A kind of micro-capacitance sensor wind-light storage model predictive control method
Technical field
The present invention relates to a kind of micro-capacitance sensor wind-light storage model predictive control method.
Background technology
Distributed power generation has resources and environment close friend, the advantage such as flexible of powering, and distributed power generation developed on the basis of centralized generating and bulk power grid, has become the inexorable trend of domestic and international intelligent grid development.According to national energy development plan, will reach 9% to the year two thousand twenty China distributed power generation installed capacity accounting, be the important component part in supply of electric power system.For solving the grid-connected problem brought of distributed power generation, the concept of micro-capacitance sensor is proposed.Comprehensive achievement in research both domestic and external, micro-capacitance sensor refers in a distributed manner based on generation technology, based on regenerative resource, utilizes energy storage and control device, realizes the miniature supply network of network internal balance of electric power and ener.
The uncertainty that wind-powered electricity generation and photovoltaic are exerted oneself is large, and be difficult to prediction, its precision of prediction is also very low, and the prediction time is in advance longer, and its predicated error is larger.The uncertainty that the access of wind-powered electricity generation and photovoltaic makes micro-capacitance sensor run increases, and existing grid control method requires higher to the precision of forecasting model of uncertain course, therefore in the urgent need to seek can the control method of better coping with uncertainty.
Model Predictive Control (model predictive control, MPC) is the effective way addressed this problem.Model Predictive Control is widely used in industrial stokehold always, and it is comparatively strong to the adaptability of model and robustness, is applicable to very much the large problem of answering system model uncertainty.Model Predictive Control is the finite time-domain closed loop optimal control algorithm of a class based on model in essence, in each sampling period, controller is using the system mode of current time as the initial condition controlled, based on forecast model predicting the outcome to to-be, solving one by rolling online have the optimal control problem of limit thus obtain current controlling behavior, making following output minimum with the difference of reference locus.
Summary of the invention
The present invention is in order to solve the problem, propose a kind of micro-capacitance sensor wind-light storage model predictive control method, this method uses for reference the thought of Model Predictive Control, establish the Controlling model of prediction-on-line optimization-feedback, by the maximum output of Wind turbines and photovoltaic generation in micro-capacitance sensor in forecast model prediction following certain period; The wind-powered electricity generation predicted by forecast model and photovoltaic maximum output, as constraints, are exerted oneself to Wind turbines, photovoltaic generation and energy-storage battery three in micro-capacitance sensor and are carried out on-line optimization, and the reference providing three in following certain period is exerted oneself; According to the real-time, tunable capacity of Wind turbines and photovoltaic generation, instantaneously with reference to exerting oneself, feedback adjusting is carried out to wind-light storage three in control.Embodiment analysis shows, the method can tackle the uncertainty that wind-powered electricity generation and photovoltaic are exerted oneself preferably, effectively improves the operation characteristic of micro-capacitance sensor.
To achieve these goals, the present invention adopts following technical scheme:
A kind of micro-capacitance sensor wind-light storage model predictive control method, comprises the following steps:
(1) forecast model is set up, by the maximum output of Wind turbines and photovoltaic generation in micro-capacitance sensor in the forecast model prediction following setting period;
(2) using the wind-powered electricity generation predicted and photovoltaic maximum output as constraints, exert oneself to Wind turbines, photovoltaic generation and energy-storage battery three in micro-capacitance sensor and carry out on-line optimization, the reference providing three is exerted oneself;
(3) according to the real-time, tunable capacity of Wind turbines and photovoltaic generation, with reference to exerting oneself, feedback adjusting is carried out to Wind turbines, photovoltaic generation and energy-storage battery three.
In described step (1), forecast model realizes the power prediction of Wind turbines and photovoltaic generation by neural network technology, according to the historical information of process and the following output valve of following input prediction process.
In the present invention, the method for building up of forecast model is neural network technology, and the method process is for known by the industry.
The target function of the on-line optimization model in described step (2): target function f
1for micro-capacitance sensor and power distribution network, to exchange power deviation minimum, to ensure that micro-capacitance sensor becomes the stable power supply of power distribution network or load, reduces the control difficulty of power distribution network, and at micro-capacitance sensor under net state, arranging micro-capacitance sensor and power distribution network, to exchange power be 0, target function f
2for the difference of energy-storage battery first end period state-of-charge is minimum, to ensure that energy-storage battery has higher electricity;
minf
2=SOC(1)-SOC(N)
Wherein, N optimizes time hop count total in time domain; P
li () is the load power in period i; P
tiei () is the exchange power of micro-capacitance sensor and power distribution network in period i, on the occasion of expression micro-capacitance sensor to power distribution network injecting power, negative value represents that power distribution network is to micro-capacitance sensor injecting power; P
wi () is exerted oneself for the plan of Wind turbines in period i; P
pVi () is exerted oneself for the plan of photovoltaic generation at period i; P
bi () is exerted oneself for the plan of energy-storage battery in period i, during charging this value be on the occasion of, during electric discharge, this value is negative value; SOC (1), SOC (N) are the energy-storage battery state-of-charge of the 1st period, N period.
The constraints of the on-line optimization model in described step (2): comprise that Wind turbines is exerted oneself, photovoltaic generation is exerted oneself and energy-storage battery discharge and recharge count constraint:
0<P
w(i)<P
w_pre(i)
0<P
PV(i)<P
PV_pre(i)
N
ch_dis<N
battery
Wherein, P
w_prei () to be exerted oneself predicted value for the Wind turbines in period i; P
pv_prei () to be exerted oneself predicted value for the photovoltaic generation in period i; N
ch_disfor energy-storage battery discharge and recharge number of times; N
batteryfor energy-storage battery maximum permission discharge and recharge number of times.
The derivation algorithm concrete grammar of the on-line optimization model in described step (2) is: this Optimized model comprises 2 target functions, belong to multi-objective optimization question, multi-objective optimization algorithm has 3 Performance Evaluating Indexes: 1. tried to achieve solution will as far as possible close to Pareto optimal solution; 2. to keep distributivity and the diversity of separating colony as far as possible; 3. to prevent the Pareto optimal solution obtained from losing in solution procedure, adopt NSGA-II Algorithm for Solving on-line optimization model.
The concrete grammar of the real-time, tunable capacity in described step (3) is: Wind turbines real-time, tunable capacity △ P
wfor:
ΔP
w=P
w_est-P
w
Wherein, P
wfor Wind turbines is with reference to exerting oneself; P
w_estfor Wind turbines is exerted oneself estimated value in real time, this estimated value calculates fast by real-time wind speed and fan operation state;
Photovoltaic generation real-time, tunable capacity △ P
pvfor:
ΔP
pv=P
pv_est-P
pv
Wherein, P
pvfor photovoltaic generation is with reference to exerting oneself; P
pv_estfor photovoltaic generation is exerted oneself estimated value in real time, this estimated value by real-time lighting intensity and temperature computation out.
Feedback adjusting in described step (3): according to Wind turbines and photovoltaic generation with or without variable capacity, feedback adjusting link specifically comprises following 4 kinds of situations:
A () Wind turbines and photovoltaic generation all have variable capacity, namely variable capacity is just, are now directly handed down to Wind turbines, photovoltaic generation and energy-storage battery with reference to exerting oneself;
B () Wind turbines and photovoltaic generation are all without variable capacity, namely variable capacity is negative, are now directly handed down to Wind turbines and photovoltaic generation with reference to exerting oneself, energy-storage battery compensation power vacancy;
C () Wind turbines has variable capacity, photovoltaic generation is without variable capacity, and photovoltaic generation has superfluous plan, now the plan of photovoltaic generation surplus is transferred to Wind turbines;
D () photovoltaic generation has variable capacity, Wind turbines is without variable capacity, and Wind turbines has superfluous plan, now the plan of Wind turbines surplus is transferred to photovoltaic generation.
In described step (c), with reference to exerting oneself, adjustment is as follows:
P
w_sch=P
w+min(ΔP
w,-ΔP
pv)
P
pv_sch=P
pv-min(ΔP
w,-ΔP
pv)
Wherein, P
w_schfor Wind turbines last minute planning is exerted oneself; P
pv_schfor photovoltaic generation last minute planning is exerted oneself.
In described step (d), with reference to exerting oneself, adjustment is as follows:
P
pv_sch=P
pv+min(ΔP
pv,-ΔP
w)
P
w_sch=P
w_est-min(ΔP
pv,-ΔP
w)。
Beneficial effect of the present invention is:
(1) use for reference the thought of Model Predictive Control, the invention provides a kind of micro-capacitance sensor wind-light storage model predictive control method, establish the Controlling model of prediction-on-line optimization-feedback;
(2) this control method have employed the on-line optimization strategy be based upon on actual output feedack basis, makes control procedure can make correction to the impact of predicated error in time;
(3) compared with traditional grid control method, this control method reduces the requirement of the precision of forecasting model to uncertain course, compensate for the insoluble wind-powered electricity generation of traditional control method and photovoltaic precision of forecasting model is low, uncertain strong defect of exerting oneself, effectively improve the operation characteristic of micro-capacitance sensor.
Accompanying drawing explanation
Fig. 1 is micro-capacitance sensor wind-light storage model predictive control method schematic flow sheet provided by the invention;
Fig. 2 is wind-powered electricity generation prediction maximum output and actual maximum output curve synoptic diagram in the embodiment of the present invention;
Fig. 3 is photovoltaic prediction maximum output and actual maximum output curve synoptic diagram in the embodiment of the present invention;
Fig. 4 is that in the embodiment of the present invention, wind-light storage is optimized with reference to power curve schematic diagram;
Fig. 5 is honourable with reference to power curve schematic diagram after feedback adjusting in the embodiment of the present invention;
Fig. 6 is the real-time power curve schematic diagram of wind-light storage in the embodiment of the present invention;
Fig. 7 is energy-storage battery SOC change curve schematic diagram in the embodiment of the present invention.
Embodiment:
Below in conjunction with accompanying drawing and embodiment, the invention will be further described.
A kind of micro-capacitance sensor wind-light storage model predictive control method, comprises the steps:
Step (1): by the maximum output of Wind turbines and photovoltaic generation in micro-capacitance sensor in forecast model prediction following certain period;
Step (2): the wind-powered electricity generation predict step (1) and photovoltaic maximum output are as constraints, exert oneself to Wind turbines, photovoltaic generation and energy-storage battery three in micro-capacitance sensor and carry out on-line optimization, the reference providing three in following certain period is exerted oneself;
Step (3): according to the real-time, tunable capacity of Wind turbines and photovoltaic generation, is controlling to carry out feedback adjusting to the wind-light storage three of step (2) with reference to exerting oneself instantaneously.
Forecast model in described step (1): the function of forecast model is the following output valve of historical information according to process and following input prediction process, for the optimization of Model Predictive Control provides priori.Forecast model only focuses on the function of model, and does not focus on the form of model, as long as have the model of the following dynamic function of prognoses system, no matter which type of form of expression it has, and all can be used as forecast model.In the present invention, forecast model realizes the power prediction of Wind turbines and photovoltaic generation by neural network technology.
The target function of the on-line optimization model in described step (2): target function f
1for micro-capacitance sensor and power distribution network, to exchange power deviation minimum, to ensure that micro-capacitance sensor becomes the stable power supply of power distribution network or load, reduces the control difficulty of power distribution network.At micro-capacitance sensor under net state, arranging micro-capacitance sensor and power distribution network, to exchange power be 0.Target function f
2for the difference of energy-storage battery first end period state-of-charge is minimum, to ensure that energy-storage battery has higher electricity.
minf
2=SOC(1)-SOC(N)
Wherein, N optimizes time hop count total in time domain; P
li () is the load power in period i; P
tiei () is the exchange power of micro-capacitance sensor and power distribution network in period i, on the occasion of expression micro-capacitance sensor to power distribution network injecting power, negative value represents that power distribution network is to micro-capacitance sensor injecting power; P
wi () is exerted oneself for the plan of Wind turbines in period i; P
pVi () is exerted oneself for the plan of photovoltaic generation at period i; P
bi () is exerted oneself for the plan of energy-storage battery in period i, during charging this value be on the occasion of, during electric discharge, this value is negative value; SOC (1), SOC (N) are the energy-storage battery state-of-charge of the 1st period, N period.
The constraints of the on-line optimization model in described step (2): comprise that Wind turbines is exerted oneself, photovoltaic generation is exerted oneself and energy-storage battery discharge and recharge count constraint.
0<P
w(i)<P
w_pre(i)
0<P
PV(i)<P
PV_pre(i)
N
ch_dis<N
battery
Wherein, P
w_prei () to be exerted oneself predicted value for the Wind turbines in period i; P
pv_prei () to be exerted oneself predicted value for the photovoltaic generation in period i; N
ch_disfor energy-storage battery discharge and recharge number of times; N
batteryfor energy-storage battery maximum permission discharge and recharge number of times, be traditionally arranged to be 1.
The derivation algorithm of the on-line optimization model in described step (2): this Optimized model comprises 2 target functions, belongs to multi-objective optimization question.Multi-objective optimization algorithm has 3 main Performance Evaluating Indexes: 1. tried to achieve solution will as far as possible close to Pareto optimal solution; 2. to keep distributivity and the diversity of separating colony as far as possible; 3. to prevent the Pareto optimal solution obtained from losing in solution procedure.Correspondingly, nondominated sorting genetic algorithm II (NSGA-II) has 3 key technologies to become a kind of outstanding multi-objective optimization algorithm, i.e. quick non-dominated ranking, individual crowding distance and elitism strategy, therefore the present invention adopts NSGA-II Algorithm for Solving on-line optimization model.
Real-time, tunable capacity in described step (3): the on-line optimization stage is relative to the control moment, its lead is larger, predicated error is also larger, its Wind turbines optimizing out and photovoltaic generation may higher than the actual maximum output of Wind turbines and photovoltaic generation with reference to power curve, cause cannot accurately completing with reference to power curve, add the power back-off pressure of energy-storage battery.In order to improve Wind turbines and photovoltaic generation with reference to the completeness of exerting oneself, controlling the real-time, tunable capacity of moment according to Wind turbines and photovoltaic generation, with reference to exerting oneself, feedback adjusting being carried out to the two, thus makes energy-storage battery as far as possible according to on-line optimization curve motion.
Wind turbines real-time, tunable capacity △ P
wfor:
ΔP
w=P
w_est-P
w
Wherein, P
wfor Wind turbines is with reference to exerting oneself; P
w_estfor Wind turbines is exerted oneself estimated value in real time, this estimated value calculates fast by real-time wind speed and fan operation state.
Photovoltaic generation real-time, tunable capacity △ P
pvfor:
ΔP
pv=P
pv_est-P
pv
Wherein, P
pvfor photovoltaic generation is with reference to exerting oneself; P
pv_estfor photovoltaic generation is exerted oneself estimated value in real time, this estimated value calculates fast by real-time lighting intensity and temperature etc.
Feedback adjusting in described step (3): according to Wind turbines and photovoltaic generation with or without variable capacity, feedback adjusting link specifically comprises following 4 kinds of situations.
(1) Wind turbines and photovoltaic generation all have variable capacity, and namely variable capacity is just, are now directly handed down to Wind turbines, photovoltaic generation and energy-storage battery with reference to exerting oneself.
(2) Wind turbines and photovoltaic generation are all without variable capacity, and namely variable capacity is negative, are now directly handed down to Wind turbines and photovoltaic generation with reference to exerting oneself, energy-storage battery compensation power vacancy.
(3) Wind turbines has variable capacity, and photovoltaic generation is without variable capacity, and photovoltaic generation has superfluous plan, and now the plan of photovoltaic generation surplus is transferred to Wind turbines, with reference to exerting oneself, adjustment is as follows:
P
w_sch=P
w+min(ΔP
w,-ΔP
pv)
P
pv_sch=P
pv-min(ΔP
w,-ΔP
pv)
Wherein, P
w_schfor Wind turbines last minute planning is exerted oneself; P
pv_schfor photovoltaic generation last minute planning is exerted oneself.
(4) photovoltaic generation has variable capacity, and Wind turbines is without variable capacity, and Wind turbines has superfluous plan, and now the plan of Wind turbines surplus is transferred to photovoltaic generation, with reference to exerting oneself, adjustment is as follows:
P
pv_sch=P
pv+min(ΔP
pv,-ΔP
w)
P
w_sch=P
w_est-min(ΔP
pv,-ΔP
w)。
According to the micro-capacitance sensor wind-light storage model predictive control method flow process shown in Fig. 1, work out micro-capacitance sensor wind-light storage Model Predictive Control Algorithm and realized program.In embodiment, test parameter arranges as follows: fan capacity is 66kW, and photovoltaic capacity is 200kW, and energy storage is 90kW/270kWh, and load capacity is 120kW, and it is 60kW that power distribution network and micro-capacitance sensor exchange power, and the initial SOC of energy-storage battery is 0.5.
The prediction maximum output of Wind turbines and photovoltaic generation and actual maximum output curve are respectively as shown in Figures 2 and 3.The predicted value of wind power is bigger than normal, is greater than actual maximum output within all the period of time.The predicting the outcome of photovoltaic generation can be coincide the change of actual maximum output preferably, and precision of prediction is relatively high.
When arranging on-line optimization, hop count is 15, and each period duration is 1 minute, and optimizing total duration is 15 minutes.Adopt NSGA-II algorithm to be optimized, population number is 400, and algebraically is 200.Optimized algorithm consuming time 1 minute 10 seconds, can meet the requirement of On-line Control.Typical Pareto prioritization scheme is as shown in table 1.
The typical Pareto prioritization scheme of table 1
|
f
1/kW
|
f
2 |
Scheme 1 |
0.1613 |
0.4907 |
Scheme 2 |
2.7910 |
0.6017 |
With target function f
1for main preference, target function f
2for secondary preference, selection scheme 1 is optimal control scheme.In scheme 1, the optimization of Wind turbines, photovoltaic generation and energy-storage battery with reference to power curve as shown in Figure 4.
Exert oneself according to the reference of the conditions such as real-time wind speed, illumination and temperature to Wind turbines and photovoltaic generation and adjust, as shown in Figure 5, power surplus plan is transferred to photovoltaic generation to the reference power curve after adjustment by Wind turbines.
Exerting oneself in real time of Wind turbines and photovoltaic generation can be exerted oneself in the reference preferably after tracking adjustment, as shown in Figure 6.Energy-storage battery can run according to the reference power curve that on-line optimization is given substantially, and effectively reduce the power back-off pressure of energy-storage battery, energy-storage battery discharge process as shown in Figure 7.
By reference to the accompanying drawings the specific embodiment of the present invention is described although above-mentioned; but not limiting the scope of the invention; one of ordinary skill in the art should be understood that; on the basis of technical scheme of the present invention, those skilled in the art do not need to pay various amendment or distortion that creative work can make still within protection scope of the present invention.