制定一篇金融專業的留學生dissertation-The Depth Analysis by New Model

 

Considering the wind

As a commodity, the electricity price is affected by the flow and distribution of all kinds of resources in the entire market. With the development of renewable energy generation technology and the increase of its share, such as the wind power, electricity price curve is more complex than before. The tariff fluctuations are very sensitive to any changes in grid-connected wind power, which will lead to rose or fell of the electricity price. Due to the limitation of GBM estimates, the neural network model can effectively deal with the multi-variable and nonlinear problems, which is qualified for the analysis of grid-connected wind power system.作為一種商品，電力價格是由各種資源在整個市場的流動和分布的影響。隨著可再生能源發電技術的發展和增加其份額，如風力發電，電價曲線比以前更加復雜。電價波動對并網風電的任何變化，這將導致大漲或大跌的電價非常敏感。由于GBM估計的局限性，神經網絡模型可以有效地處理多變量，非線性問題，這是合格的并網風電系統的分析。

 

The analysis main consider the factors like grid-connected wind power, load and historical clearing price. Taking into account a different variation of the tariff at different time period, it is divided into 48 sub-time tariff sequences. 

Prior to electricity price forecasting, we need to make clear of the relevance between the factor and tariff. The correlation coefficient can be used to measure the degree of association, which is defined as follows:  .Where, ρ is the correlation coefficient, f for a certain influence factor, σf for factors of standard deviation, p for electricity price, σp is the standard deviation of the tariff, Cov (f, p) is the covariance between factor f and tariff p. 該分析主要考慮像電網風電，負載和歷史清算價格的因素。考慮到資費在不同的時間段的一個不同的實施方案中，它被劃分成48個子時間資費序列。

此前電價預測，我們需要明確的因素和關稅之間的相關性。相關系數可以用來測量關聯程度，其定義如下：式中， ρ為相關系數，F為一定的影響因子， ΣF為標準偏差，對電力價格的因素， ΣP是資費標準差，冠狀病毒（ F，P ）為系數f和關稅P間的協方差。

 

According to the British electricity market data, we can calculate the correlation coefficient between the wind power and related load ratio (ρ1) and the correlation with historical clearing price (ρ2). Based on the strong correlation with tariff, we consider the equivalent load history clearing price, load, wind power and the load ratios, historical clearing price as the neural network input factor to forecast market sub-period clearing price.

 

Neural network is a parallel, distributed information processing structure consisting by the processing unit. Artificial neuron is to simulate the basic characteristics of the neurons in the brain, which has multi-input / single output nonlinear unit with a certain internal state and the threshold.

 

Radial basis function network (RBF) has only one hidden layer, the output is a weighted sum of the hidden layer [12-13]. The most commonly used radial basis function is a Gaussian function. Generalized regression neural network (GRNN) is a variation of the RBF network [14-15]. The theoretical basis is non-linear regression analysis. We can obtain the non-independent variable y with respect to the regression analysis of independent variable x. As a forward-feed neural network, GRNN has input layer, a hidden layer and output layer. Hidden layer used Gaussian transformation to control the output, thereby inhibiting the activation of the output unit. Gaussian function belongs to accepted domain in the input space. The influence input neuron was attenuation because of the distance between the input vector and the network output.

 

To calculate the evaluation error, the traditional method of mean absolute percentage error (MAPE) is not enough. We substitute the predictive value with the mean price to decrease the error caused by the fraction. 

 .

However, we need the probability index to evaluate the credibility of the forecast results [13]. And the error distribution function F（ε＜ε(n)）can be fitted by S(n)（ε） as the following.

 

Where ε represents the errors. 

 

Considering the storage

The generation, transmission, distribution and use of traditional electricity production is almost preceded at the same time, which significantly influenced the planning, construction, scheduling, operation and control of power system. The application of large-capacity storage technology will break the limitation of the real-time power supply and demand balance. The development of storage has become an inevitable trend of energy storage technologies in power system [1.5].

Currently, there are several storage devices. 

 

Pumped storage reserves the pumped water at the upstream during low load hours, and generates electricity using the stored water during the peak load hours. The efficiency is about 75 percent. Limited by location places and construction period, it is a bit difficult for large-scale application. 

 

The flywheel energy storage device combines the motor with the flywheel, which can store energy into a high speed rotation, and transfer into the electricity when necessary. The efficiency can reach 85percent to 90percent[7]. The series of commercial products practice this storage style. Benefited by the rapid response performance, it can be used as small capacity, short discharge time situation. [1]#p#分頁標題#e#

The essence of compressed air energy storage (CASE) is gas turbine power plants, which compress the air and store it in the high-pressure sealed space such as underground lava caves. And the released gas will drive a turbine to generate electricity in the peak time. The limitation is the suitable gas storage place.

The main principle of the superconducting magnetic energy storage (SMES) technology is to store the energy into the superconducting coil in the form of electromagnetic energy. With comprehensive high efficiency of 95 percent, it can be good choice to control the voltage stability and power quality.

 

Super-capacitor energy storage device is similar with the conventional capacitor. After special processing, the super capacitor can have greater electric capacity. But the expensive and the technology limitation, it is only used in the switching station. [1,9]

 

Also known as electrochemical energy storage, battery energy storage device can be divided by the different types of battery. Lead-acid battery is mature and low cost. Considering short life and low energy density, it will be abandoned in the future. The similar nickel-cadmium battery is limited by the pollution and self-discharge phenomenon. The sodium sulfur battery is increased used in recent years because of the long life cycle and high energy density. The flow battery can change the capacity but is limited by the high cost. The lithium battery and nickel-cadmium battery still need to be examined. Although the immature situation, the battery storage method will have more potential in the future.

Due to the several methods, we cannot satisfy the overall storage demand of the power system. The power system requires the storage device to complete the load shifting, frequency adjustment and standby mode, the stability control and power quality adjustment. All of these subjects need the higher storage capacity and fast response speed. However, Most of the devices are used in American, who has the geographic advantages. 

 

In Great British, we must combine the renewable energy style with different storage devices. The wind and solar energy has characteristics of random, intermittent, and fast output variation. When this kind of energy is put into use, we must face with the possible accident caused by the shock of Instability. However, due to the depth analysis in the previous chapter, we must balance the peak and valley period of electric use by storage method. The load shifting and grid stability control can improve power efficiency, improve the load rate, and smooth the price. 

 

Considering the existing power storage device, we can assume that the storage technology has been qualified to the demand of the power system. And we must consider the transformation cost and the purchasing cost generated by the application of the electricity storage method. And the model can be established as optimization problem with the objective function of minimum annual purchasing fee. 

We use the storage technology to achieve the situation of eliminating the line capacity limitation or mitigating network congestion. Suppose some part of the grid has exceeded the load, and it need congestion management device. Now, we install the storage fittings, which can release energy when it is over load, and store energy when it is low load. Then the congestion situation can be eliminated.

In the previous chapter, we have found that the RTP will change due to the influence of the load factors. Generally speaking, the price is higher when the load is high, which will lead to the emergence of the peak price. And the higher price also means higher purchasing cost, which can be treated as the implicit cost of power congestion. 

 

Based on the assumption, the optimization goals are the minimum congestion costs, which are also minimum annual purchase costs and also the minimum output casts. In the formula,  , we consider pt and Pst respectively represent for the price and power during the time period t. And X = { Pes …} is the power vector of energy storage. 

 

We should also make clear the constraints. 

(1) The sum of output power and storage output power must be less than or equal to the corresponding total load.  

(2) The transmission capacity can not exceed the maximum capacity of the transmission grid, and it can not be negative.  

(3) The storage output power can not exceed its maximum capacity and can not release more than its maximum depth of discharge. P est ≤ P esmax; BODt ≤ BODmax.

(4) Considering the efficiency, it must be balance between the storage and release,  .

(5) The energy storage device must charge and dispose every time. And we had better ignore the influence of distance due to the storage constraints. 

Then we can solve the equation under the constraints by using the software.