm of the current gradient vector is calculated and compared with the preset convergence tolerance. When the L2 norm is less than the preset convergence tolerance, the iteration is stopped and the current solution is determined to be the optimal solution after convergence. The optimal solution is converted into a sequence of control commands corresponding to distributed power sources, energy storage devices and adjustable loads within the prefabricated substation, and the sequence of control commands is issued as the optimal load scheduling scheme.

9. A load dispatching system for a prefabricated substation, characterized in that, Includes the following modules: A module is built to acquire historical load data, transformer status data, and renewable energy output forecast data for the period to be dispatched from the prefabricated substation, and to build an objective function that includes a power generation cost term and a consistency penalty term. The calculation module is used to calculate the historical load projection benchmark and weighted dispersion for the current time to be dispatched based on load data from multiple historical moments. Based on the interval width of the renewable energy output forecast data and the weighted dispersion, an allowable fluctuation interval centered on the historical load projection benchmark is jointly determined. When the planned load value for the time to be dispatched exceeds the allowable fluctuation interval, the difference between the planned load value and the nearest endpoint of the allowable fluctuation interval is taken as the independent variable, and a consistency penalty value is calculated based on a piecewise penalty function. When the planned load value is within the allowable fluctuation interval, the consistency penalty value is zero. The scheduling module is used to predict the winding hot spot temperature of the next scheduling period based on the transformer status data, perform asymmetric adjustment on the segmented penalty function based on the winding hot spot temperature, and solve the optimal load scheduling scheme based on the objective function and in combination with the power flow constraints and capacity constraints of the box-type substation.

10. The system according to claim 9, characterized in that, The acquisition of historical load data, transformer status data, and renewable energy output forecast data for the period to be dispatched from the prefabricated substation includes: Read the actual active power values ​​at the same time and the adjacent previous time within the past 30 days from the database of the data acquisition and monitoring control system as historical load data; Sensors installed in the prefabricated substation are used to collect real-time data on the transformer top oil temperature, ambient temperature, and three-phase load current as transformer status data. Extract the predicted output power of wind power generation, the predicted output power of photovoltaic power generation, and their corresponding prediction intervals for each scheduling time within the scheduling period. Determine the predicted output power of renewable energy based on the sum of the predicted output power of wind power generation and the predicted output power of photovoltaic power generation. Determine the predicted output interval of renewable energy based on the sum of the upper and lower limits of the prediction intervals of the two.