New energy power generation, represented by wind and solar power, exhibits significant power output fluctuations and uncertainties. Both wind and solar power outputs are directly affected by local weather conditions, prone to power output spikes or drops, posing challenges to the grid connection frequency of the power system.
Due to power fluctuations and relatively complex grid impedance characteristics, under normal conditions of large-scale centralized grid connection or random power output, power oscillations are likely to occur, leading to power system stability issues. This affects load and the performance of planned new energy production systems across a wide area, necessitating sufficient reserve capacity in the system to avoid impacting the ability to integrate new energy sources, which is crucial for achieving both planning and economic efficiency.
The integration of energy storage and new energy sources mainly focuses on three aspects: First, by releasing grid-level loads over short time periods, it enables 10-minute-level power regulation of the power grid, alleviating short-term fluctuations and fully leveraging the existing grid's capacity to connect to new energy sources. Second, by developing minute-level plans that include new energy output forecasts, and based on short-term day-ahead power generation forecasts, it effectively incorporates new energy sources into ultra-short-term power forecasts. This improves the rational operation and scheduling of various generating units within the grid, reduces the demand for rapid frequency regulation resources, enhances the accuracy and stability of grid forecasts, and smooths out minute-level real-time fluctuations in new energy sources, minimizing the impact on the normal operation of conventional generating units.
Peak Shaving and Valley Filling
Compared to conventional power generation, renewable energy generation has a relatively low utilization rate of its equipment or units. Taking the "Three Norths" region of my country as an example, according to wind resource statistics, the probability of a wind farm's total output exceeding 60% of its total installed capacity is generally less than 5%. To improve line utilization, line capacity planning typically aims to meet 95% of wind power transmission needs or 60% of the total installed capacity of wind farms. The situation is even more severe for photovoltaics. Therefore, a certain percentage of wind power will be curtailed due to insufficient transmission capacity, and solar power will be curtailed due to load mismatch (anti-peak-shaving characteristics).
Renewable energy generation, with its relatively long-term fluctuations on an hourly basis throughout the day, and the arrival of peak electricity demand in the evening (generally 7-10 PM), will increase the system's upward and downward capacity requirements. Wind power, on the other hand, often reaches full output around midnight, when the load is at its lowest point of the day. Therefore, to eliminate the uncertainty in forecasting renewable energy generation, both the power grid and conventional generating units must bear significant risks associated with deep peak shaving.
Peak shaving and valley filling utilize the time-shifting characteristics of energy storage to maximize line transmission capacity, reduce the need to match load trends, and decrease the demand for increased and decreased capacity from conventional generating units.
By summing the given daily load curve P_l with the renewable energy generation output curve P_{NE}, we can obtain the final system equivalent load curve ∑P_i, i.e., ∑P_i = P_l - P_{NE}. However, considering the output regulation range of conventional power plants and peak-shaving power plants, and the maximum power P_L that the regional interconnection line can transmit or obtain to the external grid, the maximum effective power P_{max} of the grid-connected units is:
P_{max} = μ(P_f + P_b + P_L) (3-3)
Where:
the only place you'll find outside the home
- P_f-Maximum output power of peak-shaving units;
- P_b-Minimum output of units that cannot participate in peak shaving;
- μ-Grid transmission and operating efficiency.
In the formula, C represents the output power regulation coefficient of the peak-shaving unit. The power relationships are shown in the figure.
The minimum effective power P_{min} of the grid-connected units is:
During the lowest load period t₁–t₂, the downward regulation capacity reserved by conventional peak-shaving units is the maximum renewable energy power P'_{NE} that the grid can accept during this period, i.e., P'{NE} = P{max} - P_{min} (3-5) where P_{min} is the minimum daily output (without energy storage, renewable energy generation during t₁–t₂ can only be achieved through wind/solar curtailment).
It can be seen that without energy storage, renewable energy output during t₁–t₂ can only be limited; however, with energy storage, charging during t₁–t₂ and discharging during t₃–t₄ shifts the effective equivalent load curve ∑P_i within the range of P_{min} and P_{max}, avoiding renewable energy output limitations and wind/solar curtailment, improving renewable energy absorption capacity, reducing the grid's demand for reserve capacity, and improving overall system efficiency. The power P_{BESS} of a BESS (Battery Energy Storage System) is:
P_{BESS} = max( P_{min} - ∑P_{min}, ∑P_{max} - ∑P_{max} ) (3-6)
The energy E_{BESS} of a BESS is:
E_{BESS} = max{ μ_c ∫{t₁}^{t₂} (P{min} - ∑P_i) dt , 1/μ_d ∫{t₃}^{t₄} (∑P_i - P{max}) dt } (3-7)
Where:
- μ_c -- Charging efficiency of the energy storage system;
- μ_d -- Discharging efficiency of the energy storage system.
Further research in a broader sense shows that for load peaks and valleys that are often prolonged, configuring an energy storage system of a certain capacity can effectively reduce the peak-valley difference, as shown in the figure.
The improvement level of load peak-valley difference is:
- Where Pimax is the maximum expected load;
- Pimax is the minimum expected load.
The energy storage system configuration method is similar to the previous one and will not be repeated.
Improve prediction accuracy
According to NBT32011-2013 "Technical Requirements for Power Prediction System of Photovoltaic Power Station", the root mean square error of short-term prediction during the power generation period of a photovoltaic power station (excluding periods with limited output) should be less than 0.15, and the monthly pass rate should be greater than 80%; the root mean square error of the fourth hour of ultra-short-term prediction should be less than 0.1, and the monthly pass rate should be greater than 85%.
According to the "Interim Measures for the Administration of Wind Farm Power Forecasting and Early Warning", the maximum error of the daily forecast curve of a wind farm shall not exceed 25%, the real-time forecast error shall not exceed 15%, and the root mean square error of the forecast for the whole day shall be less than 20%.
Both short-term and ultra-short-term forecasts provide forecast data at 15-minute intervals. Therefore, the output of new energy sources can be segmented and controlled at 15-minute intervals, with 96 control segments throughout the day. The allowable control error bandwidth ΔP is established based on the maximum allowable error in the relevant forecasting technical specifications. As shown in Figure 3-8, P(1) and Pe(2) are the predicted power values for the first and second 15-minute intervals, respectively, while AP is the allowable error bandwidth, set to 15% of the installed capacity of new energy power generation.
Short-term variation smoothing of new energy power generation
The short-time rate of change of new energy power generation should also meet the requirements of power system stability. The current power grid limits for the active power variation of new energy grid-connected power generation are shown in the table below.
Table 3-2: Limits on Active Power Change for Grid-Connected New Energy Power Generation
| Installed Capacity of New Energy Power Station (MW) | Maximum Change in Active Power over 10 Minutes (MW) | Maximum Change in Active Power over 1 Minute (MW) |
|---|---|---|
| < 30 | 10 | 3 |
| 30 ~ 150 | Installed Capacity / 3 | Installed Capacity / 10 |
| > 150 | 50 | 15 |
In renewable energy smoothing applications, the BESS (Power Equipped Element System) is used to store and release renewable energy power generation, suppressing minute-level power fluctuations in the renewable energy grid-connected system. This ensures that the combined output P fluctuation of the energy storage PBEss (Power Element System) and the renewable energy Pv (Power V) meets the aforementioned technical requirements, with the control time interval mostly set to 1 minute. However, unlike algorithms that improve prediction accuracy, this approach primarily focuses on the power fluctuations of renewable energy output. Therefore, in selecting the specific rated power of the BESS, the data sample source for statistical analysis and probability analysis will be the minute-level and 10-minute-level active power changes of renewable energy output.
The power and capacity design of BESS can still be based on the probability statistics of past power changes and the cumulative changes in power consumption, aiming to meet the smoothing requirements in 80% to 90% of cases. This will not be repeated here. To ensure that the power fluctuation range meets the above requirements, two main BESS power control algorithms are used:
- One is the point-by-point limiting method;
- The other is the low-pass filtering method.
Point-by-point limit method
Taking the figure as an example, the figure shows a large comparison between the new energy output Pne(j) at time j and the combined output P(J-n) over the past 10 minutes. It can be seen that at time (j-3), that is, the change between P(j-3) and Pne(j) is the largest, and it exceeds the maximum of 10 minutes. The comparison shows that △P10.
Therefore, in order to meet the 10-minute power fluctuation limit, the output range of BESS (positive for charging, negative for discharging) is:
Low-pass filtering method
Based on the filtering principle in signal processing, as shown in the figure, the low-pass filter makes the output signal smoother by adding or subtracting the amplitude of the input signal. Similarly, the access of BESS will also achieve smoothing of the fluctuation of the output power of the new energy power station through its charging and discharging control, so as to meet the relevant technical requirements.
The expected value of the total grid-connected power ∑P\sum P∑P is given by:
Discretize the data, where t is the control period, and we take 1 minute：
∑P(j) = (τ / (τ + t)) * ∑P(j-1) + (t / (τ + t)) * P_ne(j)
Given ∑P(j) = P_ne(j) - P_bess(j)
P_bess(j) = (τ / (τ + t)) * (P_ne(j) - ∑P(j-1))
P_bess(j) = (τ / (τ + t)) * (∑P(j) - ∑P(j-1))
According to grid-connected power fluctuation technical requirements, the minute-level fluctuation range of ∑P(j)must satisfy:
|∑P(j) - ∑P(j-1)| ≤ min(ΔP_i, 0.1 P_0)
Substituting the calculation formula for Pbess(j) we obtain:
