Air conditioning type virtual power plant day-ahead scheduling method considering temperature prediction error

Abstract

This invention discloses a day-ahead scheduling method for an air-conditioning virtual power plant considering temperature prediction errors. The method includes: acquiring input information for virtual energy storage modeling of the air conditioner; constructing a virtual energy storage model and calculating the predicted baseline power and the predicted virtual energy storage state transition equation for each time period; fitting the time-series recursive relationship of the error based on historical temperature prediction error data, converting the temperature prediction error into a virtual energy storage state deviation and propagating it hourly along the state transition equation; generating corrected time-sharing safe operating boundaries for each time period based on a preset maximum allowable exceedance probability; incorporating the predicted baseline power, state transition equation, and safety boundaries, along with the safe operation constraints of the distribution network within the virtual power plant, into a distribution network optimization scheduling model, solving for the minimum operating cost, generating time-sharing power adjustment commands for each air-conditioning device, and issuing them for execution. This invention achieves continuous transmission of temperature prediction errors to the air-conditioning operating state, and coordinated optimization of air-conditioning adjustment plans and distribution network safety.

Description

Technical Field

[0001] This invention belongs to the field of new power system dispatch and control technology, and specifically relates to a day-ahead dispatch method for an air-conditioned virtual power plant that takes into account temperature prediction errors. Background Technology

[0002] With the continuous expansion of installed capacity of new energy power generation, its output volatility and forecasting deviations place higher demands on the power balance and safe operation of the power system. The power system urgently needs to develop demand-side flexible resources with fast response speeds, low regulation costs, and the ability to continuously participate in operation control. Air conditioning features large load capacity, concentrated operating periods, and rapid regulation response. It can also utilize the thermal inertia of building envelopes to shift power consumption during specific time periods while maintaining indoor temperatures within the user's permissible range. Virtual power plants can centrally manage air conditioning equipment, distributed photovoltaic systems, and baseloads distributed across different nodes, enabling demand-side resources to participate in peak shaving and valley filling, power balance, and operating cost optimization. This plays a crucial role in improving the operational flexibility of the new power system and the local consumption capacity of new energy sources.

[0003] Against this backdrop, the key to enabling air conditioning loads to truly participate in the day-ahead dispatch of the virtual power plant lies in transforming their temperature-dependent electricity demand into calculable, constrained, and executable adjustment capabilities. Air conditioning loads maintain indoor temperature by consuming electrical energy, and their power regulation is not entirely free; it is constrained by both building thermal inertia and the allowable range of indoor temperature. Utilizing the building's thermal storage characteristics, indoor temperature changes can be equivalently represented as the virtual energy storage state of the air conditioning system. The power required to maintain indoor thermal balance is used as the baseline power, and adjustments to increase or decrease electricity consumption are planned based on this baseline power. However, the baseline power is typically calculated based on day-ahead outdoor temperature forecasts. When the actual temperature deviates from the forecast value, the actual cooling demand of the air conditioning changes accordingly, causing the virtual energy storage state to deviate from the planned value. Because building temperature states have a coupling characteristic between different time periods, temperature prediction errors generated in a certain period not only affect the current period but also continue to affect subsequent periods along the state change process, thereby altering the safe adjustment space that air conditioning can provide. Therefore, accurately characterizing the cumulative temporal impact of temperature prediction errors on the air conditioning state is a crucial foundation for ensuring the safe execution of day-ahead adjustment plans.

[0004] Currently, air-conditioning virtual power plant dispatching methods mainly include economic dispatching methods based on deterministic temperature forecasts, robust or stochastic dispatching methods oriented towards system uncertainties such as electricity prices and loads, and methods that describe air-conditioning regulation capabilities through adjustable power boundaries or stochastic capacity ranges. Deterministic dispatching methods typically calculate the air-conditioning baseline power directly based on day-ahead temperature forecasts, without fully considering the impact of actual temperature deviations on the feasibility of the dispatch plan. System-level uncertain dispatching methods often simplify flexible resources such as air conditioning into equivalent units with fixed regulation capacities, making it difficult to reflect the continuous state changes caused by building thermal inertia. Methods based on regulation boundaries mainly describe air-conditioning flexibility from the perspective of external power ranges, lacking direct calculation of the hourly propagation process of temperature errors through baseline power offsets to the virtual energy storage state. Since meteorological forecast errors usually have significant temporal correlations, state deviations generated in previous periods may continue to affect subsequent periods. Using static boundaries or mutually independent random errors can easily underestimate the cumulative degree of state deviations, leading to overly optimistic day-ahead regulation plans and indoor temperature exceeding limits during actual execution. Furthermore, since each air conditioner is distributed across different nodes in the internal power distribution network of the virtual power plant, its power adjustment will also change the node load, branch power flow, and node voltage. Existing methods cannot simultaneously guarantee the executability of building air conditioning adjustment commands, the operational safety of the power distribution network, and the operational economy of the virtual power plant. Summary of the Invention

[0005] To address the aforementioned technical challenges, this invention proposes a day-ahead dispatching method for air-conditioned virtual power plants that considers temperature prediction errors. This provides a feasible technical path for air conditioning to participate in peak shaving and valley filling, load shifting, and coordinated operation of the distribution network. It has significant engineering application value and promotional potential for improving the refined control of flexible resources on the demand side, supporting the safe and economical operation of virtual power plants, and promoting the large-scale development and utilization of air conditioning resources.

[0006] To achieve the above objectives, the present invention adopts the following technical solution: In a first aspect, the present invention proposes a day-ahead dispatching method for air-conditioning type virtual power plants that considers temperature prediction errors, comprising the following steps: The system obtains the virtual energy storage modeling input information required for air conditioning operation, constructs an air conditioning virtual energy storage model based on the virtual energy storage modeling input information, and calculates the predicted baseline power and predicted virtual energy storage state transition equations for each time period. The input information includes building thermal parameters, equipment operating parameters, temperature control requirements, and day-ahead outdoor temperature prediction data. Based on historical temperature prediction error data, the time-series recursive relationship of temperature prediction error is fitted, the temperature prediction error is converted into virtual energy storage state deviation and propagated hourly along the virtual energy storage state transition equation, and the time-sharing safe operation boundary after correction for each time period is generated according to the preset maximum allowable over-limit probability. The predicted baseline power, the predicted virtual energy storage state transition equation, the time-sharing safe operation boundary, and the safe operation constraints of the distribution network within the virtual power plant are all incorporated into the distribution network optimization scheduling model. The distribution network optimization scheduling model is solved with the goal of minimizing the operating cost of the virtual power plant, and time-sharing power adjustment commands for each air conditioning device are generated and issued for execution.

[0007] Furthermore, the detailed process of issuing and executing the command is as follows: The power plans for each resource are organized according to the equipment number, access node, and scheduling period, and the corresponding control commands are sent to the corresponding execution units; the execution units include air conditioning equipment, distributed photovoltaic resources, and the upstream power grid. For air conditioning equipment, the compressor operating frequency is adjusted or the set temperature is changed to ensure that each air conditioning unit operates according to the optimized time-sharing target power. For distributed photovoltaic resources, the power is reduced by adjusting the inverter output setpoint to ensure that it operates according to the optimized time-sharing plan. For power interaction with the upper-level power grid, the amount of electricity purchased from the upper-level power grid in each time period is adjusted according to the optimized time-sharing power purchase plan, and the operation of internal resources is coordinated to ensure that the actual exchange power of the tie line matches the power purchase plan.

[0008] Furthermore, based on the virtual energy storage modeling input information, an air conditioning virtual energy storage model is constructed, specifically as follows: Obtain the building equivalent thermal resistance, building equivalent heat capacity, cooling energy efficiency ratio, upper and lower limits of allowable operating power, indoor temperature setpoint, and upper and lower limits of allowable indoor temperature for each air conditioning unit. Based on the first-order equivalent thermal parameter model, the product of the building's equivalent heat capacity and the rate of change of indoor temperature is equal to the difference between outdoor temperature and indoor temperature divided by the building's equivalent thermal resistance, and then the product of the cooling energy efficiency ratio and the power consumption of air conditioning is subtracted to establish the continuous time state transition relationship of indoor temperature. The continuous-time state transition relationship of indoor temperature is discretized according to the day-ahead scheduling step size, and the virtual energy storage state is defined by the position of indoor temperature relative to the temperature allowable range, and the virtual energy storage state transition equation of air conditioning is established.

[0009] Furthermore, the predicted baseline power and the predicted virtual energy storage state transition equations for each time period are calculated, specifically as follows: Historical outdoor temperature data is acquired, and a day-ahead temperature prediction model is established using regression fitting methods to generate a day-ahead outdoor temperature prediction sequence for the cost scheduling cycle. Substitute the predicted outdoor temperature values ​​for each time period into the baseline power function, and calculate the predicted baseline power for each time period hour by hour. The baseline power function is expressed as the product of the difference between the outdoor temperature and the indoor set temperature divided by the cooling energy efficiency ratio and the building's equivalent thermal resistance. Substitute the predicted baseline power for each time period into the virtual energy storage state transition equation to form the virtual energy storage state transition equation under the day-ahead forecast conditions, and calculate the initial value of the virtual energy storage state based on the measured indoor temperature at the start of dispatch.

[0010] Furthermore, before fitting the time-series recursive relationship of temperature prediction error based on historical temperature prediction error data, the following steps are also included: The daytime outdoor temperature forecast data and the corresponding actual outdoor temperature data for multiple consecutive historical dates in the area where the air conditioner is located are aligned according to the scheduling time interval. Let the number of historical sample dates be . Each scheduling cycle contains In a certain period of time, Indicates the first Historical dates Outdoor temperature forecast for the period of time, in Indicates the first Historical dates Actual outdoor temperature value for the time period, historical temperature prediction error Represented as: ; Calculate the sample mean of all historical temperature prediction errors.

[0011] ; The sample mean is used to correct the deviation of the daytime temperature forecast curve, resulting in the corrected daytime outdoor temperature forecast and a temperature forecast residual with a mean close to zero. ; ; in, This represents the revised day-ahead outdoor temperature forecast. This represents the residual of historically centered temperature prediction.

[0012] Furthermore, a time-series recursive relationship for temperature prediction errors is fitted based on historical temperature prediction error data, specifically as follows: Based on historical error samples between consecutive scheduling periods, a first-order autoregressive process is used to establish the state recursive relationship of temperature prediction error: ; in, Indicates the first Zero-mean temperature prediction error for a given period; Indicates the first Zero-mean temperature prediction error for a given period; This represents the autoregressive coefficient of temperature prediction error, and ; Indicates random perturbation; Temperature prediction error autoregressive coefficient Represented as: ; in, Indicates the first The first historical date The centered temperature prediction residual for the time period, i.e. the temperature prediction residual with a mean close to zero; Indicates the first The first historical date Centralized temperature prediction residuals for different time periods; The calculated Substituting into the first-order autoregressive process, we obtain the random perturbation residuals for each historical period. : ; Then calculate the variance of random disturbances. : ; Based on the autoregressive coefficients and the steady-state variance of the temperature prediction error, calculate the autocovariance among the temperature prediction errors for different scheduling periods: ; ; in, Indicates the first time; Indicates the first time; This represents the steady-state variance of the temperature error.

[0013] Furthermore, the temperature prediction error is converted into a virtual energy storage state deviation and propagated hourly along the virtual energy storage state transition equation, specifically as follows: The current scheduling day The actual outdoor temperature is obtained by superimposing the predicted outdoor temperature for the specified time period with the temperature prediction error. ; in, Indicates the first Actual outdoor temperature at the time of the event; Indicates the first Predicted outdoor temperature for the specified time period; Indicates the first Error in predicting zero-mean temperature over a given period; Substituting the actual outdoor temperature into the baseline power function, we obtain the actual baseline power: ; in, Indicates the first Actual baseline power for the time period; Indicates the indoor temperature setting; Indicates the cooling energy efficiency ratio; Indicates the building's equivalent thermal resistance; Indicates the first Time period predicted baseline power; Indicates the temperature-power sensitivity coefficient; Substituting the actual baseline power into the virtual energy storage state transition equation, we obtain the actual virtual energy storage state: ; in, Indicates the first Actual virtual energy storage status during the time period; Indicates the first Actual virtual energy storage status during the time period; Indicates the state retention coefficient; Indicates the electrical energy conversion coefficient; Indicates the first Air conditioning power consumption during different time periods; Indicates the reference bias constant; By combining the actual virtual energy storage state with the virtual energy storage state under the predicted conditions, the virtual energy storage state deviation is calculated: ; in, Indicates the first Virtual energy storage status deviation over time period; Indicates the first Virtual energy storage status deviation over time period; Will Timeline recursion expansion: ; Since the temperature prediction error follows a normal distribution, the actual virtual energy storage state also follows a normal distribution, i.e. ; in, Indicates the first The expected value of the actual virtual energy storage state during the time period; Indicates the first Variance of virtual energy storage state deviation over time period ; in, Indicates the first Virtual energy storage status deviation over time period; Indicates the first Each historical period number; Indicates the first Zero-mean temperature prediction error for a given period; Indicates the first Zero-mean temperature prediction error for a given period; Indicates the first Time period and the Covariance among time-period temperature prediction errors; Formula Substituting into the above equation, we can obtain the final analytical expression for the state variance: ; in, This represents the steady-state variance of the temperature prediction error.

[0014] Furthermore, based on the preset maximum permissible probability of exceeding limits, the corrected time-sharing safe operation boundaries for each time period are generated, specifically as follows: Set the maximum permissible probability of a single side exceeding the upper limit or falling below the lower limit for the actual virtual energy storage state of the air conditioner. Establish opportunity constraints for the upper limit of the state and the lower limit of the state respectively: ; in, Indicates the probability of an event occurring; Indicates the upper limit of the virtual energy storage state; Indicates the lower limit of the virtual energy storage state; The chance constraint is transformed into a deterministic linear constraint using the inverse cumulative distribution function of the standard normal distribution: ; in, Indicates the first Time Period Prediction Virtual Energy Storage Status ; Let be the inverse cumulative distribution function of the standard normal distribution, and let... To define a safe multiplier corresponding to the maximum allowable probability of exceeding the limit, we can define: Dynamic safety margin for different time periods for: ; The revised time-sharing safety state boundary is represented as follows: ; in, Indicates the first Dynamic safety margin over time period; Indicates the first Predict the virtual energy storage status over a specific time period.

[0015] Furthermore, the objective function of the power distribution network optimization scheduling model is to minimize the sum of the cost of purchasing electricity from the upstream power grid and the cost of regulating air conditioning load. The constraints of the power distribution network optimization scheduling model include: Tie line power constraint: The active power exchanged between the virtual power plant and the upstream power grid shall not exceed the tie line capacity limit; Distributed photovoltaic (PV) operation constraints: The actual active power output of PV shall not exceed the maximum power output predicted before the day, and the reactive power output shall be limited by the inverter capacity; Air conditioning operation constraints: The actual dispatch power of each air conditioner is between the upper and lower limits of the allowable power, and the corresponding virtual energy storage state is within the corrected time-sharing safety state boundary; Node power balance constraint: The active and reactive power injected into each node are equal to the sum of the adjustable loads, non-adjustable loads, photovoltaic output, and power injected into the upstream grid connected to that node; Power flow constraints: AC power flow equations are used to describe the relationship between active and reactive power of branches and node voltage magnitude and phase angle difference; Distribution network safety constraints: The transmission power of each branch shall not exceed the line capacity, and the voltage amplitude of each node shall be within the allowable upper and lower limits.

[0016] Furthermore, the distribution network optimization scheduling model is solved with the objective of minimizing the operating cost of the virtual power plant, specifically as follows: The nonlinear terms in the distribution network optimization scheduling model are linearized. The linearization process includes: introducing non-negative auxiliary variables to linearize the absolute value terms in the objective function, using the double circumscribed square method to linearize the apparent power capacity constraint, and using the improved DC power flow method to linearize the AC power flow constraint. The optimization solver is invoked to solve the linearized model, and the network loss factor is updated iteratively until the average absolute error rate of branch power converges to the preset accuracy threshold, thereby obtaining the time-of-use power plan, distributed photovoltaic power output plan and upstream power grid purchase plan for each air conditioning unit.

[0017] The effects described in the invention are merely those of the embodiments, and not all the effects of the invention. One of the above technical solutions has the following advantages or beneficial effects: 1. Construct a virtual energy storage model to realize the quantitative and dispatchable thermal dynamics of air conditioning. This invention constructs a virtual energy storage model for air conditioning based on a first-order equivalent thermal parameter model. It transforms the thermal inertia of the building envelope and indoor temperature changes into virtual energy storage state variables with clear physical meaning. This allows the heat storage characteristics of the air conditioning load to be directly embedded into the optimal scheduling model in the form of state equations, overcoming the deficiency in traditional methods where the thermal dynamics of air conditioning are difficult to quantify and describe by the scheduling model. The constructed virtual energy storage state transition equation accurately characterizes the temporal coupling relationship between air conditioning power, outdoor temperature, and indoor temperature, providing a precise mathematical basis for subsequent temperature error propagation calculations and day-ahead optimal scheduling. This significantly improves the computability and schedulability of the flexible adjustment capability of the air conditioning load.

[0018] 2. Characterize the temporal correlation of errors to avoid the systematic underestimation of uncertainty. This invention employs a first-order autoregressive model to fit the temporal recursive relationship of temperature forecast errors and utilizes the least squares method to identify autoregressive coefficients and random disturbance variances, accurately capturing the correlation characteristics of meteorological forecast errors between adjacent scheduling periods. Compared to existing technologies that treat forecast errors in different time periods as independent, this invention fully considers the temporal cumulative effect of temperature errors, avoiding the problem of overly optimistic day-ahead adjustment plans due to underestimating error persistence, and making the description of temperature uncertainty more consistent with the statistical laws of actual meteorological errors.

[0019] 3. Analyze the cumulative impact of propagation state deviation and quantification error on air conditioning status. This invention derives an analytical recursive expression for the virtual energy storage state deviation by propagating temperature prediction errors hourly along the virtual energy storage state transition equation. Based on the covariance structure of the temperature error, it accurately calculates the variance of the virtual energy storage state deviation for each time period. This technique directly quantifies the cumulative impact of historical temperature prediction errors on the current air conditioning operating state, enabling a fully analytical process of temperature uncertainty transmission from meteorological forecasting to air conditioning operation. It obtains the statistical characteristics of the state deviation without relying on extensive scenario sampling, exhibiting high computational efficiency and good numerical stability, thus providing a theoretical basis for the reliability assessment of air conditioning regulation capabilities.

[0020] 4. Construct dynamic security boundaries to achieve a flexible balance between economy and security. This invention transforms the opportunity constraint of the virtual energy storage state into a deterministic linear constraint by setting a maximum permissible exceedance probability and utilizing the inverse cumulative distribution function of the standard normal distribution. The quantile of this inverse cumulative distribution function at a given exceedance probability is used as the safety multiplier, and the product of the safety multiplier and the standard deviation of the virtual energy storage state defines the dynamic safety margin, generating a modified time-sharing safety boundary. This dynamic safety boundary can adaptively adjust based on the cumulative variance of temperature prediction errors in different scheduling periods. It automatically shrinks the available adjustment range during periods with large cumulative variance and allows for more full utilization of air conditioning regulation capacity during periods with small cumulative variance, thus achieving flexible configuration of air conditioning regulation plans under different risk appetite levels. Compared to traditional methods using fixed safety margins, this invention avoids both the waste of regulation capacity due to excessive conservatism and the indoor temperature exceedance caused by excessive optimism, effectively balancing the economy and safety of day-ahead dispatching of virtual power plants.

[0021] 5. Collaboratively optimize network constraints to ensure the executability of scheduling plans at the network level. This invention integrates the state transition equations of the air conditioning virtual energy storage, the time-sharing dynamic safety boundary, and the node power balance constraints, power flow constraints, node voltage constraints, branch capacity constraints, distributed photovoltaic operation constraints, and tie-line power constraints of the internal power distribution network into a power distribution network optimization scheduling model. This achieves coordinated optimization of air conditioning load regulation plans and power distribution network operation safety. This technique overcomes the shortcomings of existing methods where air conditioning regulation capacity calculations and network operation constraints are separated. It ensures that day-ahead scheduling results meet both building indoor temperature control requirements and the network security requirements of the virtual power plant's internal power distribution network, while also guaranteeing the executability of time-sharing power commands for each air conditioning unit at the network level. This avoids problems such as voltage exceeding limits or line overloads during actual execution of the scheduling plan due to neglecting network constraints.

[0022] 6. Implement efficient linearization solutions to enhance engineering practicality and promotional value. This invention linearizes the absolute value term, apparent power capacity constraint, and AC power flow constraint in the objective function before calling an optimization solver. This results in a distribution network optimization scheduling model with high computational efficiency and reliable convergence performance. It can directly generate and execute time-of-use power commands for each air conditioning unit, distributed photovoltaic power output plans, and upstream power grid purchase plans, demonstrating good engineering practicality and scalability. Simulation results show that the day-ahead scheduling plan generated using this invention can strictly control the actual exceedance rate within the preset allowable range under different risk control requirements. Furthermore, the statistical results of the exceedance rate highly match the theoretical set values, verifying the effectiveness and reliability of the proposed method in terms of risk control accuracy.

[0023] 7. Summary of Overall Benefits In summary, this invention significantly improves the availability and reliability of flexible load regulation for air conditioning while ensuring indoor temperature comfort and the security of the distribution network within the virtual power plant. It also enhances the feasibility and economic efficiency of day-ahead planning within the virtual power plant. This invention provides a feasible technical path for air conditioning to participate in peak shaving, load shifting, and coordinated operation of the distribution network. It has significant engineering application value and promotional potential for improving the refined control of flexible demand-side resources, supporting the safe and economical operation of virtual power plants, and promoting the large-scale development and utilization of air conditioning resources. Attached Figure Description

[0024] Figure 1 This is a flowchart of a day-ahead dispatching method for an air-conditioning type virtual power plant that considers temperature prediction errors, as proposed in Embodiment 1 of the present invention. Figure 2 The flowchart of air conditioner model construction and baseline power calculation based on virtual energy storage proposed in Embodiment 1 of this invention; Figure 3 This is a flowchart of the time-sharing safety boundary generation method for air conditioning based on temperature error time-series propagation proposed in Embodiment 1 of the present invention; Figure 4 This is a flowchart of the day-ahead scheduling and execution process of a virtual power plant based on internal power distribution network security constraints, as proposed in Embodiment 1 of the present invention. Figure 5 This is a topology diagram of the IEEE 33-node system in an application example of the present invention; Figure 6 This is a baseline power curve for air conditioning prediction in an application example of the present invention; Figure 7 This is a diagram showing the day-ahead air conditioning scheduling results in an application example of the present invention; Figure 8 This is a collection of outdoor temperature curves from an application example of the present invention; Figure 9 This is a Monte Carlo trajectory evolution diagram of AC1 SOC in an application example of the present invention; Figure 10 This is a schematic diagram of a day-ahead dispatching device for an air-conditioning type virtual power plant that takes into account temperature prediction errors, according to Embodiment 2 of the present invention. Detailed Implementation

[0025] To clearly illustrate the technical features of this solution, the invention will be described in detail below through specific embodiments and in conjunction with the accompanying drawings. The following disclosure provides many different embodiments or examples for implementing different structures of the invention. To simplify the disclosure of the invention, components and arrangements of specific examples are described below. Furthermore, reference numerals and / or letters may be repeated in different examples. This repetition is for simplification and clarity and does not in itself indicate a relationship between the various embodiments and / or arrangements discussed. It should be noted that the components illustrated in the drawings are not necessarily drawn to scale. Descriptions of well-known components, processing techniques, and processes are omitted in this invention to avoid unnecessarily limiting the invention.

[0026] Example 1 The present invention provides a day-ahead scheduling method for air-conditioned virtual power plants that considers temperature prediction errors. This method addresses the technical problems in existing day-ahead scheduling of air-conditioned virtual power plants, such as the difficulty in transmitting outdoor temperature prediction errors to the air conditioning operation status hourly along the building's thermal inertia process, the lack of coordinated optimization between the air conditioning regulation plan and the safe operation of the internal power distribution network, which leads to a high risk of indoor temperature exceeding limits and poor network executability when the day-ahead scheduling results are actually executed.

[0027] This invention, in Embodiment 1, proposes a day-ahead scheduling method for air-conditioned virtual power plants considering temperature prediction errors. This method is applicable to day-ahead scheduling within air-conditioned virtual power plants that consider temperature uncertainties. It mainly comprises three parts: air-conditioned model construction and baseline power calculation based on virtual energy storage; generation of time-sharing safety boundaries for air-conditioned systems based on the time-series propagation of temperature errors; and day-ahead scheduling and execution of the virtual power plant based on internal power distribution network security constraints. The first part investigates the thermal parameters, equipment parameters, and temperature operation requirements of the buildings corresponding to the air-conditioned systems, establishing a dynamic relationship between outdoor temperature, air-conditioned power consumption, and indoor temperature. The air-conditioned systems are converted into virtual energy storage resources with state and power constraints, and the baseline power for each time period is calculated based on the day-ahead outdoor temperature prediction. The second part uses historical temperature prediction and actual values ​​to characterize the time-series variation of temperature prediction errors, converting these errors into air-conditioned baseline power offsets. The cumulative impact of these offsets on the operating state of subsequent time periods is calculated along the virtual energy storage state transition process, and the safety operating boundaries for each time period are generated based on a preset maximum permissible exceedance probability. The third part investigates the operating parameters of the virtual power plant's internal distribution network, base load, distributed photovoltaic, electricity price, and tie lines. Taking the operating cost of the virtual power plant as the optimization objective, the safe operating boundary, along with air conditioning operating constraints, node power balance constraints, power flow constraints, distribution network safety constraints, and tie line power constraints, are incorporated into the day-ahead optimization model. After linearizing the model, the solution is obtained, generating time-of-use adjustment commands for each air conditioning unit and a power interaction plan between the virtual power plant and the upper-level power grid. The adjustment commands are then sent to the corresponding air conditioning control systems for execution.

[0028] Figure 1 This is a flowchart of a day-ahead dispatching method for an air-conditioning type virtual power plant that considers temperature prediction errors, as proposed in Embodiment 1 of the present invention. In step 1, the virtual energy storage modeling input information required for the operation of the air conditioner is obtained, the air conditioner virtual energy storage model is constructed based on the virtual energy storage modeling input information, and the predicted baseline power and the predicted virtual energy storage state transition equation for each time period are calculated. This step first obtains the building thermal parameters, equipment operating parameters, temperature control requirements, and day-ahead outdoor temperature forecast data required for air conditioning operation. Then, based on a first-order equivalent thermal parameter model, a dynamic relationship between outdoor temperature, indoor temperature, and air conditioning power consumption is established. The virtual energy storage state of the air conditioning is defined using the allowable range of indoor temperature. Based on the indoor thermal balance relationship, a baseline power function and a complete virtual energy storage state transition model suitable for different outdoor temperature inputs are established. Finally, the day-ahead outdoor temperature forecast curve is substituted into the baseline power function, and the predicted baseline power is calculated hourly to form the virtual energy storage state transition equation under the day-ahead forecast conditions. This provides a deterministic calculation benchmark for subsequent temperature forecast error propagation, air conditioning time-sharing safety boundary generation, and day-ahead optimized scheduling of the virtual power plant.

[0029] Figure 2 The flowchart of air conditioner model construction and baseline power calculation based on virtual energy storage proposed in Embodiment 1 of this invention; Specifically, the equivalent thermal resistance, equivalent heat capacity, cooling efficiency ratio, upper and lower limits of allowable operating power, indoor temperature setpoint, and allowable upper and lower limits of indoor temperature for each air conditioning unit are obtained. Based on the first-order equivalent thermal parameter model, the product of the equivalent heat capacity and the rate of change of indoor temperature is equal to the difference between the outdoor temperature and the indoor temperature divided by the equivalent thermal resistance, and then the product of the cooling efficiency ratio and the power consumption of the air conditioning is subtracted to establish the continuous-time state transition relationship of indoor temperature. The continuous-time state transition relationship of indoor temperature is discretized according to the day-ahead scheduling step size, and the virtual energy storage state is defined by the position of the indoor temperature relative to the temperature allowable range, thus establishing the virtual energy storage state transition equation of the air conditioning system.

[0030] Step 1.1 Obtain the virtual energy storage modeling input information required for air conditioning operation; the input information includes building thermal parameters, equipment operating parameters, temperature control requirements, and day-ahead outdoor temperature forecast data.

[0031] Among them, building thermal parameters include the building's equivalent thermal resistance. Equivalent heat capacity of building It is used to characterize the heat transfer characteristics between the air-conditioned building and the outdoor environment, as well as the heat storage capacity inside the building.

[0032] Air conditioning equipment parameters include cooling energy efficiency ratio. Lower limit of permissible operating power and upper limit It is used to limit the range of electrical power adjustment of air conditioning equipment.

[0033] Temperature control parameters include the indoor target set temperature of each air conditioner. Lower limit of indoor temperature Upper limit of indoor temperature and indoor temperature during the initial scheduling period ; Further investigation was conducted into historical outdoor temperature data for the area where air conditioning units are located. This historical data includes measured outdoor temperatures for different dates and time periods within the day. Data was cleaned, missing values ​​were added, and the time scale was standardized according to the day-ahead scheduling interval. Based on this, a day-ahead outdoor temperature prediction model for the region was established using regression fitting methods. Historical outdoor temperature variation patterns, date types, and day-ahead meteorological information were used as model inputs to calculate the predicted outdoor temperature values ​​for each time period within the day-ahead scheduling cycle. .

[0034] After obtaining the above parameters, the dimensions and time resolution of the parameters are uniformly configured, and the correspondence between the air conditioning equipment number, operating parameters, and temperature prediction data is established to form the parameter input for each air conditioning virtual energy storage model.

[0035] Step 1.2 Construction of Virtual Energy Storage Model for Air Conditioning Based on Equivalent Thermal Parameter Model Based on the building equivalent thermal resistance, building equivalent heat capacity, air conditioning cooling efficiency ratio, and indoor temperature control parameters obtained in step 1.1, an equivalent thermal parameter model for the air conditioning operation process is established: (1) in, express Real-time indoor temperature; express outdoor temperature at all times; This refers to the electrical power of the device.

[0036] By precisely discretizing equation (1) according to the day-ahead scheduling step size, the discrete state transition equation for indoor temperature is obtained: (2) In the formula, Indicates the first Indoor temperature during the period; Indicates the first Indoor temperature during the period; Indicates the first Outdoor temperature at that time of day; For scheduling step size; For scheduling periods; Indicates the first Air conditioning power consumption during different time periods.

[0037] To convert the cooling capacity of the building corresponding to the air conditioning into a state variable that can be used for day-ahead optimization scheduling, a virtual energy storage state is introduced. : (3) To maintain indoor temperature in thermal equilibrium, that is... The baseline power can be obtained as: (4) In the formula, Indicates baseline power.

[0038] Substituting equations (3) and (4) into equation (2), we obtain the standard form of the state transition equation for air conditioning virtual energy storage: (5) in: (6) In the formula, Indicates the first Virtual energy storage status during a given time period; and These are the temperature state thresholds that characterize the upper and lower limits of the comfortable indoor temperature range, respectively, and are set to 1 and 0 when using the complete allowable indoor temperature range; Indicates the state retention coefficient; Indicates the electrical energy conversion coefficient; This represents the reference bias constant.

[0039] Step 1.3 Calculation of the predicted baseline power and predicted SOC state transition equations Specifically, historical outdoor temperature data is acquired, and a day-ahead temperature prediction model is established using a regression fitting method to generate a day-ahead outdoor temperature prediction sequence for the current scheduling cycle. The predicted outdoor temperature values ​​for each time period are substituted into the baseline power function, and the predicted baseline power for each time period is calculated hourly. The baseline power function is expressed as the product of the difference between the outdoor temperature and the indoor set temperature divided by the cooling energy efficiency ratio and the building's equivalent thermal resistance. The predicted baseline power for each time period is substituted into the virtual energy storage state transition equation to form the virtual energy storage state transition equation under the day-ahead prediction conditions, and the initial value of the virtual energy storage state is calculated based on the measured indoor temperature at the start of the scheduling.

[0040] The daytime outdoor temperature forecast obtained in step 1.1 Substituting the baseline power function (4) established in step 1.2, the predicted baseline power of each air conditioner under the day-ahead forecast conditions is calculated hourly: (7) In the formula, Indicates the first Forecast baseline power for the time period. Under day-ahead forecast conditions, to forecast baseline power. The general baseline power in substitution (5) This forms the predicted SOC state transition equation: (8) In the formula, This indicates the day before the outdoor temperature forecast. Predicted virtual energy storage status over a given period; This indicates the first day under the current outdoor temperature forecast conditions. The predicted virtual energy storage state for the time period is calculated based on the measured indoor temperature at the start of dispatch: (9) Thus, the predicted baseline power curves of each air conditioning unit and the SOC state transition equation under the day-ahead prediction conditions are obtained. The predicted baseline power and predicted SOC state constitute a deterministic operating benchmark without considering temperature prediction errors.

[0041] In step 2, the time-series recursive relationship of temperature prediction error is fitted based on historical temperature prediction error data. The temperature prediction error is converted into virtual energy storage state deviation and propagated hourly along the virtual energy storage state transition equation. The time-sharing safe operation boundary after correction for each time period is generated according to the preset maximum allowable over-limit probability.

[0042] This step first investigates historical outdoor temperature prediction data and corresponding actual temperature data for the region where the air conditioners are located, calculates historical prediction error samples, and identifies the recursive relationship, autoregressive coefficient, and random disturbance variance of temperature prediction errors between adjacent scheduling periods. Then, the outdoor temperature prediction errors are sequentially converted into actual baseline power offset and virtual energy storage state deviation. The cumulative impact of prediction errors on subsequent states is calculated hourly along the virtual energy storage state transition process of the air conditioners, obtaining the state deviation variance and standard deviation of each air conditioner in each scheduling period. Finally, based on the pre-set maximum allowable temperature exceedance probability, the random virtual energy storage state constraints are converted into deterministic constraints containing dynamic safety margins, generating the upper and lower limits of the corrected state and the safe adjustment range of each air conditioner in each scheduling period, and using them as inputs for the day-ahead optimization scheduling of the virtual power plant in step 3. Figure 3 This is a flowchart of the time-sharing safety boundary generation method for air conditioning based on temperature error time-series propagation proposed in Embodiment 1 of the present invention; Step 2.1 Calculation of Historical Temperature Prediction Error and Fitting of Time Series Model Parameters The daytime outdoor temperature forecast data and the corresponding actual outdoor temperature data for multiple consecutive historical dates in the area where the air conditioner is located are aligned according to the scheduling time interval. Let the number of historical sample dates be . Each scheduling cycle contains In a certain period of time, Indicates the first Historical dates Outdoor temperature forecast for the period of time, in Indicates the first Historical dates Actual outdoor temperature value for the time period, historical temperature prediction error Represented as: (10) Calculate the sample mean of all historical temperature prediction errors. : (11) The sample mean was used to correct the deviation of the daytime temperature forecast curve, resulting in the corrected daytime outdoor temperature forecast and the temperature forecast residual with a mean close to zero. (12) (13) in, This indicates the predicted temperature after deviation correction. This represents the temperature prediction residual with a mean close to zero; This indicates the predicted temperature after deviation correction.

[0043] The time-series recursive relationship of temperature prediction error is fitted based on historical temperature prediction error data, specifically as follows: Based on historical error samples between consecutive scheduling periods, a first-order autoregressive process is used to establish the state recursive relationship of temperature prediction error: (14) in, Indicates the first Zero-mean temperature prediction error for a given period; Indicates the first Zero-mean temperature prediction error for a given period; This represents the autoregressive coefficient of temperature prediction error, and ; Represents random perturbation; follows the mean of . variance is The Gaussian distribution, i.e. ; Temperature prediction error autoregressive coefficient The least squares method can be used to calculate based on historical centered error samples: (15) in, Indicates the first The first historical date The centered temperature prediction residual for the time period, i.e. the temperature prediction residual with a mean close to zero; Indicates the first The first historical date Centralized temperature prediction residuals for different time periods; The calculated Substituting into the first-order autoregressive process, formula (14) yields the random disturbance residuals for each historical period. : (16) Then calculate the variance of random disturbances. : (17) Based on the autoregressive coefficients and the steady-state variance of the temperature prediction error, the autocovariance among the temperature prediction errors in different scheduling periods is calculated as follows: (18) (19) in, Indicates the first time; Indicates the first time; This represents the steady-state variance of the temperature error.

[0044] When multiple air conditioners are located in the same or similar meteorological areas, common features can be identified based on unified historical temperature data. and When the meteorological conditions of different air conditioner locations differ, the historical temperature data of the corresponding areas of each air conditioner are used to execute equations (10) to (19) to form their respective temperature prediction error time series models.

[0045] After completing this step, output the temperature forecast deviation correction value, autoregressive coefficient, random disturbance variance, temperature error steady-state variance, and covariance between each scheduling period for each meteorological region, as input for the state deviation propagation calculation.

[0046] Step 2.2 Hourly propagation of temperature error to baseline power and virtual energy storage state The current scheduling day The actual outdoor temperature is obtained by superimposing the predicted outdoor temperature for the specified time period with the temperature prediction error. (20) in, Indicates the first Actual outdoor temperature at the time of the event; Indicates the first Predicted outdoor temperature for the specified time period; Indicates the first Error in predicting zero-mean temperature over a given period; Substituting the actual outdoor temperature into the baseline power function, we obtain the actual baseline power: ;(twenty one) in, Indicates the first Actual baseline power for the time period; Indicates the indoor temperature setting; Indicates the cooling energy efficiency ratio; Indicates the building's equivalent thermal resistance; Indicates the first Time period predicted baseline power; The temperature power sensitivity coefficient represents the additional power load caused by each degree Celsius deviation of the outdoor temperature. The calculation formula is as follows: ; Substituting the actual baseline power into the virtual energy storage state transition equation constructed in step 1, we obtain the actual virtual energy storage state: ;(twenty two) in, Indicates the first Actual virtual energy storage status during the time period; Indicates the first Actual virtual energy storage status during the time period; Indicates the state retention coefficient; Indicates the electrical energy conversion coefficient; Indicates the first Air conditioning power consumption during different time periods; Indicates the reference bias constant; By combining the actual virtual energy storage state with the virtual energy storage state under the predicted conditions, the virtual energy storage state deviation is calculated: ;(twenty three) in, Indicates the first Virtual energy storage status deviation over time period; Indicates the first Virtual energy storage status deviation over time period; Will Timeline recursion expansion: ;(twenty four) in, Indicates the first Each historical period number; Indicates the first Error in predicting zero-mean temperature over a given period; Since the temperature prediction error follows a normal distribution, the actual virtual energy storage state also follows a normal distribution, i.e. (27) in, Indicates the first The expected value of the actual virtual energy storage state during the time period; Indicates the first Variance of virtual energy storage state deviation over time period.

[0047] True variance Accurate calculation is achieved through double summation of the covariance matrix: (25) in, Indicates the first Virtual energy storage status deviation over time period; Indicates the first Each historical period number; Indicates the first Zero-mean temperature prediction error for a given period; Indicates the first Zero-mean temperature prediction error for a given period; Indicates the first Time period and the Covariance among time-period temperature prediction errors.

[0048] Substituting the covariance formula (18) of the AR(1) process into the above equation, we can obtain the final analytical expression for the state variance: (26) in, This represents the steady-state variance of the temperature prediction error.

[0049] Step 2.3 Time-sharing safety boundary generation based on allowable exceedance probability Set the maximum permissible probability of a single side exceeding the upper limit or falling below the lower limit for the actual virtual energy storage state of the air conditioner. Establish opportunity constraints for the upper limit of the state and the lower limit of the state respectively: (28) in, Indicates the probability of an event occurring; Indicates the upper limit of the virtual energy storage state; Indicates the lower limit of the virtual energy storage state; The chance constraint is transformed into a deterministic linear constraint using the inverse cumulative distribution function of the standard normal distribution: (29) In the formula, Indicates the first Time-period predicted virtual energy storage status (based on the baseline value of day-ahead temperature forecast). ; It is the inverse cumulative distribution function of the standard normal distribution.

[0050] make To define a safe multiplier corresponding to the maximum allowable probability of exceeding the limit, we can define: Dynamic safety margin for different time periods for: (30) The safety multiplier is the value of the inverse cumulative distribution function at the maximum permissible probability of exceeding the limit. The dynamic safety margin is defined by the product of the safety multiplier and the standard deviation of the virtual energy storage state, and the modified time-sharing safety state boundary is generated. (31) in, Indicates the first Dynamic safety margin over time period; Indicates the first Predict the virtual energy storage status over a specific time period.

[0051] The time-sharing security boundary is stored according to the air conditioning equipment access node and scheduling period, and then passed to step 3.

[0052] In step 3, the predicted baseline power, the predicted virtual energy storage state transition equation, the time-sharing safe operation boundary, and the safe operation constraints of the distribution network within the virtual power plant are all incorporated into the distribution network optimization scheduling model. The distribution network optimization scheduling model is solved with the goal of minimizing the operating cost of the virtual power plant, and time-sharing power adjustment commands for each air conditioning device are generated and issued for execution.

[0053] This step first obtains the topology and parameters of the internal distribution network of the virtual power plant, the day-ahead forecast data of distributed photovoltaics, the access location of each air conditioning unit, and the day-ahead electricity purchase price, as well as other information required for scheduling. Then, the predicted baseline power and predicted SOC state transition model of each air conditioning unit obtained in step 1, and the dynamic safety margin and corrected SOC boundary of each time period obtained in step 2 are matched with the corresponding air conditioning units and access nodes. Next, with the goal of minimizing the day-ahead operating cost of the virtual power plant, the air conditioning state and power constraints, the internal distribution network operation constraints, the distributed photovoltaic operation constraints, and the tie-line interaction power constraints are uniformly incorporated into the day-ahead optimization model. Finally, the model of the AC power flow network operation relationship is linearized, the optimization solver is called to obtain the day-ahead scheduling results, and time-of-use power instructions for each air conditioning unit and the power interaction plan with the upper-level power grid are generated according to the equipment number and access node for execution in each scheduling period of the next day.Figure 4 This is a flowchart of the day-ahead scheduling and execution process of a virtual power plant based on internal power distribution network security constraints, as proposed in Embodiment 1 of the present invention. Step 3.1 Obtain the virtual power plant's internal network and day-ahead dispatch parameters. Based on the recent optimization and scheduling requirements of the virtual power plant, basic parameters related to the internal power distribution network, various distributed resources, and interactions with the upper-level power grid were surveyed and obtained, and then uniformly organized according to node number, equipment number, and scheduling period. The required parameters mainly include the following categories: (1) Internal power distribution network parameters. Obtain the topology and node numbers of the internal power distribution network of the virtual power plant. Branch road number Total number of branch roads Branch connection relationship and the conductivity of each line susceptance Line transmission capacity Parameters such as these are obtained, along with the upper and lower limits of the allowable voltage at each node. Safety operating parameters, etc.

[0054] (2) Basic load forecast parameters. Obtain the active power of the basic load of each node within the day-ahead scheduling cycle. Reactive power prediction data The daily load forecast curves for each node are generated according to the set scheduling time intervals.

[0055] (3) Distributed photovoltaic parameters. Obtain the set of access nodes for each distributed photovoltaic system within the virtual power plant. Power factor angle Inverter capacity And the predicted maximum power output for each time period in recent days. .

[0056] (4) Air conditioning operation and access parameters. Obtain the set of access nodes for each air conditioning unit in the internal power distribution network. The equipment numbers are then mapped to the models and calculation results obtained in steps 1 and 2, including the upper and lower limits of air conditioning power. Predicted baseline power Virtual energy storage first, second and third state transition parameters Initial state and the SOC security boundaries corrected for each time period .

[0057] (5) Interconnection lines and upstream power grid interaction parameters. Obtain the set of common connection points between the virtual power plant and the upstream power grid. Main transformer or tie line capacity Other operating parameters.

[0058] (6) Economic and dispatch parameters. Obtain the time-of-use electricity purchase price of the superior power grid for each time period within the day-ahead dispatch cycle. Air conditioning adjustment compensation price Economic parameters, and scheduling cycle length Scheduling time interval Isotime parameters.

[0059] Step 3.2 Constructing the Distribution Network Optimization Scheduling Model In this invention, the objective function of the distribution network optimization scheduling model is to minimize the sum of the cost of purchasing electricity from the upstream power grid and the cost of regulating air conditioning load. The constraints of the distribution network optimization scheduling model include: Tie line power constraint: The active power exchanged between the virtual power plant and the upstream power grid shall not exceed the tie line capacity limit; Distributed photovoltaic (PV) operation constraints: The actual active power output of PV shall not exceed the maximum power output predicted before the day, and the reactive power output shall be limited by the inverter capacity; Air conditioning operation constraints: The actual dispatch power of each air conditioner is between the upper and lower limits of the allowable power, and the corresponding virtual energy storage state is within the corrected time-sharing safety state boundary; Node power balance constraint: The active and reactive power injected into each node are equal to the sum of the adjustable loads, non-adjustable loads, photovoltaic output, and power injected into the upstream grid connected to that node; Power flow constraints: AC power flow equations are used to describe the relationship between active and reactive power of branches and node voltage magnitude and phase angle difference; Distribution network safety constraints: The transmission power of each branch shall not exceed the line capacity, and the voltage amplitude of each node shall be within the allowable upper and lower limits.

[0060] Specifically: (32) in, This represents the total day-ahead operating cost of the virtual power plant; This indicates the total number of time periods within the current day's scheduling cycle; This represents the set of common connection points that connect to the upper-level power grid. This represents the set of nodes connected to the air conditioning equipment. Indicates the first Time-of-use electricity price of the upstream power grid during the specified period; Indicates the first Time period nodes Active power supplied by the upper-level power grid; Indicates the first Time period nodes Reactive power supplied from the upper-level power grid; This indicates the price for air conditioning load adjustment compensation. Indicates the first Time period nodes Actual air conditioning power dispatched at the location; Indicates the first Time period nodes Predicted baseline power for air conditioning; Indicates the scheduling step size; Represents a node The apparent power capacity limit of the connection line; Indicates the first Time period nodes The photovoltaic system actually contributes power. Indicates the first Time period nodes The actual reactive power output of the photovoltaic system is zero. Indicates the first Time period nodes The maximum predicted power output of photovoltaic power generation is located at [location]. Represents a node Apparent power capacity of photovoltaic inverter; Indicates the photovoltaic power factor angle; Indicates the first Time period nodes Predicting virtual energy storage status for air conditioning; Indicates the first Time period nodes Predicting virtual energy storage status for air conditioning; Represents a node The lower limit of the allowable operating power of the air conditioner; Represents a node The maximum allowable operating power of the air conditioner; Represents a node The virtual energy storage state of the air conditioner is at its limit; Represents a node The upper limit of the virtual energy storage state of the air conditioner; Indicates the first Time period nodes Dynamic safety margin of air conditioning; Indicates the first Time period nodes Active power injection; Indicates the first Time period nodes Reactive power injection; Indicates the first Time period nodes Base load active power; Indicates the first Time period nodes reactive power of base load; Indicates the first Time-of-day branch The active power; Indicates the first Time-of-day branch reactive power; Represents a node and nodes The electrical conductivity; Represents a node and nodes The susceptance; Indicates the first Time period nodes The voltage amplitude; Indicates the first Time period nodes The voltage amplitude; Indicates the first Time period nodes and nodes The voltage phase angle difference between them; Indicates the total number of branches; Represents the elements of the node-branch association matrix; Indicates a branch The apparent power transmission capacity limit; Represents a node Lower limit of allowable voltage amplitude; Represents a node The upper limit of the voltage amplitude.

[0061] In the objective function, the costs are, in order, electricity purchase cost and air conditioning load adjustment cost.

[0062] Equation (a) represents the tie line constraint. The power exchange between the virtual power plant and the upstream main grid is limited by the substation capacity, and it is assumed that there is no active power backfeeding.

[0063] Equations (b) to (c) represent the equipment operation constraints, specifically the constraints for distributed photovoltaic equipment and air conditioning equipment, respectively. Equation (d) represents the node power balance constraint. During any scheduling period, the node injected power consists of adjustable load, non-adjustable load, main grid power generation, and photovoltaic output.

[0064] Equation (e) represents the power flow constraint, which is expressed by the power flow equation.

[0065] Equation (f) represents the safety constraints of the distribution network, including branch capacity constraints and node voltage constraints.

[0066] Step 3.3: Solve the model linearization problem, specifically as follows: The nonlinear terms in the distribution network optimization scheduling model are linearized. The linearization process includes: introducing non-negative auxiliary variables to linearize the absolute value terms in the objective function, using the double circumscribed square method to linearize the apparent power capacity constraint, and using the improved DC power flow method to linearize the AC power flow constraint. The optimization solver is invoked to solve the linearized model, and the network loss factor is updated iteratively until the average absolute error rate of branch power converges to the preset accuracy threshold, thereby obtaining the time-of-use power plan, distributed photovoltaic power output plan and upstream power grid purchase plan for each air conditioning unit.

[0067] (1) Linearization of the absolute value term For the absolute value term in the objective function Linearization is achieved by introducing a non-negative auxiliary variable. The regulating power is defined as: (33) In the formula, for The amount of air conditioning power adjustment during a given time period can be represented by a non-negative auxiliary variable. , express: (34) Then adopt Equivalent replacement This allows for the linearization of the absolute value term.

[0068] (2) Capacity constraint linearization Capacity constraints mainly refer to apparent power not exceeding equipment capacity, which is a circular constraint and can be uniformly represented as: (35) In the formula, , , It serves as a unified representation of the active power, reactive power, and capacity limit of photovoltaic systems, interconnecting lines, and branch lines. Linearizing it using a double circumscribed square yields: (36) (3) Linearization of AC current flow constraints The power flow constraints are linearized based on the improved DC power flow method, and the specific results are as follows: (37) In the formula, For defined independent variables; and All are node voltage magnitudes used in linearized power flow (current values ​​during the iteration process).

[0069] Indicates the parameter of the first network loss factor; Indicates the second network loss factor parameter, This represents the parameter of the third network loss factor; This represents the fourth network loss factor parameter; This represents the parameter of the fifth network loss factor; The sixth network loss factor parameter is represented by the following expression: (38) In the formula, Represents a node Initial value of voltage amplitude; Represents a node Initial value of voltage amplitude; node node The initial value of the phase angle difference between them needs to be updated iteratively during the solution process.

[0070] Initial values ​​for voltage amplitude and phase angle difference are set to calculate the initial value of the network loss factor. Subsequently, the linearized optimized scheduling model is solved using the CPLEX solver, and the average absolute error rate of the power of each branch in the system is evaluated using equation (39). If the error exceeds the preset accuracy threshold, the network loss factor is updated using the currently obtained voltage amplitude and phase angle difference, and the next round of optimization is continued. The above process is continuously iterated until the average absolute error converges to the set accuracy range, and finally the day-ahead scheduling result is obtained.

[0071] (39) In the formula, Indicates the first Average absolute error rate of branch power during time period; Indicates the first Time-of-day branch The theoretical accurate value of active power (obtained from AC power flow calculation); Indicates the first Time-of-day branch The theoretical accurate value of reactive power (obtained from AC power flow calculation).

[0072] Step 3.4: Issuance and execution of scheduling results. After completing the day-ahead optimization model solution, the power plans for each air conditioning unit, the output plans for distributed photovoltaic power, and the power purchase plans from the upper-level power grid are compiled according to the equipment number, access node, and scheduling period. Corresponding control commands are then issued to the corresponding equipment or operation control system. The specific execution method is as follows: For air conditioning equipment, by adjusting the compressor operating frequency and changing the set temperature, each air conditioning unit can operate according to the optimized time-sharing target power.

[0073] For distributed photovoltaic resources, power reduction is achieved by adjusting the output setpoint of the inverter, so that it can operate according to the optimized time-sharing plan.

[0074] For power exchange with the upper-level power grid, the amount of electricity purchased from the upper-level power grid in each time period is adjusted according to the optimized time-of-use power purchase plan, and the operation of internal resources is coordinated to ensure that the actual exchange power of the tie line matches the power purchase plan.

[0075] Embodiment 1 of this invention addresses the problem of changes in actual power demand and adjustable capacity of air conditioners caused by outdoor temperature prediction deviations. It constructs a day-ahead dispatching system that includes virtual energy storage modeling for air conditioners, temperature error state propagation, time-sharing safety boundary generation, and collaborative optimization of the internal power distribution network. This system incorporates the continuous impact of temperature prediction errors on the operating status of air conditioners into the dispatching plan formulation process in advance, ensuring the executability of air conditioner adjustment commands under actual temperature fluctuation conditions and the safe and economical operation of the internal power distribution network of the virtual power plant.

[0076] Example 2 To fully illustrate the implementation process of the day-ahead dispatching method for an air-conditioned virtual power plant considering temperature prediction errors disclosed in Embodiment 1 of this invention, an air-conditioned virtual power plant operation scenario is constructed using the IEEE 33-node distribution system to verify the effectiveness of this patented method. The specific steps are as follows: Obtain the internal power distribution network topology of the virtual power plant, as well as the connection status of basic loads, distributed photovoltaics, and air conditioning equipment. Figure 5 This is a topology diagram of the IEEE 33-node system in an application example of the present invention; the total basic non-adjustable load is 3.59MW, the total installed capacity of distributed photovoltaic is 6MW, node 1 serves as a common connection point for power interaction with the upper-level grid, and four large-scale commercial building air conditioners are connected to nodes 8, 24, 25 and 32 respectively, denoted as AC1 to AC4.

[0077] The building parameters and equipment operating parameters corresponding to each air conditioner were investigated, as shown in Table 1. Based on the obtained parameters, first-order equivalent thermal parameter models for AC1 to AC4 were established respectively. The virtual energy storage state of charge (SOC) was defined according to step 1.2, and an independent virtual energy storage state transition model was formed for each air conditioner.

[0078] Table 1: Air Conditioning Equipment and Building Thermodynamic Parameters

[0079] Further investigation was conducted on historical outdoor temperature data for the target area. Based on historical intraday temperature variation patterns, a 24-hour outdoor temperature forecast curve was obtained. The corresponding baseline power curve was calculated based on step 1.3. Figure 6 As shown, Figure 6 This is a baseline power curve for air conditioning prediction in an application example of the present invention; Subsequently, the historical daytime temperature forecasts and corresponding actual temperatures for the region were investigated. Based on step 2.1, a first-order autoregressive model of the temperature prediction error was fitted. The autoregressive coefficient of the AR(1) model was 0.939, and the standard deviation of the temperature prediction error was 0.711℃. The covariance between temperature prediction errors in different time periods was calculated based on this error model. The temperature error was first converted into the prediction baseline power offset. Based on step 2.2, the propagation was carried out hourly along the state transition equations of each air conditioning virtual energy storage system. The variance and standard deviation of the SOC state deviation of AC1 to AC4 in 24 scheduling periods were calculated respectively.

[0080] Based on the above state error propagation results, the maximum permissible over-limit probability is set to 30% and 10% respectively. The corresponding dynamic safety margin is calculated based on step 2.3, and the original SOC operating boundary of each air conditioner is corrected hourly to obtain the time-sharing safety boundary under the two risk control requirements.

[0081] After completing the time-of-use safety boundary calculation for air conditioning, further investigation is conducted into the internal power distribution network and economic operating parameters required for the day-ahead scheduling of the virtual power plant. These operating parameters are then input into the day-ahead optimization scheduling model constructed in step 3. In this embodiment, the model is built using Pyomo in a Python environment, and the CPLEX solver is called to obtain the 24-hour day-ahead operating plan.

[0082] The main operational results obtained from the solution are shown in Table 2, and the day-ahead scheduling results for air conditioning are as follows: Figure 7 As shown.

[0083] Table 2: Comparison of System Operation Results

[0084] Figure 7 The diagram shows the day-ahead scheduling results of air conditioning in an application example of this invention. The results indicate that as the permissible over-limit probability decreases, the time-sharing safety boundaries of each air conditioning unit further shrink inward, correspondingly reducing the air conditioning adjustment space available for day-ahead optimization, thus increasing scheduling costs. When the maximum permissible over-limit probability is set to 30%, the day-ahead operating cost of the virtual power plant is 69,202.98 yuan; when the maximum permissible over-limit probability decreases to 10%, the operating cost increases to 73,004.03 yuan.

[0085] To verify the protective effect of the generated time-sharing safety boundary against actual temperature prediction errors, while keeping the aforementioned day-ahead power plan unchanged, 1000 sets of outdoor temperature curves satisfying the established AR(1) time-series characteristics were generated using the Monte Carlo method, as shown below. Figure 8 As shown, Figure 8 This is a set of outdoor temperature curves used in an application example of the present invention; and the virtual energy storage state transition model of four air conditioners is re-introduced for a 24-hour operation simulation. The number of SOC exceedances for each air conditioner in each time period is counted, and the average cross-sectional exceedance rate of the four devices is further calculated. The risk control effect of the corresponding day-ahead plan is evaluated based on the maximum average cross-sectional exceedance rate within 24 hours.

[0086] The verification results are shown in Table 3 and Figure 9 As shown.

[0087] Table 3: Comparison of Maximum Average Cross-Section Exceedance Rate in Various Scenarios

[0088] Figure 9 This is the Monte Carlo trajectory evolution diagram of AC1 SOC in an application example of this invention. When the maximum permissible exceedance probability is set to 30%, the maximum average cross-sectional exceedance rate under 1000 random temperature scenarios is 31.15%, and the highest indoor temperature simulated by AC1 is 34.23℃. When the maximum permissible exceedance probability is set to 10%, the maximum average cross-sectional exceedance rate decreases to 10.40%, and the highest indoor temperature of AC1 decreases to 32.57℃. In both cases, the exceedance rates obtained by actual statistics are close to the set maximum permissible exceedance probability, indicating that the dynamic safety margin generated based on the time-series propagation results of temperature error can effectively adjust the range of available air conditioning status according to different risk control requirements.

[0089] Example 3 Based on the day-ahead dispatching method for air-conditioned virtual power plants considering temperature prediction errors proposed in Embodiment 1 of the present invention, Embodiment 3 of the present invention further proposes a day-ahead dispatching device for air-conditioned virtual power plants considering temperature prediction errors. Figure 10 This is a schematic diagram of a day-ahead dispatching device for an air-conditioning type virtual power plant that takes into account temperature prediction errors, as proposed in Embodiment 3 of the present invention.

[0090] At the hardware level, the electronic device 100 includes a processor 110, and optionally, an internal bus 120, a network interface 130, and memory. The memory may include main memory, such as high-speed random-access memory (RAM), or it may also include non-volatile memory, such as at least one disk drive. Of course, the electronic device may also include other hardware required for its functions.

[0091] The processor 110, network interface 130, and memory can be interconnected via an internal bus 120. This internal bus 120 can be an Industry Standard Architecture (ISA) bus, a Peripheral Component Interconnect (PCI) bus, or an Extended Industry Standard Architecture (EISA) bus, etc. The bus can be categorized as an address bus, data bus, control bus, etc. For ease of illustration, only a single bidirectional arrow is used in this diagram, but this does not imply that there is only one bus or one type of bus. The memory is used to store programs. Specifically, the program can include program code, which includes computer operation instructions. The memory can include main memory 140 and non-volatile memory 150, and provides instructions and data to the processor 110.

[0092] Processor 110 reads the corresponding computer program from non-volatile memory 150 into memory 140 and runs it, forming a day-ahead dispatching device for an air-conditioning type virtual power plant at the logical level. Processor 110 executes the program stored in memory and specifically performs the following: In step 1, the virtual energy storage modeling input information required for the operation of the air conditioner is obtained, the air conditioner virtual energy storage model is constructed based on the virtual energy storage modeling input information, and the predicted baseline power and the predicted virtual energy storage state transition equation for each time period are calculated. In step 2, the time-series recursive relationship of temperature prediction error is fitted based on historical temperature prediction error data. The temperature prediction error is converted into virtual energy storage state deviation and propagated hourly along the virtual energy storage state transition equation. The time-sharing safe operation boundary after correction for each time period is generated according to the preset maximum allowable over-limit probability.

[0093] In step 3, the predicted baseline power, the predicted virtual energy storage state transition equation, the time-sharing safe operation boundary, and the safe operation constraints of the distribution network within the virtual power plant are all incorporated into the distribution network optimization scheduling model. The distribution network optimization scheduling model is solved with the goal of minimizing the operating cost of the virtual power plant, and time-sharing power adjustment commands for each air conditioning device are generated and issued for execution.

[0094] Figure 1It can be applied to processor 110, or implemented by processor 110. The processor may be an integrated circuit chip with signal processing capabilities. In the implementation process, each step of the above method can be completed by the integrated logic circuit in the processor or by instructions in the form of software. The processor mentioned above can be a general-purpose processor, including a central processing unit (CPU), a network processor (NP), etc.; it can also be a digital signal processor (DSP), an application-specific integrated circuit (ASIC), a field-programmable gate array (FPGA), or other programmable logic devices, discrete gate or transistor logic devices, or discrete hardware components. It can implement or execute the various methods, steps, and logic block diagrams disclosed in the embodiments of this application. The general-purpose processor can be a microprocessor or any conventional processor. The steps of the method disclosed in the embodiments of this application can be directly embodied as being executed by a hardware decoding processor, or executed by a combination of hardware and software modules in the decoding processor. The software module can reside in a mature storage medium in the field, such as random access memory, flash memory, read-only memory, programmable read-only memory, electrically erasable programmable memory, or registers. This storage medium is located in memory, and the processor reads information from the memory and, in conjunction with its hardware, completes the steps of the above method.

[0095] Of course, in addition to software implementation, the electronic device of this application does not exclude other implementation methods, such as logic devices or a combination of hardware and software, etc. In other words, the execution subject of the following processing flow is not limited to each logic unit, but can also be hardware or logic devices.

[0096] The description of the relevant parts of the day-ahead dispatching equipment for an air-conditioned virtual power plant that considers temperature prediction error provided in Embodiment 3 of this application can be found in the detailed description of the corresponding parts of the day-ahead dispatching method for an air-conditioned virtual power plant that considers temperature prediction error provided in Embodiment 1 of this application, and will not be repeated here.

[0097] It should be noted that, in this document, relational terms such as "first" and "second" are used merely to distinguish one entity or operation from another, and do not necessarily require or imply any such actual relationship or order between these entities or operations. Furthermore, the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that the elements inherent in a process, method, article, or apparatus that includes a list of elements are included. Without further limitations, an element defined by the phrase "comprising one..." does not exclude the presence of other identical elements in the process, method, article, or apparatus that includes said element. Additionally, portions of the technical solutions provided in the embodiments of this application that are consistent with the implementation principles of corresponding technical solutions in the prior art have not been described in detail to avoid excessive elaboration.

[0098] While specific embodiments of the present invention have been described above in conjunction with the accompanying drawings, this is not intended to limit the scope of protection of the present invention. Those skilled in the art can make other modifications or variations based on the above description. It is neither necessary nor possible to exhaustively describe all embodiments here. Various modifications or variations that can be made by those skilled in the art without creative effort based on the technical solutions of the present invention are still within the scope of protection of the present invention.

Claims

1. A day-ahead dispatching method for an air-conditioning type virtual power plant considering temperature prediction errors, characterized in that, Includes the following steps: The system obtains the virtual energy storage modeling input information required for air conditioning operation, constructs an air conditioning virtual energy storage model based on the virtual energy storage modeling input information, and calculates the predicted baseline power and predicted virtual energy storage state transition equations for each time period. The input information includes building thermal parameters, equipment operating parameters, temperature control requirements, and day-ahead outdoor temperature prediction data. Based on historical temperature prediction error data, the time-series recursive relationship of temperature prediction error is fitted, the temperature prediction error is converted into virtual energy storage state deviation and propagated hourly along the virtual energy storage state transition equation, and the time-sharing safe operation boundary after correction for each time period is generated according to the preset maximum allowable over-limit probability. The predicted baseline power, the predicted virtual energy storage state transition equation, the time-sharing safe operation boundary, and the safe operation constraints of the distribution network within the virtual power plant are all incorporated into the distribution network optimization scheduling model. The distribution network optimization scheduling model is solved with the goal of minimizing the operating cost of the virtual power plant, and time-sharing power adjustment commands for each air conditioning device are generated and issued for execution.

2. The method according to claim 1, characterized in that, The detailed process of issuing and executing the command is as follows: The power plans for each resource are organized according to the equipment number, access node, and scheduling period, and the corresponding control commands are sent to the corresponding execution units; the execution units include air conditioning equipment, distributed photovoltaic resources, and the upstream power grid. For air conditioning equipment, the compressor operating frequency is adjusted or the set temperature is changed to ensure that each air conditioning unit operates according to the optimized time-sharing target power. For distributed photovoltaic resources, the power is reduced by adjusting the output setpoint of the inverter to ensure that it operates according to the optimized time-sharing plan. For power interaction with the upper-level power grid, the amount of electricity purchased from the upper-level power grid in each time period is adjusted according to the optimized time-sharing power purchase plan, and the operation of internal resources is coordinated to ensure that the actual exchange power of the tie line matches the power purchase plan.

3. The method according to claim 1, characterized in that, Based on the aforementioned virtual energy storage modeling input information, an air conditioning virtual energy storage model is constructed as follows: Obtain the building equivalent thermal resistance, building equivalent heat capacity, cooling energy efficiency ratio, upper and lower limits of allowable operating power, indoor temperature setpoint, and upper and lower limits of allowable indoor temperature for each air conditioning unit. Based on the first-order equivalent thermal parameter model, the product of the building's equivalent heat capacity and the rate of change of indoor temperature is equal to the difference between outdoor temperature and indoor temperature divided by the building's equivalent thermal resistance, and then the product of the cooling energy efficiency ratio and the power consumption of air conditioning is subtracted to establish the continuous time state transition relationship of indoor temperature. The continuous-time state transition relationship of indoor temperature is discretized according to the day-ahead scheduling step size, and the virtual energy storage state is defined by the position of indoor temperature relative to the temperature allowable range, and the virtual energy storage state transition equation of air conditioning is established.

4. The method according to claim 1, characterized in that, The predicted baseline power and the predicted virtual energy storage state transition equations for each time period are calculated as follows: Historical outdoor temperature data is acquired, and a day-ahead temperature prediction model is established using regression fitting methods to generate a day-ahead outdoor temperature prediction sequence for the cost scheduling cycle. Substitute the predicted outdoor temperature values ​​for each time period into the baseline power function, and calculate the predicted baseline power for each time period hour by hour. The baseline power function is expressed as the product of the difference between the outdoor temperature and the indoor set temperature divided by the cooling energy efficiency ratio and the building's equivalent thermal resistance. Substitute the predicted baseline power for each time period into the virtual energy storage state transition equation to form the virtual energy storage state transition equation under the day-ahead forecast conditions, and calculate the initial value of the virtual energy storage state based on the measured indoor temperature at the start of dispatch.

5. The method according to claim 1, characterized in that, Before fitting the time-series recursive relationship of temperature prediction errors based on historical temperature prediction error data, the following steps are also included: The daytime outdoor temperature forecast data and the corresponding actual outdoor temperature data for multiple consecutive historical dates in the area where the air conditioner is located are aligned according to the scheduling time interval. Let the number of historical sample dates be . Each scheduling cycle contains In a certain period of time, Indicates the first Historical dates Outdoor temperature forecast for the period of time, in Indicates the first Historical dates Actual outdoor temperature value for the time period, historical temperature prediction error Represented as: ； Calculate the sample mean of all historical temperature prediction errors. ； The sample mean is used to correct the deviation of the daytime temperature forecast curve, resulting in the corrected daytime outdoor temperature forecast and a temperature forecast residual with a mean close to zero. ； ； in, This represents the revised day-ahead outdoor temperature forecast. This represents the residual of historically centered temperature prediction.

6. The method according to claim 5, characterized in that, The time-series recursive relationship of temperature prediction error is fitted based on historical temperature prediction error data, specifically as follows: Based on historical error samples between consecutive scheduling periods, a first-order autoregressive process is used to establish the state recursive relationship of temperature prediction error: ； in, Indicates the first Zero-mean temperature prediction error for a given period; Indicates the first Zero-mean temperature prediction error for a given period; This represents the autoregressive coefficient of temperature prediction error, and ; Indicates random perturbation; Temperature prediction error autoregressive coefficient Represented as: ； in, Indicates the first The first historical date The centered temperature prediction residual for the time period, i.e. the temperature prediction residual with a mean close to zero; Indicates the first The first historical date Centralized temperature prediction residuals for different time periods; The calculated Substituting into the first-order autoregressive process, we obtain the random perturbation residuals for each historical period. : ； Then calculate the variance of random disturbances. : ； Based on the autoregressive coefficients and the steady-state variance of the temperature prediction error, calculate the autocovariance among the temperature prediction errors for different scheduling periods: ； in, Indicates the first time; Indicates the first time; This represents the steady-state variance of the temperature error.

7. The method according to claim 6, characterized in that, The temperature prediction error is converted into a virtual energy storage state deviation and propagated hourly along the virtual energy storage state transition equation, specifically as follows: The current scheduling day The actual outdoor temperature is obtained by superimposing the predicted outdoor temperature for the specified time period with the temperature prediction error. ； in, Indicates the first Actual outdoor temperature at the time of the event; Indicates the first Predicted outdoor temperature for the specified time period; Indicates the first Error in predicting zero-mean temperature over a given period; Substituting the actual outdoor temperature into the baseline power function, we obtain the actual baseline power: ； in, Indicates the first Actual baseline power for the time period; Indicates the indoor temperature setting; Indicates the cooling energy efficiency ratio; Indicates the building's equivalent thermal resistance; Indicates the first Time period predicted baseline power; Indicates the temperature-power sensitivity coefficient; Substituting the actual baseline power into the virtual energy storage state transition equation, we obtain the actual virtual energy storage state: ； in, Indicates the first Actual virtual energy storage status during the time period; Indicates the first Actual virtual energy storage status during the time period; Indicates the state retention coefficient; Indicates the electrical energy conversion coefficient; Indicates the first Air conditioning power consumption during different time periods; Indicates the reference bias constant; By combining the actual virtual energy storage state with the virtual energy storage state under the predicted conditions, the virtual energy storage state deviation is calculated: ； in, Indicates the first Virtual energy storage status deviation over time period; Indicates the first Virtual energy storage status deviation over time period; Will Timeline recursion expansion: ； Since the temperature prediction error follows a normal distribution, the actual virtual energy storage state also follows a normal distribution, i.e. ； in, Indicates the first The expected value of the actual virtual energy storage state during the time period; Indicates the first Variance of virtual energy storage state deviation over time period ； in, Indicates the first Virtual energy storage status deviation over time period; Indicates the first Each historical period number; Indicates the first Zero-mean temperature prediction error for a given period; Indicates the first Zero-mean temperature prediction error for a given period; Indicates the first Time period and the Covariance among time-period temperature prediction errors; Formula Substituting into the above equation, we can obtain the final analytical expression for the state variance: ； in, This represents the steady-state variance of the temperature prediction error.

8. The method according to claim 7, characterized in that, The time-sharing safe operation boundaries for each time period are generated based on the preset maximum permissible probability of exceeding the limit, specifically as follows: Set the maximum permissible probability of a single side exceeding the upper limit or falling below the lower limit for the actual virtual energy storage state of the air conditioner. Establish opportunity constraints for the upper limit of the state and the lower limit of the state respectively: ； in, Indicates the probability of an event occurring; Indicates the upper limit of the virtual energy storage state; Indicates the lower limit of the virtual energy storage state; The chance constraint is transformed into a deterministic linear constraint using the inverse cumulative distribution function of the standard normal distribution: ； in, Indicates the first Time Period Prediction Virtual Energy Storage Status ; Let be the inverse cumulative distribution function of the standard normal distribution, and let... To define a safe multiplier corresponding to the maximum allowable probability of exceeding the limit, we can define: Dynamic safety margin for different time periods for: ； The revised time-sharing safety state boundary is represented as follows: ； in, Indicates the first Dynamic safety margin over time period; Indicates the first Predict the virtual energy storage status over a specific time period.

9. The method according to claim 1, characterized in that, The objective function of the power distribution network optimization scheduling model is to minimize the sum of the cost of purchasing electricity from the upstream power grid and the cost of regulating air conditioning load. The constraints of the power distribution network optimization scheduling model include: Tie line power constraint: The active power exchanged between the virtual power plant and the upstream power grid shall not exceed the tie line capacity limit; Distributed photovoltaic (PV) operation constraints: The actual active power output of PV shall not exceed the maximum power output predicted before the day, and the reactive power output shall be limited by the inverter capacity; Air conditioning operation constraints: The actual dispatch power of each air conditioner is between the upper and lower limits of the allowable power, and the corresponding virtual energy storage state is within the corrected time-sharing safety state boundary; Node power balance constraint: The active and reactive power injected into each node are equal to the sum of the adjustable loads, non-adjustable loads, photovoltaic output, and power injected into the upstream grid connected to that node; Power flow constraints: AC power flow equations are used to describe the relationship between active and reactive power of branches and node voltage magnitude and phase angle difference; Distribution network safety constraints: The transmission power of each branch shall not exceed the line capacity, and the voltage amplitude of each node shall be within the allowable upper and lower limits.

10. The method according to claim 9, characterized in that, The optimal scheduling model for the distribution network is solved with the objective of minimizing the operating cost of the virtual power plant, specifically as follows: The nonlinear terms in the distribution network optimization scheduling model are linearized. The linearization process includes: introducing non-negative auxiliary variables to linearize the absolute value terms in the objective function, using the double circumscribed square method to linearize the apparent power capacity constraint, and using the improved DC power flow method to linearize the AC power flow constraint. The optimization solver is invoked to solve the linearized model, and the network loss factor is updated iteratively until the average absolute error rate of branch power converges to the preset accuracy threshold, thereby obtaining the time-of-use power plan, distributed photovoltaic power output plan and upstream power grid purchase plan for each air conditioning unit.

Images