Port power system dispatching depolymerization method considering net load uncertainty

Abstract

The invention provides a port power system dispatching depolymerization method considering net load uncertainty. The method aims to solve the technical problems of scheduling response lag, low system utilization rate, high operation cost and the like caused by neglecting the uncertainty of net load fluctuation in traditional port power scheduling. According to the method, a two-stage risk avoidance model is adopted, the first stage is day-ahead scheduling, the second stage is real-time scheduling, and collaborative peak clipping and scheduling are achieved in combination with a Gaussian mixture model optimized through an algorithm. According to the method, the risk perception capability is obviously enhanced, the resource scheduling precision is obviously improved, the comprehensive economy advantage is prominent, multiple targets are uniformly optimized, and the economy is optimal while the system safety is guaranteed. The method is suitable for being used as a port power system dispatching depolymerization method considering the net load uncertainty.

Description

Technical Field

[0001] This invention relates to system dispatching in the field of power, specifically a port power system dispatching de-aggregation method that considers net load uncertainty. Background Technology

[0002] As the core hub of the maritime transport system, ports are key drivers of economic development in port cities. Their core function is to support cross-regional cargo transshipment and build logistics connections between surrounding areas and inland hinterlands, possessing irreplaceable hub value in the global supply chain network. According to publicly available statistics from the International Maritime Organization, over 90% of global trade relies on maritime transport; currently, maritime transport remains the most cost-effective mode of long-distance transport of goods and raw materials globally. Simultaneously, ports are energy-intensive industrial and commercial complexes. The entire transportation sector, encompassing road, rail, air, and sea transport, accounts for nearly 20% of global energy consumption, with maritime transport accounting for approximately 4% of total energy consumption in the transportation sector. However, traditional port energy infrastructure systems generally operate independently, resulting in low overall system energy efficiency and high operating costs. Furthermore, due to the massive turnover of bulk energy goods such as crude oil, ports are significantly affected by greenhouse gas emissions such as carbon dioxide, facing prominent environmental pollution problems.

[0003] Based on the energy issues in port development mentioned above, in order to achieve the goal of sustainable port development, it is urgent to propose an economical port area energy solution to address the technical pain points of low efficiency and high operating costs of existing port energy systems.

[0004] Currently, power dispatching schemes in port areas typically take into account multiple aspects such as shore power load fluctuations and energy storage coordination. The following are some of the main contents and methods: Multi-source coordinated scheduling model: Considering the upper limit of the shore power main transformer capacity, the real-time power generation output of renewable energy, the current charging and discharging status of the storage system, and the priority levels and berthing durations of different ships, a real-time optimal power allocation model based on mixed integer programming is established to achieve reasonable scheduling of power resources in the port area.

[0005] Hybrid Energy Storage Coordination and Dispatch: Establish a hybrid energy storage system consisting of fuel cells, electrolyzers, and lithium batteries. Based on the port area's traffic-energy coupling characteristics, first establish a port area berth-quay crane joint dispatch model with the goal of minimizing the total berth time of ships. Then, based on the degradation cost model of hybrid energy storage, formulate a day-ahead dispatch strategy with the goal of optimizing the daily operating cost of the port area, thereby balancing the power imbalance between new energy sources and loads.

[0006] Demand-side response scheduling: An intermittent power supply strategy is adopted to regulate the storage temperature of adjustable equipment such as refrigerated containers, ensuring they operate within the permissible temperature range and avoiding peak electricity consumption caused by simultaneous operation of equipment. Simultaneously, referencing fluctuations in main grid electricity prices, the port area is guided to store energy during periods of low electricity prices and release it during periods of high electricity prices, shifting electricity consumption to off-peak periods and reducing electricity costs.

[0007] Although the above-mentioned dispatching schemes each have their own advantages, they still have some shortcomings. For example, since traditional shore power often uses a fixed power allocation method, the time when ships arrive at the shore has a large degree of randomness, making it impossible to respond in a timely manner. The dispatching response is delayed, resulting in low system utilization and power imbalance. The port wind speed, light intensity and ship berthing load have a large degree of uncertainty. Traditional schemes do not take this uncertainty into account, which can easily lead to deviations, increase actual operating costs, and bring safety hazards. Summary of the Invention

[0008] To improve the stability and security of port power system dispatching and avoid energy waste, this invention proposes a port power system dispatching de-aggregation method that considers net load uncertainty. This method aims to solve the technical problems of delayed dispatching response, low system utilization, and high operating costs caused by traditional port power dispatching methods that ignore net load fluctuation uncertainty.

[0009] The solution adopted by this invention to solve the technical problem is: For port power supply considering net load uncertainty, a dispatch de-aggregation method is first provided, including the following steps: S1: A two-stage optimization operation method is proposed for the power system, including port cranes, port shore power, transfer vehicles and other port power system equipment, under the worst risk conditions. The first stage is day-ahead scheduling. In the day-ahead scheduling stage, the port dynamically deaggregates scheduling instructions by quantifying the characteristic differences between resources and demand, so as to achieve collaborative peak shaving.

[0010] In step S1, after the main grid issues a peak shaving demand, the port declares its adjustable capacity and receives the dispatch instructions. Then, the differences in power demand characteristics of each power-consuming device are quantified, the response priority is calculated, resource coordination is achieved by de-aggregating dispatch instructions, and finally, deviation assessment is performed based on the settlement results and the main grid.

[0011] S2: The port uses feature matching and mixed-integer linear programming optimization to de-aggregate scheduling instructions into power resources, achieving coordinated peak shaving scheduling with optimal comprehensive cost while satisfying operational constraints.

[0012] In step S2, the total port scheduling demand is decomposed into various power resources using a feature matching method. First, a collaborative scheduling scheme is established, iteratively allocating the total demand to available resources. This approach satisfies basic operational constraints while minimizing power exchange with the grid, optimizing resource willingness costs, and maximizing port peak-shaving benefits. Then, a general feature matching model is constructed, mapping the operating characteristics of power resources to a comparable format. A discrete selection model is used to prioritize resource and scheduling commands, ensuring that the most compatible resources respond first. Finally, a mixed-integer linear programming method is used to de-aggregate and settle scheduling commands, simultaneously calculating resource response costs and port peak-shaving benefits. This achieves optimized and efficient port system operation while ensuring power balance.

[0013] Furthermore, the day-ahead scheduling steps of steps S1 and S2 are as follows: based on the peak shaving demand instructions of the main power grid, obtain the power resource information required by the port, and dynamically decompose and allocate the total scheduling instructions to multiple power-consuming devices by quantifying the characteristic differences between the required power resources and the peak shaving demand. The power resources include one or more of the following: port cranes, port shore power facilities, electric transfer vehicles, and energy storage systems.

[0014] S3: A two-stage optimization operation method for power systems containing electrical equipment such as port cranes, port shore power, and transfer vehicles under worst-case risk conditions. The second stage is real-time dispatch, i.e., the real-time operation and control cost F is... lsm As a loss function, for each scenario Introducing auxiliary variables The objective function and constraints are adjusted to achieve scheduling.

[0015] In step S3, the real-time scheduling is based on a net load probability scenario set, comprehensively considering the power deviation between prediction and actual power, as well as the flexibility and ramp-up capabilities of each control entity. Optimization decisions are made under worst-case risk conditions to ensure the system's operational stability and economy in response to real-time load fluctuations. A risk-avoidance optimization model is established by setting the real-time control cost as a loss function and introducing auxiliary variables for each scenario.

[0016] Furthermore, in step S3, real-time scheduling involves adjusting each electrical device based on the real-time uncertainty of the net load, with the goal of minimizing real-time operation and control costs, to address real-time power deviations.

[0017] S4: Create a Gaussian mixture model, use a density-preserving hierarchical expectation-maximization algorithm to reduce the model, and modify the covariance matrix of the Gaussian mixture model to weaken the peak characteristics.

[0018] In step S4, the Gaussian mixture model is created by using a density-preserving hierarchical expectation-maximization algorithm to reduce and optimize the model. This algorithm preserves the main density features of the original data distribution through hierarchical clustering, and simultaneously regularizes the covariance matrix of the Gaussian mixture model during the expectation-maximization iteration process, effectively suppressing the peak characteristics of the probability density function and enhancing the model's smoothness and generalization ability.

[0019] S5: The scheduling method is simulated by processing the data using a modified Gaussian mixture model.

[0020] In step S5, a probabilistic model is used to model the uncertainty of the net load in the real-time scheduling step S3 in order to optimize the scheduling decision.

[0021] The method of modeling the uncertainty of net load using a probabilistic model includes: collecting historical net load data; fitting the probability distribution of the net load data using a Gaussian mixture model; and further includes the step of reducing the computational complexity of the Gaussian mixture model.

[0022] In this step S5, the real-time scheduling step involves generating multiple net load probability scenarios based on the Gaussian mixture model and making scheduling decisions based on worst-case risk conditions.

[0023] Compared with existing technologies, the port power system dispatching method described in this invention has significant advantages in three aspects: uncertainty handling, dispatching accuracy, and operational economy. I. Significantly Enhanced Risk Perception Capability. Traditional methods rely on deterministic predictions, which are poorly adaptable to net load fluctuations. This invention employs a two-stage risk aversion model, combined with an algorithmically optimized Gaussian mixture model, to accurately characterize net load uncertainty. By minimizing the expected cost in the worst-case scenario, it proactively avoids extreme fluctuation risks, significantly improving system reliability.

[0024] II. Significantly Improved Resource Scheduling Accuracy. Traditional scheduling methods are inefficient due to the diverse characteristics of port resources. This invention proposes a feature-matching-based de-aggregation strategy. By constructing a multi-dimensional feature matrix of resources and demands, and utilizing a mathematical model for precise matching and priority ranking, dynamic optimal decomposition of scheduling instructions is achieved.

[0025] Third, it boasts significant comprehensive economic advantages. This invention establishes a comprehensive cost function to optimize multiple objectives in a unified manner, achieving optimal economic efficiency while ensuring system security.

[0026] In summary, this method is suitable for application as a de-aggregation method for port power system dispatch considering net load uncertainty. Attached Figure Description

[0027] Figure 1 This is a schematic diagram of the overall block diagram of the present invention. Detailed Implementation

[0028] To make the objectives, technical solutions, and advantages of the embodiments of the present invention clearer, the technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some, not all, of the embodiments of the present invention. The components of the embodiments of the present invention described and shown in the accompanying drawings can generally be arranged and designed in various different configurations. Therefore, the following detailed description of the embodiments of the present invention provided in the accompanying drawings is not intended to limit the scope of the claimed invention, but merely to illustrate selected embodiments of the invention. All other embodiments obtained by those skilled in the art based on the embodiments of the present invention without inventive effort are within the scope of protection of the present invention.

[0029] In this invention, all embodiments, implementation methods, and features can be combined with each other without contradiction or conflict. In this invention, conventional equipment, devices, and components can be commercially available or self-made according to the disclosure of this invention. In this invention, to highlight the key points, some conventional operations and equipment, devices, and components are omitted or only briefly described.

[0030] This invention proposes a two-stage optimized operation method for power systems containing equipment such as port cranes, shore power, and transfer vehicles. In the day-ahead dispatch phase, the port receives peak-shaving demand instructions from the main power grid. By quantifying the characteristic differences between various internal power resources and dispatch demands, and utilizing feature matching and mixed-integer linear programming optimization, the macro-level dispatch instructions are dynamically and accurately decomposed and allocated to specific power resources such as port cranes, shore power, and transfer vehicles, achieving coordinated peak shaving. In the real-time dispatch phase, to address the deviation between real-time net load and predicted values, a real-time dispatch model based on a net load probability scenario set is constructed. With the goal of minimizing the expected control cost under worst-case conditions, the output of each resource is adjusted to ensure the stability and economy of the system's real-time operation. Through this method, the port power system's ability to perceive and avoid risks related to net load uncertainty is significantly enhanced, improving the accuracy and dynamic response speed of resource dispatch, and optimizing operational economy while ensuring system safety.

[0031] This invention provides a method for load de-aggregation of port power systems under overlapping and intersecting power grid dispatching scenarios such as power grid peak shaving, local network congestion, and renewable energy consumption. This method can improve the port power dispatching capability, provide the optimal dispatching method, and perform optimal coordinated dispatching of port load de-aggregation.

[0032] See Figure 1To address port power supply uncertainty, a dispatch de-aggregation method is first provided, including the following steps: S1: To effectively mitigate the operational risks caused by power deviations and quantification deviations between the day-ahead net load forecast and real-time demand, and to minimize the expected cost under the maximum risk, a two-stage optimization operation method is proposed for the power system containing power-consuming equipment such as port cranes, port shore power, and transfer vehicles under the worst-case risk conditions. The first stage is: day-ahead dispatch, that is, after the main grid issues peak-shaving demand, the port declares adjustable capacity and receives dispatch instructions; then, the differences in power demand characteristics of each power-consuming equipment are quantified, the response priority is calculated, resource coordination is achieved by de-aggregating dispatch instructions, and finally, deviation assessment is performed based on the settlement results and the main grid.

[0033] During the day-ahead scheduling phase, the port dynamically deaggregates scheduling instructions by quantifying the differences in resource and demand characteristics, thereby achieving coordinated peak shaving. A scheduling framework for port planning that considers microgrid-based coordinated peak shaving auxiliary services is constructed. This framework is established through the following steps: 1) The main grid declares peak shaving demand to the peak shaving ancillary services due to energy volatility and uncertainty; 2) The port declares the response power domain of the port's power resources by combining the bidding capacity and peak shaving cycle of the power resources, and obtains the dispatch instructions for peak shaving demand; 3) The port quantifies the characteristic differences between the power resources required by the power-consuming equipment and the planned demand, and calculates the response priority of the power resources to achieve coordinated dispatch; 4) Based on the characteristic matching of power resources, a day-ahead dispatch de-aggregation framework structure for the port is constructed, and the dispatch instructions of the port are de-aggregated to achieve coordinated dispatch and meet peak shaving demand; 5) Based on the settlement results, the port and the main grid conduct deviation assessment.

[0034] Specifically, the response power domain of the port power resources in 2) above can be expressed as:

[0035] Where P agg It is the feasible region of electricity; P agg It is the port's aggregate regulation power vector; R T It is the feasible region of the polymerization power in period T; b min and b max A is the port's regulation boundary vector; A is a coefficient matrix describing the relationship between power and energy; E is an identity matrix. , These are the lower and upper limit matrices for power regulation, respectively; , These are the lower and upper bound matrices for energy regulation, respectively.

[0036] Furthermore, 4) Based on the characteristic matching of power resources, the de-aggregation of port dispatch instructions is de-aggregated to achieve coordinated dispatch and meet peak shaving requirements. This de-aggregation can be expressed as:

[0037]

[0038] Where P com These are dispatch instructions for the microgrid; I m,t It is an m-dimensional vector representing the solution aggregation situation, and its elements are 0 or 1.

[0039] The daytime scheduling also includes constructing a multi-dimensional feature matrix based on the matching relationship between the operating characteristics of power resources and scheduling instructions. This matrix is ​​used to dynamically evaluate and adjust the response priority of each power-consuming device.

[0040] Furthermore, during the day-ahead scheduling phase, it is necessary to construct a comprehensive cost function by combining the port's peak-shaving benefits with the power resource response costs, and to achieve multi-objective collaborative scheduling through an optimization model.

[0041] S2: The port uses feature matching and mixed-integer linear programming optimization to de-aggregate scheduling instructions into power resources, achieving coordinated peak shaving scheduling with optimal comprehensive cost while satisfying operational constraints.

[0042] Furthermore, step S2 involves decomposing the total port scheduling demand into various power resources using a feature matching method. First, a collaborative scheduling scheme is established, iteratively allocating the total demand to available resources. This aims to meet basic operational constraints while minimizing power exchange with the grid, optimizing resource willingness costs, and maximizing port peak-shaving benefits. Then, a general feature matching model is constructed, mapping the operating characteristics of power resources to a comparable format. A discrete selection model is used to prioritize matching resources with scheduling instructions, ensuring that the most compatible resources respond first. Finally, a mixed-integer linear programming method is used to de-aggregate and settle scheduling instructions, simultaneously calculating resource response costs and port peak-shaving benefits. This achieves optimized and efficient port system operation while ensuring power balance.

[0043] Furthermore, in this step, the day-ahead dispatching de-aggregation framework for ports includes the acquisition, matching, and decomposition of dispatching demands from grid-connected microgrids. In addition, by quantifying the power resource response cost, power resource response willingness, and power resource response peak-shaving cost, the multi-objective function of the proposed method is transformed into a comprehensive cost function. Equations and inequalities representing the constraints are also established, covering traditional resources, energy storage, and peak-shaving services.

[0044] The response costs of electricity resources include economic and willingness indicators:

[0045] Where C e For the response cost of electricity, c m For power unit response, P is the port weighting coefficient. m (t) is the power resource response function.

[0046] The willingness to respond to electricity resources is expressed as an economic indicator in the following ways:

[0047] Where C WL This represents the port's willingness to respond cost; Cca, Cra, Ctm, and Cpt are the influence coefficients of willingness to respond on response potential, rate, time, and potential for electricity resources, respectively. This is the port weighting coefficient.

[0048] Calculation of peak shaving cost in response to power resources: Grid-connected microgrids participate in peak shaving ancillary services by increasing power input from the main grid. To enhance the initiative of peak shaving ancillary services, they can obtain some benefits from this:

[0049] Where Bps is the net revenue from port peak shaving, b ps To compensate port operators for peak shaving costs. F is the port weighting coefficient. ps (t) is the peak clipping power, P ps (t) represents the actual peak-shaving power at the port, including energy storage discharge power, shore power system power reduction, and load reduction power of gantry cranes and container handling equipment; P bid (t) represents the port’s day-ahead bidding capacity, and k represents the port’s benchmark ratio.

[0050] Port peak shaving costs refer to the additional expenses incurred by ports in reducing peak electricity consumption during peak periods (such as when cargo arrives at the port in concentrated periods or when loading and unloading equipment is operating at full capacity) through various technical or management measures. These costs include energy storage equipment-related costs, adjustable load management costs, and emergency and punitive costs.

[0051] At the same time, the actual peak-shaving output is assessed to ensure the port's responsiveness in participating in peak-shaving ancillary services. The peak-shaving penalty cost can be calculated as follows:

[0052] Among them, C pp The penalty cost for a port failing to meet its peak shaving commitments, where cpp is the unit penalty coefficient, and K is the penalty cost. pp (t) represents the capacity difference due to the port's failure to meet peak shaving standards.

[0053] The port peak shaving penalty cost is a punitive fee that a port must pay to the power grid if it fails to reduce its load during peak electricity consumption periods (failing to meet the standard or failing to respond at all) after signing a power supply agreement or demand response contract with the power grid company.

[0054] In a power market environment, port bidding decisions are challenging due to resource characteristics and market variability. This challenge primarily stems from the spatial and temporal characteristics of wind and solar power, making peak shaving strategies a risk management problem in this application, considering discrete revenue distributions and confidence levels. Conditional value at risk is approximately a low probability. The expected loss in the scenario can be expressed as:

[0055] in It is the value at risk; It is a preset confidence level; It is an introduced auxiliary variable, representing the peak shaving cost and The difference between them.

[0056] The overall cost function is represented by a linear weighted average of the above five indicators:

[0057] Where Y is the risk preference coefficient, representing the port's preference for the risk of peak shaving benefits.

[0058] Considering the nonlinearity of the absolute values ​​in the formulas for the response costs of electricity resources, including economic efficiency and willingness indicators, they are linearized as follows:

[0059] Where P e E ca R ra L tm Q pt It is an auxiliary variable that is introduced.

[0060] The following formula is linearized as follows:

[0061] Where a wt b wt This is a 0-1 state variable that is introduced.

[0062] The establishment of a day-ahead dispatch de-aggregation framework for ports involves two steps: First, a feature-matching power resource response method is employed to deconstruct the port's declared dispatch demand. The initial step involves establishing a collaborative dispatch scheme that decomposes the port's total dispatch demand and iteratively allocates it to available power resources. This deconstruction process aims not only to minimize power exchange with the grid but also integrates two additional objectives: the willingness cost of power resources and the port's peak-shaving benefits. These objectives ensure the optimal utilization of power resources with different operating characteristics. In this step, the total dispatch demand is iteratively decomposed while respecting the port's fundamental operating constraints and additional factors such as peak-shaving bidding capacity and actual power resource output, ensuring that the port's dispatch demand is accurately met without exceeding power resource limits. Second, a general feature-matching model is developed to quantify the willingness to respond to each power resource within the microgrid. This process begins with establishing a feature mapping framework that transforms the operating characteristics of power resources into a format comparable to port dispatch instructions. Furthermore, a discrete selection model is employed to facilitate feature matching between power resources and dispatch instructions, ensuring that the most compatible power resources are prioritized for matching. This matching is crucial for maintaining power balance within the microgrid and optimizing the peak-shaving ancillary services provided by the port. The dispatch instructions originate from the net load curve, serving as the planned demand for the upcoming feature matching iteration. The third step involves constructing the dispatch of power resources and the settlement of port operations. The de-aggregation strategy is implemented using a mixed-integer linear programming optimization method to ensure effective de-aggregation of dispatch instructions among power resources, while simultaneously achieving power balance and peak shaving objectives. In this step, the response costs of power resources and the peak shaving benefits of the port are calculated and evaluated, ensuring optimal and efficient port operation while considering both technical and economic factors.

[0063] Among them, the day-ahead scheduling steps of steps S1 and S2 are as follows: based on the peak shaving demand instructions of the main power grid, obtain the power resource information required by the port, and dynamically decompose and allocate the total scheduling instructions to multiple power-consuming equipment by quantifying the characteristic differences between the required power resources and the peak shaving demand. The step of dynamically decomposing and allocating the overall dispatch instructions to electrical equipment includes: constructing a matching relationship between the operating characteristics of the power resources and the demand characteristics of the dispatch instructions; determining the response priority of each electrical equipment based on the matching relationship; and using an optimization model to decompose and aggregate the dispatch instructions to the corresponding electrical equipment according to the response priority.

[0064] The power resources include one or more of the following: port cranes, port shore power facilities, electric transfer vehicles, and energy storage systems.

[0065] The optimization model is a mixed-integer linear programming model.

[0066] S3: A two-stage optimization operation method for power systems including port cranes, port shore power, and transfer vehicles under worst-case risk conditions, wherein the second stage is real-time dispatch. That is, the real-time operation control cost F lsm As a loss function, for each scenario Introducing auxiliary variables The objective function and constraints are adjusted to achieve scheduling.

[0067] Furthermore, in step S3: In the second-stage real-time scheduling, based on the net load probability scenario set, and comprehensively considering the power deviation between prediction and actual power, as well as the flexibility and ramp-up capabilities of each control entity, optimization decisions are made under the worst-case risk conditions to ensure the system's operational stability and economy in response to real-time load fluctuations. A risk-avoidance optimization model is established by setting the real-time control cost as a loss function and introducing auxiliary variables for each scenario.

[0068] Furthermore, in step S3, real-time scheduling involves adjusting each electrical device based on the real-time uncertainty of the net load, with the goal of minimizing real-time operation and control costs, to address real-time power deviations.

[0069] The S3 real-time scheduling also includes robust optimization methods for scheduling decisions based on a net load probability scenario set to address system operational risks under extreme fluctuation scenarios. Based on the net load probability scenario set, the deviation between predicted and actual values ​​is fully considered, as well as the flexibility and ramp-up adjustment capabilities of each entity, to conduct real-time scheduling decisions under worst-case risk conditions, ensuring the system's stability and economy in handling net load fluctuations during real-time operation. The real-time operation control cost F... lsm As a loss function, for each scenario Introducing auxiliary variables The details are as follows:

[0070]

[0071]

[0072]

[0073] In the formula: and It is a scene The upward and downward adjustments of the power output of the i-th power plant during time period t are due to the error between the predicted and actual net load values. and Adjust the unit cost upward and downward for the i-th unit; and For the scene The adjustment of the charging and discharging power of the i-th transport vehicle during time period t is due to the error between the current net load forecast and the real-time value. This refers to the battery loss cost resulting from adjusting the output of energy storage; for The probability corresponding to the scenario.

[0074] This formula is used to adjust for power system power, power constraints and costs of cranes, transport vehicles, and shore power for ships.

[0075] S4: Create a Gaussian mixture model, use a density-preserving hierarchical expectation-maximization algorithm to reduce the model, and modify the covariance matrix of the Gaussian mixture model to weaken the peak characteristics.

[0076] Furthermore, in step S4, a Gaussian mixture model is created, and a density-preserving hierarchical expectation-maximization algorithm is used to reduce and optimize the model. This algorithm preserves the main density features of the original data distribution through hierarchical clustering, and at the same time, it performs regularization correction on the covariance matrix of the Gaussian mixture model during the expectation-maximization iteration process, effectively suppressing the peak characteristics of the probability density function and enhancing the smoothness and generalization ability of the model.

[0077] Step S4: Define the net load dataset as ,in This represents a multidimensional vector. The probability distribution function constructed using a Gaussian mixture model, and considered as a base distribution, is denoted as Gaussian Mixture Model 1. The specific mathematical formula is shown below:

[0078] In the formula: Let K be the probability density function of data x under Gaussian mixture model 1; K be the number of Gaussian components in Gaussian mixture model 1. b Equal to n, Let be the weight of the i-th Gaussian component in Gaussian Mixture Model 1; and These are the mean vector and the covariance matrix, respectively.

[0079] The number of Gaussian components in Gaussian Mixture Model 1 is directly proportional to the size of the input data; the larger the size, the higher the number of components. This leads to a larger number of key parameters such as the mean, variance, and weights that need to be estimated, resulting in a more complex model solution and lower optimization accuracy. A density-preserving hierarchical expectation-maximization algorithm is used to reduce Gaussian Mixture Model 1. The specific mathematical formula is shown below:

[0080] In the formula: For the probability distribution model of x after component reduction, K m The number of Gaussian components after reduction.

[0081] Generally, a density-preserving hierarchical expectation-maximization algorithm is used to estimate the key parameters in the above equation. When solving the model, a virtual dataset following a Gaussian mixture model 1 distribution is used to seek reduced model parameters that maximize the expectation of the log-likelihood estimation function. The relevant formula is:

[0082]

[0083] In the formula, Let be the weight of the j-th Gaussian component after reduction; Let be the mean vector of the j-th Gaussian component after reduction; Let be the covariance matrix of the j-th Gaussian component after reduction; These are variational parameters, and the sum of these parameters is 1; Let be the mathematical expectation.

[0084] The covariance matrix of the Gaussian mixture model is modified to reduce the peak characteristics: Let n-dimensional random variables Its mixed central moment matrix (covariance matrix) elements Defined as:

[0085] Wherein, E(X) i ) is the variable X i The mathematical expectation, For variables and The covariance. Covariance matrix. It can be represented as:

[0086] The main diagonal elements are the variances of random variables, satisfying... The non-main diagonal elements are the covariance. The main diagonal variance determines the degree to which random variables in a Gaussian distribution are concentrated close to the mean, and the covariance determines the degree of mutual influence between variables.

[0087] The covariance matrix Decomposition into a main diagonal matrix Non-main diagonal matrix Two parts, in the form of:

[0088] For the main diagonal matrix The covariance matrix is ​​obtained by applying a correction factor K. :

[0089] Meanwhile, to ensure the regularization of the covariance matrix, a thresholding process is applied to the main diagonal element matrix to ensure that each main diagonal element is not lower than the threshold. :

[0090] in, Main diagonal matrix The diagonal element in row a, column a.

[0091] Expected steps for correction:

[0092]

[0093] Corrected weights, mean, and covariance matrices:

[0094]

[0095]

[0096] In the formula, For the corrected variational parameters, the subscripts are... subscript , tr is the trace of the matrix; the number of virtual samples M is generally taken as 10 times K. b ; It is the corrected mathematical expectation; and These are the corrected weights, mean, and covariance matrices, respectively.

[0097] S5: The scheduling method is simulated by processing the data using a modified Gaussian mixture model.

[0098] In step S5, a probabilistic model is used to model the uncertainty of the net load in the real-time scheduling step S3 in order to optimize the scheduling decision.

[0099] The method of modeling the uncertainty of net load using a probabilistic model includes: collecting historical net load data; fitting the probability distribution of the net load data using a Gaussian mixture model; and further includes the step of reducing the computational complexity of the Gaussian mixture model.

[0100] In this step, the real-time scheduling step generates multiple net load probability scenarios based on the Gaussian mixture model and makes scheduling decisions based on worst-case risk conditions.

[0101] Step S5: Case studies show that the proposed probabilistic model for modeling the uncertainty of net load has superior performance in generalization ability and fitting, improving the autonomy of microgrids. The relationship between power balance demand and peak shaving is clarified and modeled. A novel two-stage risk avoidance model for power systems based on worst-case risk conditions is proposed. By utilizing the ramp-up adjustment capability of flexible power resources, real-time power balance under net load uncertainty is achieved, and the reliability of system operation under uncertainty is improved.

[0102] Finally, it should be noted that the above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art can still modify the technical solutions described in the foregoing embodiments or make equivalent substitutions for some of the technical features. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the protection scope of the present invention.

Claims

1. A method for port power dispatching solution disaggregation considering uncertainty of net load, characterized in that, The method comprises the following steps: S1: a two-stage optimization operation method of a power system including a port crane, a port shore power, a transfer vehicle and other power system electric equipment under the worst risk condition is proposed, wherein the first stage is day-ahead scheduling, and the port quantifies the characteristic difference between resources and demand to dynamically disaggregate scheduling instructions, thereby realizing collaborative peak shaving; After the main power grid issues a peak shaving demand, the port declares the adjustable capacity and receives the scheduling instructions; then, the power demand characteristic difference of each electric equipment is quantified, the response priority is calculated, the disaggregated scheduling instructions are realized to achieve resource collaboration, and finally the clearing results are evaluated for deviation based on the main power grid; S2: the port disaggregates the scheduling instructions to the power resources through characteristic matching and mixed integer linear programming optimization, and realizes the collaborative peak shaving scheduling with optimal comprehensive cost under the satisfaction of operation constraints; S3: Two-stage optimal operation method of power system containing power supply equipment such as port crane, port shore power and transfer vehicle under the worst risk condition, wherein the second stage is real-time scheduling, that is, real-time operation control cost F lsm As a loss function, for each scenario Introducing auxiliary variables Adjust the objective function and the constraint condition to realize scheduling, as shown below: ； ； ； ； In the formula: and is the scenario The up-regulation and down-regulation power of the i th power plant in the t period generated by the error between the day-ahead net load prediction value and the actual value; and is the unit up-regulation and down-regulation cost of the i th unit; and is the scenario The charging and discharging power adjustment of the i th transfer vehicle in the t period generated by the error between the day-ahead net load prediction value and the real-time value; is the battery loss cost generated by the output adjustment of the energy storage; is The probability corresponding to the scenario; S4: a Gaussian mixture model is created, a hierarchical expectation maximum algorithm based on density reservation is used to reduce the model, and the covariance matrix of the Gaussian mixture model is modified to weaken the peak characteristic; S5: the reduced and modified Gaussian mixture model is used to process data, and the scheduling method is simulated.

2. The port power first scheduling disaggregation method considering load uncertainty according to claim 1, wherein the day-ahead scheduling in the step S1 is to quantize the characteristic difference between resources and demand to dynamically disaggregate scheduling instructions, realize collaborative peak shaving, and construct a scheduling framework for port planning processing considering micro-grid collaborative peak shaving auxiliary services, and the framework is established by the following steps: 1) the main power grid declares the peak shaving demand caused by energy fluctuation and uncertainty to the peak shaving auxiliary services; 2) the port declares the response power domain of the port power resources combined with the bidding capacity of the power resources and the peak shaving period, and obtains the scheduling instructions of the peak shaving demand; 3) the port quantifies the characteristic difference between the required power resources of the electric equipment and the planned demand, and calculates the response priority of the power resources to realize collaborative scheduling; 4) according to the characteristic matching of the power resources, a day-ahead scheduling disaggregation framework structure of the port is constructed, and the scheduling instructions of the port are disaggregated to realize collaborative scheduling and meet the peak shaving demand; 5) based on the clearing results, the port and the main power grid perform deviation evaluation; in the day-ahead scheduling, a multi-dimensional characteristic matrix is constructed based on the matching relationship between the operation characteristics of the power resources and the scheduling instructions, which is used to dynamically evaluate and adjust the response priority of each electric equipment; Further, in the day-ahead scheduling stage, a comprehensive cost function is constructed by combining the port peak shaving benefit and the response cost of the power resources, and multi-objective collaborative scheduling is realized through an optimization model.

3. The port power first scheduling disaggregation method considering load uncertainty according to claim 2, wherein the step S2: the total scheduling demand of the port is decomposed into each power resource through a characteristic matching method; first, a collaborative scheduling scheme is established, the total demand is iteratively allocated to available resources, and while meeting the basic operation constraints, the three targets of minimizing the exchange power with the power grid, optimizing the resource willingness cost and the port peak shaving benefit are considered; Then, a general feature matching model is constructed to map the operating characteristics of the power resources into a comparable format, and a discrete choice model is used to achieve priority matching of the resources and dispatch instructions, ensuring that the resources with the highest compatibility are prioritized to respond; Finally, the disaggregation and settlement of the dispatch instructions are realized by the mixed integer linear programming method, which synchronously calculates the resource response cost and the port peak shaving benefit, and realizes the optimized and efficient operation of the port system under the premise of ensuring power balance; The power resources include one or more of a port crane, a port shore power facility, an electric transfer vehicle, and an energy storage system.

4. The port power first dispatch disaggregation method considering net load uncertainty according to claim 3, wherein, The day-ahead dispatch step of the step S1 and the step S2: based on the peak shaving demand instruction of the main power grid, the information of the power resources required by the port is obtained, and the total dispatch instruction is dynamically decomposed and distributed to the power utilization equipment by quantifying the feature difference between the required power resources and the peak shaving demand; The dynamic decomposition and distribution of the total dispatch instruction to the power utilization equipment includes: constructing the matching relationship between the operating characteristics of the power resources and the demand characteristics of the dispatch instruction; based on the matching relationship, the response priority of each power utilization equipment is determined; and according to the response priority, an optimization model is used to disaggregate the dispatch instruction to the corresponding power equipment.

5. The port power first dispatch disaggregation method considering net load uncertainty according to claim 4, wherein, In the step S2, the construction of the day-ahead dispatch disaggregation framework of the port in the step S1 includes the acquisition, matching, and decomposition of the dispatch demand of the grid-connected microgrid; in addition, the multi-objective function of the proposed method is converted into a comprehensive cost function by quantifying the response cost of the power resources, the response willingness of the power resources, and the response peak shaving cost of the power resources; and equations and inequalities representing constraints are also established, covering traditional resources, energy storage, and peak shaving services.

6. The port power first dispatch disaggregation method considering net load uncertainty according to claim 5, wherein, In the second stage real-time dispatch, the step S3: based on the set of net load probability scenarios, the power deviation between prediction and actual and the flexibility ramping ability of each control subject are comprehensively considered, and optimization decision is made under the most adverse risk condition to ensure the operation stability and economy of the system in response to real-time load fluctuations; by setting the real-time control cost as a loss function and introducing auxiliary variables for each scenario, a risk-averse optimization model is established; The optimization model is a mixed integer linear programming model; Further, in the real-time dispatch in the step S3: based on the real-time uncertainty of the net load, each power utilization equipment is adjusted to respond to real-time power deviation with the objective of minimizing real-time operation control cost; In the S3 real-time dispatch, a robust optimization method is used for dispatch decision based on the set of net load probability scenarios to respond to the system operation risk under extreme fluctuation scenarios.

7. The port power first dispatch disaggregation method considering net load uncertainty according to claim 6, wherein, The step S4 creates a Gaussian mixture model, and a hierarchical expectation maximization algorithm based on density preservation is used to reduce and optimize the model.

8. The port power first-providing dispatching disaggregation method considering load uncertainty according to claim 7, characterized in that, The step S5 models the uncertainty of the load by using a probability model to optimize the dispatching decision in the real-time dispatching step S3; The step of modeling the uncertainty of the load by using a probability model includes collecting historical load data, fitting the probability distribution of the load data by using a Gaussian mixture model, and reducing the Gaussian mixture model to reduce the calculation complexity; In the step S5, the real-time dispatching step generates multiple load probability scenarios based on the Gaussian mixture model and makes a dispatching decision based on the worst-case risk condition.

Images