What Factors Affect the Price of a Cable Bill?

The short-term electricity price forecast method of the electricity market ^[1] refers to the weekly electricity price forecast, the daily electricity price forecast, and the hourly electricity price forecast under the electricity market environment. Short-term electricity price forecasting is an important part of the research on electricity price forecasting, and it is also the focus and focus of current research. At present, there are five main methods for short-term electricity price prediction. A single forecasting model cannot accurately predict electricity prices. A combined forecasting model can gather as much useful information as possible, making full use of the advantages of different mathematical methods, which will be the development trend of future electricity price forecasting methods. But we should also realize that not a combination of any two or more mathematical methods will necessarily achieve better prediction results. This requires practice to test their prediction effects in order to evaluate the application of a new mathematical method.

With the continuous development of global power marketization, the power industry has gradually shifted from monopoly to competition. As the core factor of the power market, electricity prices have also changed accordingly. Although the existing and developing power markets have different models and different electricity price formation mechanisms, as a general trend, electricity prices should be time-varying, as well as economic, meteorological, and operating conditions of power systems and users. Related to the situation, electricity prices change with demand, changes in electricity prices affect demand, and the role of electricity price adjustment mechanisms will be more significant. As a participant in the electricity market, its benefits are ultimately realized through the trading of electricity. In electricity trading and bidding, knowing the price of electricity in advance and arranging production plans and bidding strategies in advance can obtain greater benefits. Therefore, electricity price prediction It has become one of the topics to be studied and solved in the electricity market ^[2]

According to the types of forecast points, the electricity price forecast is divided into the system marginal electricity price or the market unified clearing electricity price forecast, the regional marginal electricity price forecast, and the point marginal electricity price forecast. It is often the case that what we call the electricity price prediction are predictions of the system's unified clearing electricity price. In the case of no blockage in the system, the regional clearing power price in each region and the system's unified clearing power price are the same.

According to the different prediction content, it can be divided into deterministic prediction and electricity price spatial distribution prediction. The former is a hot topic that is currently being discussed more, mainly for short-term electricity price prediction. The result of the prediction is to give a certain electricity price prediction value; the latter is mainly based on Knowledge of probability theory and mathematical statistics, to determine the possible fluctuation range of forecast results and the average price of electricity in a period of time, mainly for mid-to-long term electricity price forecasting, and there are few studies in this area at home and abroad recently ^[2]

The time series model is divided into white regression (AR) model, moving average (MA) model, ARMA model, cumulative white regression-moving average (ARMA) model. The premise of the application of these models is based on the time series as stationary random sequences. The electricity price has the characteristics of non-stationary random time series. The existing methods cannot meet the requirements of short-term electricity price prediction. Therefore, based on the existing electricity price prediction methods, iterative prediction methods that incorporate error prediction in conventional electricity price prediction models to improve accuracy: firstly establish a simple and reasonable prediction model, and then analyze the formed error sequence to establish an error prediction model. The complexity of establishing a prediction model for this sequence is lower than that of the original prediction model, but the prediction accuracy is easier to improve. The method proposed in this document is general and can push down to hourly electricity price forecasting and load forecasting and other forecasting fields. However, due to the relatively strong random volatility of electricity prices, it is generally difficult to effectively extract the non-stationary process of the electricity price time series, which greatly affects the prediction effect and makes the time series method have little advantage in the field of electricity price prediction. Of course, if the sequence can be well stabilized, the time series method can also achieve better results ^[1]

The main idea of artificial neural network applied to electricity price prediction is to use neural network to find the implicit trend and regularity from historical electricity price data, so as to achieve a better prediction effect. Artificial neural networks are less commonly used in forecasting, mainly due to the following reasons: artificial neural networks require a large amount of historical data, which can be provided by the power system; electricity price prediction is a non-linear problem that requires multiple related input , Is not conducive to the establishment of mathematical models, and artificial neural networks have strong generalization capabilities, do not need to establish mathematical models, only need to hand over the existing data to the network, the network will choose Baiji's model through training, generally can Solve the problem well; electricity price data is often accompanied by a lot of noise, and artificial neural networks are more tolerant of noise interference than other methods ^[1]

Wavelet analysis is an emerging signal processing method developed on the basis of modern harmonic analysis. It has the functions of scaling, panning, and zooming. It can perform multi-scale analysis on the information, effectively extract the required information from the signal, and realize Time Domain

High-resolution localization in the frequency and frequency domains is particularly suitable for analyzing sudden and short-term information. It is often referred to as a "mathematical microscope".

The proposed neural network short-term spot electricity price prediction method based on multi-factor wavelet analysis decomposes the historical electricity price sequence and the load sequence into wavelet decomposition, and decomposes them into an overview electricity price and a detailed electricity price, an overview load and a detailed load. Then the profile electricity price and the corresponding profile load are input to the neural network to predict the future profile electricity price; similarly, the detailed electricity price and the corresponding detail load are input to the neural network to predict the future detailed electricity price. Finally, sum up the predicted electricity price and the detailed electricity price to get the final predicted electricity price. The use of wavelet analysis technology can make the power price change rules and hierarchical characteristics more clear, and at the same time make the decomposed load more reflect the fluctuation of the degraded power price; the constructed neural network short-term spot price prediction model based on multi-factor wavelet analysis can Significantly reduce prediction errors and improve prediction accuracy ^[1]

The main feature of the gray model is that it does not require any probability distribution of the original sequence, and it can realize less data modeling.

Some improvements have been made on the basis of the GM (1, 1) model. The initial value of the differential equation is used as the initial condition, the prediction error is reduced, and the original data is smoothed by the weighted smoothing method. Based on the GM (1, 1) model, the influence of the electricity price one hour before the forecast time on the current electricity price was designed, and the GM (1, 2) model was obtained. The prediction accuracy is improved based on the GM (1, 1) model. However, this model is only suitable for electricity price prediction in markets with little historical data and relatively smooth fluctuations in electricity price sequences.

When predicting a complex system, various prediction models are established from different perspectives, and then based on these individual prediction models, a coordination model different from these models is established to achieve the effect of merits. This is the idea of combined prediction . The method of combining multiple prediction models to establish a coordination model is called a combination method, and the established coordination model is called a combination prediction model. The combined prediction model can be used to organically combine the models, and the advantages of each model can be combined to effectively improve the model's fitting ability and prediction accuracy under certain conditions.

A few days ago, the combination prediction methods proposed at home and abroad mainly include the following: one is the combination prediction method with fixed weights; the other is the combination prediction method with variable weights. The main idea of these two combined prediction methods is to combine the prediction results of various methods to finally obtain an optimal prediction result. The key is to obtain the optimal weight. These two methods are used more in power load forecasting, but they are less used in power price forecasting. On the one hand, the huge fluctuation of power prices leads to the optimal weighting.

It is very difficult to obtain. On the other hand, because the current electricity price prediction accuracy obtained by various methods is not high, it is very likely that the accuracy of the final electricity price prediction will be reduced by combining these prediction results again.

Therefore, the main combined forecasting methods in the day-to-day electricity price forecasting model are: (1) According to the characteristics of the historical data of the research object, select a suitable model method to separate the different components of the data, and then make separate predictions, and combine the prediction results The final prediction result; (2) Considering that each method has its advantages and disadvantages, a complementary method is used to combine two or more methods into a new method to predict ^[1]

Through the above analysis, we can know that a single forecasting model cannot accurately predict electricity prices. A combined forecasting model can gather as much useful information as possible, making full use of the advantages of different mathematical methods, and it will be the development trend of future electricity price forecasting methods. . But we should also realize that not a combination of any two or more mathematical methods will definitely achieve better prediction results. This requires practice to test their prediction effects in order to evaluate the application of a new mathematical method.