Smoothing ramp events in wind farm based on dynamic programming in energy internet

Abstract

The concept of energy internet has been gradually accepted, which can optimize the consumption of fossil energy and renewable energy resources. When wind power is integrated into the main grid, ramp events caused by stochastic wind power fluctuation may threaten the security of power systems. This paper proposes a dynamic programming method in smoothing ramp events. First, the energy internet model of wind power, pumped storage power station, and gas power station is established. Then, the optimization problem in the energy internet is transformed into a multi-stage dynamic programming problem, and the dynamic programming method proposed is applied to solve the optimization problem. Finally, the evaluation functions are introduced to evaluate pollutant emissions. The results show that the dynamic programming method proposed is effective for smoothing wind power and reducing ramp events in energy internet.

Keywords

energy internet / wind power / ramp events / dynamic programming

Introduction

With the high penetration of renewable energy in power systems, the concept of energy internet has been proposed, which can manage energy and share power among different renewable energy resources [¹]. From the view of global energy allocation, energy internet can be defined as the advanced stage of the smart grid, which is based on renewable energy, various types of power sources, and flexible access of users. The topics of energy internet can be classified into four layers, such as the energy layer, data layer, business layer, and application layer. The research topics in energy layer include energy planning, energy complementation systems, energy management systems, and integrated energy systems. The research topics in data layer include artificial intelligence, information systems, big data technology, etc. The topics in the business layer include market modeling, energy exchange, block chains, etc. The topics in the application layer include flexible DC technology, intelligent equipment, industrial internet, intelligent cities, etc.

This paper focuses on the smoothing ramp events caused by wind power fluctuation in 15 min, 30 min, or 1 h. Ramp events can be defined as abnormal wind power fluctuation threatening system operation possibly in a short period [²]. When power change is upward, the events are called ramp-up events; when power change is downward, the events are referred to as ramp-down events.

The common definition for ramp events are as follows [³–⁷]:

(1) Whether the absolute value of wind power difference between time t + Dt and t exceeds the given threshold value p_(tsd) in period time Dt.

(1)\left|p_{(t+\Delta t)}-p_{\left(t\right)}\right| > p_{\left(\mathrm{tsd}\right)}.

Equation (1) focuses on the wind power values between the starting and ending points of the period of time, and ignores the changes during this period.

(2) Whether the difference between the maximum and minimum wind power values exceeds the given threshold value in a period of time.

(2)\begin{array}{l}\max p_{\left(t\mathrm{1}\right)}-\min p_{\left(t\mathrm{2}\right)} > p_{\left(\mathrm{tsd}\right)},\hfill \\ t\mathrm{1} ,t\mathrm{2}\in [t,t+\Delta t]. \hfill \end{array}

Equation (2) only focuses on the maximum and minimum of wind power in a period of time, and ignores other values during this period.

(3) Whether the slop of wind power values exceeds the given threshold value in a period of time.

(3){\displaystyle \frac{\left|p_{(t+\Delta t)}-p_{\left(t\right)}\right|}{\Delta t}} > p_{\left(\mathrm{tsd}\right)}.

Equation (3) can reflect the wind power fluctuation rate. This paper uses Eq. (3) to judge ramp events.

It is demonstrated that wind power and ramp events may cause a power imbalance in power systems and even threaten the security of power systems [⁸–¹¹]. Reference [¹²] employs wind curtailment as a tool to control the ramp. However, wind curtailment would lead to a waste of wind power resources. Ramp events have had a different effect on power systems all over the world [¹³–¹⁵]. Ramp-up events will cause a frequency drop, seriously, that can make a lot of loads to be cut.

To lighten the damages caused by ramp events, the energy storage systems can be applied to smooth the power fluctuation. Reference [¹⁶] proposes a wind power ramp control method with an energy storage system (ESS) based on wind power ramp event forecast. An optimization model is established to optimize the output power of ESS and the wind curtailment to meet a hybrid wind power ramp limit. Reference [¹⁷] presents field results and analyses quantifying the ability and the value of sodium sulfur (NAS) battery energy storage toward shifting wind generation from off-peak to on-peak, limiting the ramp rate of wind farm output, and a strategy to integrate the aforementioned goals. The integration of wind energy with energy storage devices to support the short-term shortcomings of wind energy is discussed in Ref. [¹⁸], and a turbine level hybrid configuration of an energy storage system is used to limit the power ramp rates and apply power smoothing. The analysis shows that significant improvements can be made to shape the output power of the wind farm using energy storage systems. Reference [¹⁹] proposes a smoothing controller to suppress the power fluctuation from doubly-fed induction generator (DFIG)-based wind farms. A numerical case study demonstrates the effectiveness and economy of this smoothing controller for the isolated system studied. In Ref. [²⁰], a novel state of charging (SOC) feedback strategy in wind power smoothing based on short-term forecast and scenario analysis is proposed. The effectiveness of the new SOC feedback strategy is validated in a case study by using historical wind power data of a wind farm in Gansu province of China. Reference [²¹] presents a wind power filtering approach to mitigate short- and long-term fluctuations using a hybrid energy storage system (HESS), and a novel wavelet-based capacity configuration algorithm to properly size the HESS.

The main contribution of this paper is to optimize wind power, pumped storage power station, and gas power station by using dynamic programming for alleviating ramp events under multi-stage condition. First, wind power, pumped storage power station, and gas power station models are established by using inequality constraints 2. Then, the stage, state, transition equation, and solving flowchart of dynamic programming are proposed for the optimal model under the energy internet. Finally, the performance of the algorithm is verified and future work is proposed.

Problem formulation

An energy internet system model is established which includes wind power, a pumped storage power station, and a gas power station, as is shown in Fig. 1.

In Fig. 1, p_w represents wind power, p_g represents the output of the gas power station, and p_h and p_p represent the pumped storage power station generation and pumping power respectively.

Wind power

Wind power integrated into power systems must meet the given threshold p_(tsd). Because wind power is unstable, and, therefore, when larger wind power is integrated into power systems, the power flow of power systems will be unstable. Thus, in order to keep power systems safe, the ramp threshold p_(tsd) is set in this paper when wind power is integrated into power systems. The power systems are more stable when p_(tsd) in [p_min(tsd), p_max(tsd)],

(4)

where p_min(tsd) and p_max(tsd) are the minimum and maximum ramp threshold values when wind power is integrated into power systems.

Pumped storage power station

Pumped storage power station plays an important role in energy internet. When wind power is surplus, the surplus energy is consumed to pump water from the lower reservoir to the upper reservoir, which is stored as potential energy of the reservoir. When the wind power is insufficient, the water of the upper reservoir is released, and the stored potential energy is used to generate power.

Pumped storage power station must meet the following inequality constraints.

(5)

where p_p(t) is the pumping power in the given time period t, p_pmin and p_pmax represent the minimum and maximum pumping power. In this paper, the value of p_pmin is 0 MW and p_pmax is 2 MW.

(6)p_{\mathrm{h}\min }\le p_{\mathrm{h}\left(t\right)}\le p_{\mathrm{h}\max },

where p_h(t) is the hydro turbine power in the given time period t. p_hmin and p_hmax represent the minimum and maximum of the hydro turbine power, p_hmin is set to 0 MW, and p_hmax is set to 2 MW in this paper.

(7)E_{\mathrm{c}\min }\le E_{\mathrm{c}\left(t\right)}\le E_{\mathrm{c}\max },

where E_c(t) is the storage capacity of pumped storage power station in the given time period t, and E_cmin and E_cmax represent the minimum and maximum of the storage capacity. In this paper, E_cmin is set to 0 MWh and E_cmax is set to 20 MWh.

The storage capacity constraints of pumped storage power station are

E_{\mathrm{c}(t+\mathrm{1})}=E_{\mathrm{c}\left(t\right)}+{\epsilon }_{\left(t\right)}\eta p_{\mathrm{p}\left(t\right)}\Delta t-{\gamma }_{\left(t\right)}{\displaystyle \frac{p_{\mathrm{h}\left(t\right)}}{\xi }}\Delta t,

(8){\epsilon }_{\left(t\right)},{\gamma }_{\left(t\right)}=\{\begin{array}{l}\mathrm{0}\hfill \\ \mathrm{1}\hfill \end{array},

{\epsilon }_{\left(t\right)}·{\gamma }_{\left(t\right)}=\mathrm{0},

where ε_(t) and γ_(t) are 0 or 1, and η and ξ represent pumping power efficiency and generating power efficiency. In this paper, E_c(0) = 3 MWh, η = 0.93, and ξ= 0.95.

Gas power station

Gas power station can reduce pollution when it is integrated into power systems, which has a high ramp rate of up to 20% per minute. In other words, a gas power station can generate a maximum power of 20% installed capacity within one minute.

The gas power station must satisfy the following constraints.

(9)

where p_gmin and p_gmax represent the minimum and maximum gas power station generating power. In this paper, the installed capacity of the gas power station is 5 × 1.2 MW, the value of p_gmin is 0 MW, p_gmax is 6 MW, and p_g(t) represents the gas power station generating power in the t time period.

During smoothing wind power fluctuation, due to Eqs. (5) and (6), the storage capacity of pumped storage power station may be insufficient, which will result in curtailing wind power. The gas power station belongs to renewable energy with a high ramp rate, which can compensate for the storage capacity of the pumped storage power station. Due to the high cost of the gas power station, this paper uses a gas power station to generate power as low as possible.

Dynamic programming

Dynamic programming has no strict requirements on objective function and constraints and can find the global optimal solution for nonlinear, convexity, and even discontinuous programming problem. The above inequality constraints (1)–(9) are dynamic, which can be used to solve a dynamic programming problem.

The proposed model of dynamic programming in this paper is introduced as follows.

Stage

The process of searching optimal solution is divided into many stages by time or solving-step. In this paper, stages are divided by time. Each hour is counted as a stage and there are 24 stages per day. The process starts from the first stage to the next stage, until the end of the last stage. If a decision is made at a given stage, the subsequent decision will be affected by the previous stage. In this paper, the dynamic programming algorithm starts from the first time period of the system, finds the optimal p_p(t), p_h(t), p_g(t), and E_c(t) of this period, and then considers the first two time periods in a comprehensive way to find the optimal solution, until the global optimal solution is found. The solution process diagram is depicted in Fig. 2.

The numbers in Fig. 2 represent the time periods, and black arrows represent the development direction of the time periods. This paper divides the system into many stages by time periods to simplify a complex problem. The optimal solution of the system is found from the first stage to the last stage.

State

The state represents the specific situation of each stage of the system. A stage can contain several states. Because the storage capacity pumped storage power station represents the energy, which can be dispatched during the process of smoothing wind power in energy internet, this paper selects storage capacity of pumped storage power station to represent the state. State variables are used to describe states. In this paper, state variables are E_c(t)∈ [E_cmin, E_cmax], E_cmin = 0, and E_cmax = 20 MWh. According to Eq. (8), p_p(t) and p_h(t) affect E_c(t).

State variables of the stage i are expressed as S_i, and the k state of stage i is expressed as s_ik. The set of state variables in the stage i can be expressed as

(10)S_i=\{s_{i1},s_{i2},\mathrm{...},s_{ik},\mathrm{...},s_{im}\},

where S_i = E_c(t)∈ [E_cmin, E_cmax]. By determining the state variables in each stage of the energy internet, the optimal solution of energy internet can be found more quickly.

Decision

In a given stage of the system, the decision maker can choose different results by different state variables. Decision results at the previous stage affect the states of the next stage. In energy internet, because the output of pumped storage power station and gas power station in the previous stage affect the storage capacity of pumped storage power station in the next stage, this paper selects pumped storage power station and gas power station output as decisions. Decision variables are used to describe decisions. Decisions depend on the states, therefore, d_i(s_k) is used to represent the decision variable when the system is in stage i and the state is S_k. In this paper, based on wind power forecast, p_p(t) or p_h(t) can be given when E_c(t)∈ [E_cmin, E_cmax] is given in Eq. (8). D_i(s_k) represents the decision set. The relationship between d_i(s_k) and D_i(s_k) is expressed as

(11)d_i\left(s_k\right)\in D_i\left(s_k\right)\mathrm{.}

In this paper, the decisions of pumped storage power station in stage i are shown as

(12)d_{1i}\left(s_k\right)\in D_{1i}\left(s_k\right),

where D_1i(s_k) = [–2 MW, 2 MW]. D_1i(s_k) = [–2 MW,0] represents p_h(t), which stands for pumped storage power station generates power. D_1i(s_k) = [0, 2 MW] represents p_p(t), which stands for pumped storage power station pumping.

The decisions of gas power station in stage i are shown as

(13)d_{2i}\left(s_k\right)\in D_{2i}\left(s_k\right),

where D_2i(s_k) = [0, 6 MW], which means p_g(t). p_g(t) is calculated when the pumped storage power station generation is insufficient.

In this paper, dynamic programming chooses d_1i(s_k) first, then chooses d_2i(s_k) in each stage. Because the cost of pumped storage power station is cheaper than gas power station in smoothing wind power fluctuation, the output power of that is chosen as decision variable d_2i(s_k) in ith stage.

From the initial stage to the final stage, each stage has many decisions which constitute decision making sequences, which are called policies. The policy achieves optimal effect which can be called optimal policy. The policy from the state s_i of the i stage can be expressed as

(14)P_{i,n}\left(s_i\right)=\{d_i,d_{i+1},\mathrm{...},d_n\}\mathrm{.}

In this paper, the process of dynamic programming generates a policy from the first stage is illustrated in Fig. 3.

State transition equation

Given the state of stage i, the state of next stage (i + 1 stage) can be determined when the decision of stage i is determined. The state transition equation can be understood as the relationship between the decision at the previous stage and the state of the next stage. The function is given as

(15)s_{i+1}=T_i(s_i,d_i),

s_i represents the state of stage i, s_{i+ 1} represents the state of stage i+ 1, and d_i represents the decision of stage i.

This paper chooses Eq. (8) as a state transfer equation to describe the relationship between the output and storage capacity of pumped storage power station in energy internet. In Eq. (8), d_i is p_p(t) or p_h(t), and s_i is E_c(t). ε_(t) = 1 and γ_(t) = 0, when wind power is surplus, p_p(t) is given and constraints (4)–(7) are considered; ε_(t) = 0 and γ_(t) = 1, when wind power is insufficient, p_h(t) is given and constraints (4)–(7) are considered. In this paper, based on wind power forecast, a day is divided into 24 stages. According to Fig. 2 and Eq. (8), E_c(t) can be given in every stage. Since the gas power station has no capacity balance equation, there is no state transfer equation for the gas power station.

Objective function and algorithm flowchart

This paper establishes the following objective function to represent the fluctuation of wind power.

(16)\begin{array}{l}\min z={\displaystyle \sum _{t=\mathrm{1}}^N{(p_{\mathrm{w}\left(t\right)}-u_{\left(t\right)}{p}_{\mathrm{s}\left(t\right)}+v_{\left(t\right)}{p}_{\mathrm{g}\left(t\right)}-\overline{p})}^{\mathrm{2}}},\hfill \\ u_{\left(t\right)}, v_{\left(t\right)}=\{\begin{array}{l}\mathrm{0} \hfill \\ \mathrm{1} \hfill \end{array}, \hfill \end{array}

where N is the total number of stages. In this paper, N = 24, u(t) and v(t) are 0 or 1, 0 representing not to take part in smoothing wind power, 1 representing to take part in smoothing wind power. p_s(t) represents the conversion power between wind power and pumped storage power station. p_s(t) = p_p(t) >0 means that the pumped storage power station utilizes wind power pumping; p_s(t) = p_h(t) <0 means that the pumped storage power station generates power to compensate for wind power vacancy. \overline{p} represents the average power value of all stages in energy internet. Equation (16) must meet constraints (4)–(9).

Figure 4 is the flowchart of the algorithm for dynamic programming in smoothing wind power fluctuation.

Case study

This paper uses a dynamic programming algorithm shown in Fig. 4 to smooth wind power fluctuation in an 8 MW wind farm. The MATALAB software is applied to solve the dynamic programming problem. The diagram of the case study in energy internet is demonstrated in Fig. 5 based on Ref. [²²].

Discrimination of ramp events

The day-ahead wind power forecast is shown in Fig. 6(a). This paper uses Eqs. (17) and (18) to distinguish ramp events.

(17){\displaystyle \frac{p_{(t+\Delta t)}-p_{\left(t\right)}}{\Delta t}}>p_{\max \left(\mathrm{tsd}\right)},

(18)

where p_min(tsd) is –2.7 MW/h, p_max(tsd) is 2.7 MW/h, Δt is one hour, and p_w(t) is wind power in time t.

The black arrows in Fig. 6(a) represent ramp-up events, while the red ones represent ramp-down events.

As is shown in Fig. 6(a), ramp-down events occur in 3–4 h, 6–7 h, and 18–19 h while ramp-up events occur in 5–6 h, 8–9 h, and 13–14 h. The ramp rate for each period is exhibited in Fig. 6(b).

The ramp rates of the ramp-down events are –3.15 MW/h, –4.2 MW/h, and –3.5 MW/h respectively, whereas the ramp rates of the ramp-up events are 2.8 MW/h, 5.25 MW/h, and 3.36 MW/h respectively. Figures 6(a) and 6(b) are compared with p_(tsd) respectively. It is clearly observed that the fluctuation of the wind power is very large.

Smoothing method

The steps of using the dynamic programming algorithm to smooth wind power in energy internet are as follows.

(1) The wind power values of the 24 stages are inputted.

(2) The decision values of pumped storage power station and gas power station are [–2 MW, 2 MW] and [0, 6 MW].

(3) The corresponding decision values in Step (2) are allocated for each stage of the wind power value in Step (1), p_a(t) = p_w(t) – p_s(t) is calculated, and p_s(t) is p_p(t) or p_h(t).

(4) p_a(t) is reserved, Eq. (4) is considered, and p_s(t) is reserved.

(5) E_c(t) is calculated from Eq. (8).

(6) p_g(t) is given in [0, 6 MW] when constraints (5), (6), and (7) are unsatisfied.

(7) Dynamic programming is used to calculate Eq. (16) from the first stage to the last one, and the optimal results of every stage are obtained.

The results of using dynamic programming are displayed in Fig. 7.

As is shown in Fig. 7(a), the waveform using the dynamic programming algorithm is smoother. Due to the pumped storage power station constraints (Eqs. (5)–(7)), local areas of the waveform cannot be completely smooth. In Fig. 7(b), according to Eqs. (17) and (18), it is known that ramp events have been eliminated and ramp rates have been decreased by using the dynamic programming algorithm to smooth the wind power.

According to Eqs. (7) and (8), the storage capacity of pumped storage power station is shown in Fig. 8(a). The storage capacity gradually increased in time periods 1–4 h and 10–19 h, but decreased in time periods 5–9 h and 20–24 h. The output of the pumped storage power station is plotted in Fig. 8(b). A positive value represents that the pumped storage power station uses wind power for energy storage, while a negative value represents that the pumped storage power station generates power.

As is shown in Fig. 8(b), due to the pumped storage power station constraints (5) and (6), the pumping and generate power values are below 2 MW.

Because of Eq. (6), the pumped storage power station generates a maximum value of 2 MW which may not be sufficient enough to smooth the ramp-down events. Therefore, the gas power station is applied to generate electricity. In this paper, the natural gas consumption is presented in Fig. 9.

Gas power station generates electricity from the combustion of natural gas. In this paper, the efficiency of the gas power station is 24%, the calorific value of natural gas is 35 MJ/m³, and the heat consumption rate is 14.795 MJ/kWh. Equation (19) represents the relationship between the output of gas power station and combustion of natural gas.

(19)p_{\mathrm{g}\left(t\right)}=V\left({\displaystyle \frac{\beta \alpha }{\theta }}\right),

where p_g(t) is the output of gas power station, V is combustion of natural gas, β is the calorific value of natural gas, α is the efficiency of the gas power station, and θ is heat consumption rate. The output of gas power station during smoothing wind power is shown in Fig. 10.

From Figs. 7(a) and 9, according to the results of dynamic programming algorithm, it is known that wind power is lower and the storage capacity of pumped storage power station is insufficient in 4–9 h. During that time, the gas power station generates power to smooth wind power fluctuation in the energy internet.

In Eq. (16), wind power is not smoothed, p_g(t) and p_s(t) are 0, z={\displaystyle \sum _{t=\mathrm{1}}^N{(p_{\mathrm{w}\left(t\right)}-\overline{p})}^2}, N = 24. When wind power is smoothed, z is obtained by dynamic programming. The objective function values of no smoothing and smoothing by using the dynamic programming algorithm are 69.547 and 0.645.

From Fig. 7(b) and the objective function values, it is known that ramp events have been eliminated and ramp rates have been decreased by using the dynamic programming algorithm to smooth wind power.

Evaluation functions

To evaluate the effect of the dynamic programming algorithm in the energy internet, evaluation functions are introduced in this paper.

Environmental protection

In this paper, renewable energy is applied to generate power, which can effectively reduce CO₂ and SO₂ emissions. According to general rules for calculation of comprehensive energy consumption in China, 1 kWh will consume 300 g standard coal, and 1 tons of standard coal generate 2.5 tons of CO₂, resulting in 7.4 kg of SO₂. 1 m³ natural gas will generate 1.8 kg CO₂. Wind power and pumped storage power station cannot generate CO₂ and SO₂. In this paper, Eqs. (20) and (21) are used to represent CO₂ and SO₂ emissions.

(20)T_{{\mathrm{co}}_{\mathrm{2}}}={\displaystyle \sum _{t=\mathrm{1}}^N{p}_{\mathrm{a}\left(t\right)}}\times t\times \mathrm{300}\times \mathrm{2.5}-{\displaystyle \sum _{t=\mathrm{1}}^{N}V\times t}\times \mathrm{1.8},

(21)T_{{\mathrm{so}}_{\mathrm{2}}}={\displaystyle \sum _{t=\mathrm{1}}^N{p}_{\mathrm{a}\left(t\right)}}\times t\times \mathrm{7.4}\times \mathrm{300}/\mathrm{1000},

where p_a(t) represents smoothing wind power and V is combustion of natural gas. By calculating, T_{{\mathrm{co}}_{\mathrm{2}}} = 62948 kg and T_{{\mathrm{so}}_{\mathrm{2}}} = 220.6991 kg. In this paper, by using wind power and the pumped storage power station and gas power station, CO₂ and SO₂ which are produced by coal-consumption units are reduced.

Deviation rate of wind power

The deviation rate of wind power is calculated by using

(22){\beta }_{\mathrm{w}}=\max \left({\displaystyle \frac{\left|p_{\mathrm{w}\left(t\right)}-\overline{p}\right|}{\overline{p}}}\right),

(23)\overline{p}={\displaystyle \frac{{\displaystyle \sum _{t=\mathrm{1}}^N{p}_{\mathrm{w}\left(t\right)}}}{N}},

where β_w is the deviation rate of wind power variation, p_w(t) is wind power in time period t, \overline{p} is the average value of wind power, and N = 24.

β_w reflects the local smoothing effect of the dynamic programming algorithm in the energy internet. If β_w is smaller, the local smoothing effect is better and local wind power is more stable; if β_w is larger, the local smoothing effect is worse and local wind power is more unstable.

Standard deviation of wind power

The standard deviation of wind power is calculated by using

(24){\sigma }_{\mathrm{w}}=\sqrt{{\displaystyle \frac{{\displaystyle \sum _{t=\mathrm{1}}^N{(p_{\mathrm{w}\left(t\right)}-\overline{p})}^{\mathrm{2}}}}{N}}},

(25)\overline{p}={\displaystyle \frac{{\displaystyle \sum _{t=\mathrm{1}}^N{p}_{\mathrm{w}\left(t\right)}}}{N}},

where σ_w is the standard deviation of wind power variation, p_w(t) is wind power in time period t, \overline{p} is the average value of wind power, and N = 24.

In this paper, σ_w is used to reflect the global smoothing effect of dynamic programming algorithm in the energy internet. Smoothing wind power is more stable when σ_w is smaller, which means that the global smoothing effect is better and vice versa.

The evaluation function values of no smoothing and smoothing by using the dynamic programming algorithm in the energy internet are compared in Table 1.

As is shown in Table 1, β_w and σ_w are smaller than the original wind power, which means that the dynamic programming algorithm is effective in smoothing wind power fluctuation from local and global effect in energy internet.

Conclusions

In this paper, a dynamic programming algorithm proposed is applied to smooth wind power fluctuation in energy internet which is composed of wind power, a pumped storage power station, and a gas power station. By using dynamic programming method, ramp events, stages, states and decisions are defined. The optimal results are obtained when the constraints in ramp events and the main function based on minimal power fluctuation are considered. The simulation results and evaluation functions demonstrate that the dynamic programming algorithm can effectively smooth the ramp events and can obtain the energy dispatch of pumped storage power station and gas power station. The gas power station is applied to smooth ramp-down events when the pumped storage power station generation is insufficient. Wind power in common point is more stable. Today, energy internet is a timely topic worldwide. In the future, the adaptive problem of energy internet will be studied.

References

[1] Sun Q, Han R, Zhang H, Zhou J, Guerrero J M. A multiagent-based consensus algorithm for distributed coordinated control of distributed generators in the energy internet. IEEE Transactions on Smart Grid, 2015, 6(6): 3006–3019

[2] Sevlian R, Rajagopal R. Detection and statistics of wind power ramps. IEEE Transactions on Power Systems, 2013, 28(4): 3610–3620

[3] Xiong Y, Zha X, Qin L, Ouyang T, Xia T. Research on wind power ramp events prediction based on strongly convective weather classification. IET Renewable Power Generation, 2017, 11(8): 1278–1285

[4] Gallego C, Costa A, Cuerva A. Improving short-term forecasting during ramp events by means of regime-switching artificial neural networks. Advances in Science & Research, 2011, 6(1): 55–58

[5] Ouyang T, Zha X, Qin L, Kusiak A. Optimization of time window size for wind power ramps prediction. IET Renewable Power Generation, 2017, 11(8): 1270–1277

[6] Ouyang T, Zha X, Qin L. A survey of wind power ramp forecasting. Energy & Power Engineering, 2013, 5(4): 368–372

[7] Yang Q, Berg L K, Pekour M D, Fast J K, Newsom R, Stoelinga M, Finley C. Evaluation of WRF-predicted near-hub-height winds and ramp events over a pacific northwest site with complex terrain. Journal of Applied Meteorology and Climatology, 2013, 52(8): 1753–1763

[8] Bhateshvar Y, Mathur H, Siguerdidjane H. Impact of wind power generating system integration on frequency stabilization in multi-area power system with fuzzy logic controller in deregulated environment. Frontiers in Energy, 2015, 9(1): 7–21

[9] Verma Y P, Kumar A. Dynamic contribution of variable-speed wind energy conversion system in system frequency regulation. Frontiers in Energy, 2012, 6(2): 184–192

[10] Cui M, Ke D, Sun Y, Gan D, Zhang J, Hodge B. Wind power ramp event forecasting using a stochastic scenario generation method. IEEE Transactions on Sustainable Energy, 2015, 6(2): 422–433

[11] Liu Y, Sun Y, Infield D, Zhao Y, Han S, Yan J. A hybrid forecasting method for wind power ramp based on orthogonal test and support vector machine (OT-SVM). IEEE Transactions on Sustainable Energy, 2017, 8(2): 451–457

[12] Kalantari A, Galiana F. The impact of wind power variability and curtailment on ramping requirements. In: Transmission and Distribution Conference and Exposition, Sao Paulo, Brazil, 2010: 133–138

[13] Zhao J, Abedi S, He M, Du P, Sharma S, Blevins B. Quantifying risk of wind power ramps in ERCOT. IEEE Transactions on Power Systems, 2017, 32(6): 4970–4971

[14] Couto A, Costa P, Rodrigues L V, Lopes V, Estanqueiro A. Impact of weather regimes on the wind power ramp forecast in Portugal. IEEE Transactions on Sustainable Energy, 2015, 6(3): 934–942

[15] Ganger D, Zhang J, Vittal V. Statistical characterization of wind power ramps via extreme value analysis. IEEE Transactions on Power Systems, 2014, 29(6): 3118–3119

[16] Gong Y, Jiang Q, Baldick R. Ramp event forecast based wind power ramp control with energy storage system. IEEE Transactions on Power Systems, 2016, 31(3): 1831–1844

[17] Tewari S, Mohan N. Value of NAS energy storage toward integrating wind: results from the wind to battery project. IEEE Transactions on Power Systems, 2013, 28(1): 532–541

[18] Esmaili A, Novakovic B, Nasiri A, Abdel-Baqi O. A hybrid system of li-ion capacitors and flow battery for dynamic wind energy support. IEEE Transactions on Industry Applications, 2013, 49(4): 1649–1657

[19] Lin J, Sun Y, Song Y, Gao W, Sørensen P. Wind power fluctuation smoothing controller based on risk assessment of grid frequency deviation in an isolated system. IEEE Transactions on Sustainable Energy, 2013, 4(2): 379–392

[20] Zhou Y, Yan Z, Li N. A novel state of charge feedback strategy in wind power smoothing based on short-term forecast and scenario analysis. IEEE Transactions on Sustainable Energy, 2017, 8(2): 870–879

[21] Jiang Q, Hong H. Wavelet-based capacity configuration and coordinated control of hybrid energy storage system for smoothing out wind power fluctuations. IEEE Transactions on Power Systems, 2013, 28(2): 1363–1372

[22] You L, Liu D, Zhong Q, Yu N. Research on optimal schedule strategy for active distribution network. Automation of Electric Power Systems, 2014, 38(9): 177–183 (in Chinese)

RIGHTS & PERMISSIONS

Higher Education Press and Springer-Verlag GmbH Germany, part of Springer Nature