Adjustable resource real-time control method suitable for port shore power energy system

Abstract

The invention provides an adjustable resource real-time control method suitable for a port shore power energy system. The method comprises the following steps: step (1), modeling a key energy sub-module in the system; (2) constructing a prediction model by adopting a generalized regression neural network; (3) an economic layer is used for dynamically constructing an economic cost optimization model according to the real-time electricity price, the battery state and the photovoltaic output condition after the short-term prediction load curve is obtained, and generating an optimal reference charging and discharging track of the energy storage system; and step (4), a control layer is an execution layer in the energy management system, a model prediction control method is adopted, an optimal control strategy is calculated in real time according to the optimal state generated by the economic layer, and accurate tracking and power balance adjustment of the state of the energy storage system are realized. The invention provides the adjustable resource real-time control method suitable for the port shore power energy system, and the method has good prediction precision, economic optimization capability and control responsiveness.

Description

TECHNICAL FIELD

[0001] The present application belongs to the technical field of power control, and particularly relates to a real-time control method for adjustable resources suitable for a port shore power energy system. BACKGROUND

[0002] In recent years, the ship transportation volume of the ports in the country has been continuously increasing, and the energy consumption of the ships during the berthing period has been increasingly concerned. Traditionally, the ships usually rely on the auxiliary engine fuel power generation to maintain the operation of the equipment on the ships during the berthing period, and this method has problems such as low energy efficiency, high emission, and large noise. According to research, the carbon emission generated by the ships during the berthing period accounts for more than 30% of the overall emission of the port, and becomes the main source of greenhouse gases and pollutants in the port area. Therefore, the shore power technology emerges as the times require, and by providing clean power to replace the power supply of the auxiliary engine of the ship, the emission is reduced, the air quality is improved, and the energy utilization efficiency is improved.

[0003] At present, the power supply mode of the shore power system mainly relies on direct power supply. However, with the rapid development of renewable energy technology (especially photovoltaic power generation) and the continuous decrease of the cost of energy storage system, the integration of the port shore power system and the renewable energy system becomes possible. By constructing a hybrid energy supply system complementary to “photovoltaic-energy storage-grid”, not only the dependence on the power grid can be reduced, but also the peak-valley price arbitrage and clean energy preferential consumption can be realized.

[0004] However, in actual operation, due to the high uncertainty of the ship arrival time and the berthing time, the shore power load presents obvious dynamic non-stationary characteristics; at the same time, the photovoltaic power generation is greatly affected by the weather, and the capacity of the energy storage system is limited, so how to accurately predict the shore power load of the ship under the background of limited historical data and frequent working conditions, how to develop an economic and efficient charging and discharging strategy to realize the reasonable allocation of power resources, and how to ensure the safety of power supply and the stability of energy control under system disturbance and real-time fluctuation, have become key problems to be solved in the dispatching management of the shore power hybrid energy system.

[0005] Therefore, it is urgent to develop a comprehensive dispatching method with prediction ability, economic optimization ability and real-time control ability to adapt to the complex characteristics of the uncertainty, multi-source coupling and real-time operation of the shore power hybrid energy system. SUMMARY

[0006] The present application aims to solve the problems of large load uncertainty, severe fluctuation of renewable energy and limited capacity of energy storage system in the operation of the port shore power system, and proposes a real-time energy management method for multi-source integration of “photovoltaic-energy storage-grid”, which has good prediction accuracy, economic optimization ability and control responsiveness.

[0007] To this end, the application constructs a three-layer energy scheduling framework composed of a prediction layer, an economic optimization layer and a control layer to improve the intelligent scheduling ability and operation efficiency of the system in an uncertain environment.

[0008] In the prediction layer, a generalized regression neural network (GRNN) is used to make short-term prediction of the shore power load during the future port call of the ship, which can effectively deal with the real scenario of sparse samples and frequent fluctuations.

[0009] In the economic layer, a comprehensive cost function considering multiple factors such as electricity purchase cost, energy storage loss, photovoltaic income and battery attenuation is constructed, and operation constraints such as energy balance, grid power upper limit, energy storage power range and SOC boundary are introduced to realize economic scheduling under multiple energy paths.

[0010] In the control layer, a model predictive control (MPC) strategy is introduced, combined with a linear state space model, based on the economic layer reference trajectory for rolling optimization, to solve the optimal control instructions of energy storage and grid output in real time, ensuring the stable operation of the system under load disturbance and light change.

[0011] Through the above multi-layer integrated method, the application realizes efficient, low-carbon and intelligent control of the multi-source shore power system, which has stronger prediction ability, better economy and higher control accuracy than traditional methods, and is suitable for various port shore power systems with renewable energy access.

[0012] Compared with the prior art, the beneficial effects are:

[0013] By introducing a three-layer strategy integrating prediction, optimization and control, efficient scheduling and stable energy supply of the shore power system in an uncertain environment are realized, and the economy and renewable energy utilization rate are significantly improved. BRIEF DESCRIPTION OF DRAWINGS

[0014] Figure 1 A flowchart of the adjustable resource real-time control method for the port shore power energy system according to the application;

[0015] Figure 2 A regression neural network structure diagram according to the application. DETAILED DESCRIPTION

[0016] The specific embodiments of the adjustable resource real-time control method for the port shore power energy system according to the application will be described in detail below with reference to the accompanying drawings.

[0017] As shown in Figure 1 The adjustable resource real-time control method according to the application includes the following steps:

[0018] Step (1): In order to realize the above energy management method, the present application firstly models the key energy sub-modules in the system, including the photovoltaic power generation system and the energy storage system, and constructs a system mathematical model to support the subsequent prediction and optimization control strategy design.

[0019] (I) Photovoltaic system modeling:

[0020] The photovoltaic system converts solar energy into direct current power available to the shore power system, and its output power is affected by factors such as solar irradiance and ambient temperature. The output power model of photovoltaic power generation is as follows:

[0021]

[0022] where P pv represents the current photovoltaic output power; P STC represents the photovoltaic output power under standard test conditions (STC); G t and G STC represent the current and standard solar irradiance, respectively; T t and T STC represent the current and standard temperature, respectively; α T and γ T represent the temperature coefficient and nonlinear temperature compensation coefficient of the photovoltaic module, respectively.

[0023] (II) Energy storage system modeling:

[0024] The energy storage system uses lithium-ion batteries, whose state is represented by SOC (State of Charge). The SOC is estimated by ampere-hour integration method, and its dynamic expression is:

[0025]

[0026] Meanwhile, the following constraints are satisfied:

[0027] SOC min ≤ SOC t ≤ SOC max

[0028] where P represents the power of the energy storage system at time t (positive value for discharging, negative value for charging); I bat represents the battery current, where positive value represents discharging and negative value represents charging; Δt is the time step, η dis and η ch are the discharging efficiency and charging efficiency; Q max is the maximum power capacity of the energy storage device; SOC min and SOC max are the upper and lower limits of SOC.

[0029] Step (2): In order to improve the prediction accuracy of shore power load in the case of uncertainty of ship port time and scarcity of historical data, the present application adopts a generalized regression neural network (GRNN) to construct a prediction model, as shown in Figure 2 The GRNN model has good nonlinear mapping capability and can maintain high accuracy in the case of small samples, and its network structure includes an input layer, a pattern layer, a summation layer and an output layer. The core function of the GRNN model is a radial basis function, and its output is expressed as:

[0030]

[0031] Where x represents historical data input, N represents the total number of training samples; c j is the center of the jth training sample; sigma is a smoothing factor; y j is the target output corresponding to the sample c j . is the prediction result.

[0032] The GRNN model is constructed in a four-layer structure, including an input layer, a pattern layer, a summation layer and an output layer in sequence, wherein the input layer is used to receive multi-dimensional time characteristics, the pattern layer contains multiple radial basis units, the similarity between the current input and the sample center is calculated using a Gaussian kernel function, the summation layer is used to perform weighted normalization processing on the output of all pattern layer neurons, and the output layer finally generates multi-step prediction values.

[0033] Step (3): The economic layer is used to dynamically construct an economic cost optimization model and generate the optimal reference charge-discharge trajectory of the energy storage system according to the real-time electricity price, battery state and photovoltaic output after obtaining the short-term predicted load curve. The optimization objective not only considers the current operation cost of the system, but also introduces the battery life factor and the peak-valley electricity price difference factor, so as to realize the balance between the minimum operation cost and the battery health degree.

[0034] Specifically, the objective function of the economic layer is as follows:

[0035]

[0036] Where: J cost represents the overall objective function, is the grid purchase cost; is the operation loss of photovoltaic and energy storage system; is the photovoltaic grid-connected income; is the peak-valley electricity price arbitrage income of the energy storage system.

[0037] In addition, the following constraint conditions are introduced in the economic layer optimization model:

[0038] (1) Power balance constraint (supply-demand matching):

[0039] To ensure that the shore power system is in power balance at any time, i.e. total power supply equals total load, the following relationship needs to be met:

[0040]

[0041] wherein, is the grid power supply, is the photovoltaic output power, is the energy storage power, is the load power.

[0042] (2) Grid power constraint (dispatching capacity limit):

[0043] To ensure the safe operation of the grid side, the grid power input needs to be controlled within the dispatching capacity range:

[0044]

[0045] wherein, respectively represent the minimum and maximum adjustable power of the grid side.

[0046] (3) Energy storage power and SOC constraint (operation safety and battery life):

[0047] To ensure that the battery charging and discharging behavior is within the safe range, the power and SOC of the battery need to be set as follows:

[0048]

[0049] wherein, and are the energy storage charging and discharging power, and respectively represent the maximum charging and discharging power of the energy storage.

[0050] SOC min ≤ SOC t ≤ SOC max

[0051] SOC t is the state of charge of the battery at time t, the upper and lower limits are SOC min and SOC max .

[0052] Step (4): The control layer is the execution layer in the energy management system of the present application, which adopts the model predictive control (MPC) method, calculates the optimal control strategy in real time according to the optimal state generated by the economic layer, and realizes the accurate tracking and power balance adjustment of the energy storage system state.

[0053] In this layer, the system state vector x(k), the control input vector u(k), and the system output vector y(k) are defined, and a linear state space model is constructed as follows:

[0054] x(k+1) = Ax(k) + Bu(k)

[0055] y(k) = Cx(k)

[0056] where the matrices A, B and C represent the mapping relationship of state transition, control input and system output, respectively.

[0057] The control objective is to minimize the deviation between the predicted output y(k+i) and the economic layer reference trajectory y ref (k+i), so the following rolling optimization objective function is constructed:

[0058]

[0059] where N p represents the prediction time domain length; N c represents the control time domain length; and λ is the control smoothing weight factor.

[0060] The control layer adopts an optimization execution mechanism of "rolling time domain + feedback correction", that is, an optimization calculation is performed once in each control period, only the control result of the first step is applied, and the system actual state is updated. This method can significantly improve the running robustness and response speed of the system in complex environments such as fluctuating load and uncertain light, and guarantee the economic and stable operation of the port shore power system.

[0061] The optimization problem constructed by the present application belongs to a typical constrained quadratic programming problem, that is, to find the optimal control strategy that satisfies various operating boundary conditions under the given objective function. Specifically, the problem dynamically generates the coefficient matrix and constraint boundary of the optimization model according to the current system state, economic reference trajectory and physical constraint conditions in each control period, and then calls an efficient online solver for real-time calculation. The control variables finally solved are: the power of the energy storage system and the power of the power grid. In the solving process, the system considers the limits of energy storage charging and discharging power, power grid scheduling capacity, and upper and lower limits of battery state of charge, and outputs the optimal control decision at the current time under the premise of meeting all boundary constraints.

[0062] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present application but not to limit them. Those skilled in the art should understand that the specific embodiments of the present application can be modified or replaced by equivalents, and these modifications or replacements are within the protection scope of the claims.

Claims

1. A real-time control method for adjustable resources applicable to a port shore power energy system, characterized in that: The adjustable resource real-time control method comprises the following steps: Step (1): First, model the key energy submodules in the system, including the photovoltaic power generation system and the energy storage system, and construct a system mathematical model to support the subsequent prediction and optimization control strategy design; Step (2): Use generalized regression neural network (GRNN) to build a prediction model; Step (3): After obtaining the short-term forecast load curve, the economic layer is used to dynamically build an economic cost optimization model based on the real-time electricity price, battery status, and photovoltaic output, and generate the optimal reference charge and discharge trajectory of the energy storage system; Step (4): The control layer is the execution layer in the energy management system of the present invention. It adopts the model predictive control (MPC) method to calculate the optimal control strategy in real time according to the optimal state generated by the economic layer, so as to achieve accurate tracking of the energy storage system state and power balance adjustment.

2. The method for real-time control of adjustable resources applicable to a port shore power energy system according to claim 1, characterized in that: Photovoltaic system modeling in step (1): The output power model of photovoltaic power generation is as follows: Among them, P pv Indicates the current photovoltaic output power; P STC Indicates the photovoltaic output power under standard test conditions (STC); G t and G STC Represent the current and standard solar irradiance respectively; T t and T STC Represents the current and standard temperatures respectively; α T and γ T They represent the temperature coefficient and nonlinear temperature compensation coefficient of the photovoltaic module respectively.

3. The method for real-time control of adjustable resources applicable to a port shore power energy system according to claim 2, characterized in that: Energy storage system modeling in step (1): The energy storage system uses lithium-ion batteries, whose state is represented by SOC (State of Charge). The SOC is estimated by the ampere-hour integration method, and its dynamic expression is: The following constraints must be met at the same time: SOC min ≤SOC t ≤SOC max in Represents the power of the energy storage system at time t (positive value means discharge, negative value means charge); I bat represents the battery current, where a positive value indicates discharge and a negative value indicates charge; Δt is the time step, η dis and η ch is the discharge efficiency and charging efficiency; Q max It is the maximum power capacity of the energy storage device; SOC min With SOC max are the upper and lower limits of SOC.

4. The method for real-time control of adjustable resources applicable to a port shore power energy system according to claim 1, characterized in that: The GRNN model network structure in step (2) includes an input layer, a pattern layer, a summation layer, and an output layer. The core function of the GRNN model is a radial basis function, and its output is expressed as: Where x represents the historical data input, N represents the total number of training samples; c j is the center of the jth training sample; σ is the smoothing factor; y j For sample c j The corresponding target output; To predict the results; The GRNN model is constructed with a four-layer structure, which includes an input layer, a pattern layer, a summation layer and an output layer in sequence. The input layer is used to receive multidimensional time features, the pattern layer contains multiple radial basis units, and uses a Gaussian kernel function to calculate the similarity between the current input and the sample center. The summation layer is used to perform weighted normalization processing on the outputs of all pattern layer neurons, and the output layer finally generates multi-step prediction values.

5. The method for real-time control of adjustable resources applicable to a port shore power energy system according to claim 1, characterized in that: The objective function of the economic layer in step (3) is as follows: Among them: J cost represents the overall objective function, the cost of purchasing electricity for the grid; Operation losses of photovoltaic and energy storage systems; For photovoltaic grid-connected income; It is the arbitrage income of peak and valley electricity prices for the energy storage system.

6. A method for real-time control of adjustable resources applicable to a port shore power energy system according to claim 5, characterized in that: The following constraints are introduced into the economic layer optimization model: (1) Power balance constraints: In order to ensure that the shore power system is in a power balance state at any time, that is, the total power supply is equal to the total load, the following relationship must be met: in, Power supply to the grid, is the photovoltaic output power, is the energy storage power, is the load power; (2) Grid power constraints: To ensure safe operation on the grid side, it is necessary to control the grid power input within its dispatching capacity: in, Respectively represent the minimum and maximum adjustable power on the grid side; (3) Energy storage power and SOC constraints: To ensure that the battery charging and discharging behavior is within the safe range, the following physical boundaries need to be set for its power and SOC: in, and are the energy storage charging and discharging power, and Respectively represent the maximum charging and discharging power of energy storage; SOC min ≤SOC t ≤SOC max SOC t is the state of charge of the battery at time t, with upper and lower limits of SOC min and SOC max .

7. The method for real-time control of adjustable resources applicable to a port shore power energy system according to claim 1, characterized in that: In step (4), the system state vector x(k), control input vector u(k), and system output vector y(k) are defined, and the following linear state space model is constructed: x(k+1)=Ax(k)+Bu(k) y(k)=Cx(k) Among them, matrices A, B and C represent the mapping relationship between state transition, control input and system output respectively; The control objective is to minimize the difference between the predicted output y(k+i) and the economic layer reference trajectory y ref The deviation between (k+i), so the following rolling optimization objective function is constructed: where N p Indicates the length of the prediction time domain; N c represents the length of the control time domain; λ is the control smoothing weight factor.

8. The method for real-time control of adjustable resources applicable to a port shore power energy system according to claim 7, characterized in that: The control layer adopts the optimization execution mechanism of "rolling time domain + feedback correction", that is, an optimization calculation is performed once in each control cycle, only the control result of the first step is applied, and it is updated according to the actual state of the system.

Images