2.3.1. Billing

The most vital function of the smart meter is providing accurate billing for consumers. Any data protection method which influences the accuracy of billing is useless. The current sample interval of the smart meter is 15 min up to a few seconds; however, consumers do not need such high-frequency billings, weekly or monthly basis billing is enough [21].

#### 2.3.2. Grid Operation and Management

The smart meter contributes to the smart grid by improving the efficiency and stability of the whole power system. The real-time two-way communication networks provided by the smart metering system can measure, analyse, and control the energy consumption data, and further support the smart grid to implement demand side response services and power system estimation. For grid operators, the measurement of every individual household smart meter is not compulsory. Instead, they have more interest in high aggregated level data, such as measurement at feeder level or substation level [22].

## 2.3.3. Value-Added Services

The consumers can order additional services provided by the utility or TP. The additional consumer services could be awareness, (e.g., sending a warning for exceeding power), or scheduling and control (scheduling for controllable appliances, peak shaving) [23]. Demand-side response and non-intrusive load monitoring (NILM) have received the most attention. Demand-side response [21,24,25] optimizes the strength of the grid and enhances the power quality by utilizing power plants, distributed generators, and loads and energy storages. In demand-side response, consumers can also participate in the response process by accepting the bids provided by grid operators. By turning off appliances such as air conditioners and heaters, the load would be shed during peak time. NILM is a technique to disaggregate consumers' power consumption curve into individual appliance usage. The consumer can have an insight into how electricity is consumed and can better manage their home appliances to save energy and reduce carbon dioxide emissions [26]. Normally, value-added services require consumers to submit their energy data to a server; the server would use a pre-trained model to evaluate the data and send the results back to consumers. The difficulty exists in how to share personal data with TP while guaranteeing privacy at the same time.

In [21], two privacy-preserving value-added schemes are proposed. The naive scheme down-samples the original data into multiple interval resolution data, referring to the requirement of different services. Then the different resolution data are sent to different TPs with a key [13]. The second solution is to enable services at consumers' own devices (personal computer, mobile phone) via a home area network (HAN), however, TPs have the risk of revealing their models/algorithms [21].

## 2.3.4. Time-of-Use Tariff

The TOU tariff determines the electricity price during different periods. Consumers benefit from the TOU tariff by shifting their electricity usage habits to enjoy a cheaper bill, while the energy suppliers can also reduce the power plant capacity as a result [14], so TOU is becoming the mainstream method for billing in the UK. With the installation of the smart meter, the TOU moves closer to the real-time pricing tariff, allowing it to better represent the true conditions [27]. Although TOU tariff needs high-frequency data, the price volatility is determined by the aggregated power usage of the

whole area rather than individuals. So, consumers do not have to send high-frequency data to realize the TOU tariff.

#### *2.4. Privacy Intrusion Issues*

Currently, smart metering systems could easily suffer from internal [28] and external attacks [29] and be subject to privacy intrusion [2]. All privacy intrusion issues related to the smart meters fall into two categories:

Category (i) Data sensitivity. Personal energy data that cannot be measured by a conventional electricity meter. While the traditional electricity meter measures the power consumption with a low resolution (e.g., one month) and can only provide the energy consumption information in kWh, the smart meter measures the power consumption with a high frequency (ranging from every second to every half hour, and normally every 15 min [21]), and more parameters are recorded, such as real-time active/reactive power, voltage, current, TOU tariff, etc. The high granularity data provide adversaries enough information to intrude on personal information.

Category (ii) Algorithm sensitivity. Advanced algorithms/mechanisms to intrude on privacy-sensitive features that could not be extracted from raw data using traditional data processing mechanisms. With the implementation of smart meters in smart grids to meet the above functions, and the increasing development of new services and applications by TP based on big data and artificial intelligence (AI) (e.g., Machine Learning (ML), Deep Neural Network (DNN), cloud computing), more and more sophisticated data could become available [30]. New services to better understand and monitor household behaviour include NILM [26], short-term load forecasting (STLF), distributed data mining, and others [14,30]. These advanced techniques are a double-edged sword to the consumers. The benefits espoused to consumers described above (e.g., managing their own energy consumption) and the adoption of a utilitarian ethic (i.e., ensuring the greater, collective good of energy efficiencies in smart grids) need to be weighed against potential privacy intrusion risks. Privacy intrusion would mean that not only individual but also collective freedom is compromised, given that household behaviour would be shaped and constrained by the perceived presence of digital surveillance [10,11,31].

Moreover, referring to the US National Institute of Standards and Technology (NIST) guideline NIST IR 7628v2 [32], the above two categories can be divided into four aspects as follows:

#### 2.4.1. Behaviour Patterns Identification

Behaviour patterns identification belongs to category i; it aims to identify the appliances used. The smart meter and AMI communication network enables the utility and TP to access individual energy data continuously [15]. The high granularity data can reveal information about specific appliances at certain times and locations inside the home. Based on this information, operators can further infer the activities inside the house [32]. Potential usage of the appliance information may include that the retailers would adjust the warranty policy or use the information for advertising and marketing purposes.

## 2.4.2. Real-Time Surveillance

Real-time surveillance means that by regularly accessing energy data via smart meters, power system operators/TPs can have an overall picture of the activities inside a house, and even the entire life cycles of all residents (waking/sleeping pattern, number of residents, when people leave their home). This privacy concern belongs to category ii; the surveillance relies on advanced techniques such as data mining, machine learning/deep learning algorithms [26,30,32]. This information could be abused by hackers and stolen for criminal purpose [33].
